The Decision Nobody Owns
Haute Lumière · Who Decides · Chapter One
Chapter One
A decision is often waiting not for an answer but for someone to know whose answer it is.
The Decision Nobody Owns
Haute Lumière · Who Decides · Chapter One
It is twenty-five past four on a Wednesday, and the third meeting about the delivery days is ending the way the first two ended. The head of commercial at a regional bakery chain has a printed agenda in front of her, and beside the item, in the operations manager's hand, the words for decision have been crossed out and for discussion written above them. Through the window behind the finance partner she can see the loading bay, where two of the vans are being hosed down in the low October sun. The proposal is simple. Forty shops are supplied six days a week, and the figures say five would do, with a larger Saturday load. Everyone in the room has said, in one form or another, that it makes sense.
The last slide says Next steps: align with stakeholders. It said the same in June.
Her jaw is tight in a way she has noticed before, at this point in this meeting. She does not know whether she is allowed to say we are doing it, and she suspects the operations manager does not know whether he is allowed to say it either, and that the finance partner thinks it is a matter for the two of them. She is a composite, drawn from the way these meetings are described in the research below and in the house's own practice, and no firm is meant. The vans will go out six days next week, as they did this week, because nobody has decided otherwise.
Ammanuel Santa Anna's canon has a passage on authority that reads as if it had been written about the agenda in her hand. It begins by separating authority from the chart that is supposed to show it:
The CEO who issues a directive that everyone ignores has a title but no authority. The frontline worker whose informal recommendations are consistently followed by the team has no title but genuine authority. The map is not the territory, and the org chart is not the authority structure.1
This chapter makes one claim. Much of what a firm calls slow decision-making is not slow thinking at all. It is a decision that nobody owns: everyone has been consulted, nobody has been named, and so the question returns to the agenda each month, a little more familiar and no nearer to being settled. The unowned decision is the first place a firm stops deciding, and because the remedy costs one sentence, it is also the cheapest place to begin.
In 1972 three organisational theorists, Michael Cohen, James March and Johan Olsen, published a paper with a title that has outlived most of its readers' memory of its content: A Garbage Can Model of Organizational Choice.2 They had been studying universities, which they took as a familiar kind of what they called organised anarchy: places with problematic preferences, unclear technology and fluid participation. In such organisations, their abstract says, decisions can be viewed as "collections of choices looking for problems, issues and feelings looking for decision situations in which they might be aired, solutions looking for issues to which they might be an answer, and decision makers looking for work."
The sentence is funny, and it is also exact. The bakery's meeting is a decision situation in which several issues are being aired at once. The operations manager has a feeling about the drivers' rotas. The finance partner has a solution, a cheaper van lease, looking for an issue. The area managers have a long-standing grievance about Saturday stock. Each of them is attending to the question that brought them into the room, and the question on the agenda, whether to move to five days, is only one of the things the meeting is being used for. A meeting with no owner becomes a place where every problem can be parked and none has to be closed.
A word on the strength of this evidence. The garbage can model is a computer simulation built from observation of universities, and its authors offered it as a way of seeing, not a measurement of how often firms behave this way. What it shows well is the mechanism: when participation is fluid and nobody is charged with a particular choice, the timing of a decision depends on who happens to be in the room and what else is on their minds. What it cannot show is how much of any real firm's calendar works like this. That is a question each organisation can answer about itself, by counting.
Santa Anna's canon names the pattern from the inside, in a list of the dynamics that everyone in an organisation recognises and nobody says aloud. One of them is simply the bakery's agenda across three meetings: "decisions that are made, unmade, and remade in cycles that exhaust the system." The delivery days have been agreed three times. They have not been decided once.
The crossing-out on the agenda is a small act, and the canon's book on decision-making would treat it as the largest one in the room. A decision begins, it argues, before anyone gives an opinion, at the moment someone frames what is being decided, and "The person or group that controls the frame controls the decision — often without anyone realizing it." When for decision became for discussion, the operations manager did not mean to take anything. He meant to be careful; the rotas were not ready and he did not want to be pressed. But the frame he wrote told nine people that nothing would be settled that afternoon, and so nothing was. The canon adds the reason nobody objected, in a sentence about how frames are usually set: "Most framing happens unconsciously, shaped by organizational culture, professional training, and the cognitive habits of whoever happens to speak first." He spoke first. He usually does.
The right to decide, and the power
Haute Lumière · Who Decides · Chapter One
There is a distinction in economics that makes the bakery's difficulty easier to see. In 1997 Philippe Aghion and Jean Tirole published a theory of what they called formal and real authority in organisations.3 Formal authority, in their abstract's words, is "the right to decide". Real authority is "the effective control over decisions". The two can sit with different people, and the paper's central point is that "Real authority is determined by the structure of information": whoever knows enough to make a proposal that the formal holder will accept without checking holds the decision in practice, whatever the chart says.
Read the third meeting through that distinction. The head of commercial probably holds the formal authority; the delivery schedule is a commercial matter, and the managing director gave her the review. But the operations manager holds the van rota, the drivers' contracts and the knowledge of which shops can take a larger Saturday load, and without his information her decision would be a guess. He holds a large share of the real authority. Neither of them has said so, and each is waiting for the other to move. The finance partner, who holds neither, is the only person in the room who is sure the decision is not hers.
Aghion and Tirole add an observation that will matter later in this book. Among the factors their abstract lists as increasing a subordinate's real authority is "overload": a superior with too much on her desk accepts more proposals without checking them. A decision that is never formally owned is often quietly made by whoever is least overloaded, by default, and nobody is accountable for it, because nobody was asked.
The canon puts the same point in the language of design. Authority, it says, belongs to a domain of work rather than a person or a title, and a well-drawn domain is specific: "It says what authority is held and, by implication, what authority is not held." The bakery has never drawn the domain around the delivery schedule. It has drawn a meeting around it instead.
Santa Anna is also careful not to let the remedy turn into its own kind of evasion. Writing about firms that distribute authority without drawing domains, the canon describes the result in two sentences: "If a decision goes wrong, who is responsible? Everyone and no one." The bakery is not a flat organisation, and nobody there has read about self-management. It has arrived at the same place by a more ordinary road, through courtesy. Each person in the room is being careful not to assume an authority the others might think is theirs, and the sum of nine courtesies is a decision that belongs to no one.
The best-known practical answer to this problem came from a consultancy. In January 2006 Paul Rogers and Marcia Blenko of Bain & Company published an article in the Harvard Business Review called Who Has the D?4 Its opening observation is the bakery's agenda in a sentence: "At many companies, decisions routinely get stuck inside the organization like loose change." Their remedy was a set of five roles, now taught under the acronym RAPID. Bain's own current description gives them plainly. The person who recommends "drives the process, gathering relevant input and developing a recommendation for the decider." Those with input provide "expertise, experience, or information that shapes the recommendation." A small number must agree, which in the framework is a real veto. Some will perform the decision once it is made. And one person decides, the D, who "makes the final decision and commits the organization to action." Bain's page adds the sentence the bakery most needs: "Ideally, there should only be one decider for each decision."5
The grade of this evidence should be stated before it is used. RAPID is a practitioner's framework, drawn from a consultancy's engagements, and the article is an argument from cases, not a study. Rogers and Blenko, with Michael Mankins, later published a book, Decide & Deliver (2010), which reports survey work linking how well firms decide with how well they perform; it is survey evidence gathered by a firm that sells the remedy, and it is best read as a well-informed opinion.6 The framework's strength does not depend on those figures. It depends on something anyone can check in their own building: whether, for a given decision, the people in the room could each name the person who will make it.
What the framework does well, and what the bakery's meetings lack, is the separation of being heard from deciding. In the third meeting everyone gave input, and because input was the only role anyone had, everyone's input carried the weight of a vote. The area managers' Saturday grievance had the same standing as the operations manager's rota, and the finance partner's lease had the same standing as both. Once a person is named as the one who decides, the others become something more useful than a committee. They become witnesses, each with a piece of the picture, each heard, and none of them responsible for closing the question.
It also changes what the decider owes the others, and this is the part of the framework least often quoted. A person who knows she will decide has a reason to hear the input fully, because she will be the one answering for the result. A person who believes the decision is collective has a reason to hear it politely, because nothing will rest on her. The difference shows in the questions asked. In the third meeting nobody asked the operations manager what the rota change would do to the two drivers whose contracts guaranteed six days, because nobody needed to know. The person deciding would have asked it first.
The framework's commonest misuse is just as easy to see. Give too many people the right to agree and the veto spreads until the decision needs a unanimous room; name a committee as the decider and the loose change is back in the lining. The house's Luminous 100, its reading of the frameworks every consultancy sells, describes Holacracy, the most radical of the self-management systems, in a phrase that holds the balance RAPID is reaching for: "nobody needs permission, everybody has explicit accountabilities." The second half is the part most firms skip. Permission is easy to grant in a slide. An accountability someone can name is the thing that moves the vans.
It is tempting to treat the unowned decision as a matter of manners, and to answer it with a reminder about meeting discipline. It is more expensive than that, and most of the expense never appears on any report the bakery produces.
There is the plain cost first. Three meetings of nine people, ninety minutes each, is a working week of senior time spent arriving at the same agreement three times. The vans still run six days. If the figures are right, every week of the sixth day is fuel, driver hours and waste the bakery has already decided, in principle, not to spend. None of this needs a consultant's estimate to be real, and none of it should be given one here: the bakery's own fuel cards and rotas would show it, and they are the only source that matters to the bakery.
Then there is the cost the canon is more interested in, because it lasts longer. Santa Anna's book on management records a software engineer describing his first manager: "He wanted compliance wrapped in the language of collaboration. He'd ask for input, then do what he'd already decided." The bakery's failure is the mirror of that one. Its people are asked for input and then nothing is decided at all. Both teach the same lesson, which is that giving input does not change anything, and both produce the same quiet withdrawal. By the fourth meeting, the area managers will stop bringing the Saturday figures, because bringing them has made no difference three times.
The canon gives the reason the unowned decision survives so long, and it is not laziness. "Decisions get delayed because no one wants to be wrong and vulnerability is dangerous." To own a decision is to be the person named when it goes wrong. In a firm where a wrong decision is remembered longer than a slow one, the rational course for every person in the room is to keep the decision collective, which is to say unmade. The head of commercial's jaw knows this before she does. Nobody has told her she would be blamed if Saturdays went badly. Nobody has told her she would not.
The cost does not stay in the room. In the forty shops, the managers have heard that the delivery days are being reviewed, because news of that kind always travels faster than decisions. Some have stopped ordering for Saturday the way they used to, in anticipation of a change that has not come. One has told her staff that the Monday delivery is ending, because an area manager said so in March. The canon's governance chapter has a line about authority that describes the shops' position exactly: "Authority that is not recognized is not authority — it is merely an assertion of will." In the shops, rumour has more authority than the head office meetings, because rumour at least says something.
That is also why the remedy is not exhortation. Asking people to take more ownership, in a room where nobody has been given the decision, asks each of them to seize an authority they are not sure they hold. The Margin, another of the house's books, shows how [deciding that a piece of work is finished is itself a decision somebody has to own](https://hautelumiere.life/editions/rsd-master-rsd-in-acedemia-final-manuscriptt/page/), and the same holds at the scale of a schedule. Ownership is not a trait people lack. It is a sentence nobody has said.
Naming the owner
Haute Lumière · Who Decides · Chapter One
What changes this is a single line at the top of a page, and the conversation it takes to write it.
Picture the head of commercial the following Monday, who has decided to do one thing differently. Before the fourth meeting is scheduled, she asks the managing director for fifteen minutes and a straight answer to one question: whose decision is the delivery schedule? The managing director is surprised to be asked. In her mind it was always the head of commercial's, and she had been wondering why it was taking so long. That is the first discovery most people make when they ask: the person above them thought the matter was settled, and has been mildly puzzled by the delay. The second discovery comes when she asks the operations manager the same question. He thought it was the managing director's.
She then writes, at the top of the paper for the next meeting, one line in four parts. Decision: whether to move the forty shops from six deliveries a week to five, with a larger Saturday load. Decided by: the head of commercial. By: the meeting on the fourteenth. Heard first: operations, on rotas and drivers; the area managers, on Saturday stock; finance, on cost. Beneath it a second line names the one person who must agree, the food safety lead, because a larger Saturday load changes how long bread sits in a shop's back room, and that is a matter of law, not opinion.
The order of what she did matters more than the line itself. She asked upward before she asked across, so the answer carried the managing director's authority and nobody had to take her word for it. She named herself, which cost her something, rather than proposing a working group, which would have cost nobody anything and decided nothing. She gave the one veto to the person whose objection would be about harm, not preference. And she put a date on it, because a decision without a date is a topic. The canon's chapter on governance describes the same moment in a different setting, a team acknowledging who holds a domain, and it is candid about what kind of act it is: "This may sound ceremonial. It is. And ceremonies matter." The line at the top of the page is a small ceremony. It tells the room, before anyone speaks, what their speaking is for.
The fourth meeting is different from the first three in one respect, and it is the respect that matters. The operations manager, who knows he is being heard rather than asked to decide, gives the rota figures in full, including the awkward one about the two drivers whose contracts guarantee six days. The area managers bring the Saturday stock figures for the first time since June. The finance partner asks a question about the lease and then, noticing that nobody needs her to settle anything, drops it. The head of commercial decides, in the room, with a change: five days, from the first of the month after next, the two drivers moved to the new Saturday run.
There are predictable ways this goes wrong, and Chapter 7 carries them with the run sheet, in the order they tend to appear. The commonest is the owner who, once named, goes back to consulting, because consulting feels safer than deciding. The second is the senior person who names an owner and then reopens the decision when it lands on her desk as a complaint. The third is the line written for every trivial matter, until the ceremony becomes paperwork. The canon's warning about the misuse of its own language applies here with some force: "When leaders invoke "self-organization" to avoid making difficult decisions, that is not living-systems thinking — it is abdication." Naming an owner is not a way for a leader to step back from a decision. It is a way to make sure someone is standing in it.
Naming the owner is only the first place a firm's decisions wait. The second is the queue that forms when a decision is sent upward for a signature (Chapter 2), and the third, subtler, is the room that has an owner and still cannot close (Chapter 3).
The chapter ended with one decision owned. The book can go further, into the finer points a leader meets when she tries this across her own organisation, and the first is knowing which of her decisions are unowned at all.
There are signs, and none of them needs a survey. The same item appears on three agendas in a row. Papers are marked for discussion when everyone knows a choice is being asked for. Minutes record that something was agreed and nobody can say who will act on it. People ask, in the corridor, who is signing this off? A proposal is sent to a working group that has no date to report. A decision is described by its topic, the delivery schedule, rather than by its choice, whether to move to five days. Any one of these can happen in a well-run firm. Three or four together are the garbage can at work.
The second finer point is the wording of the line itself. A decision is a sentence with a verb and a choice in it: whether to, which of, how many. A topic is not a decision, and a line that names a topic will produce a discussion. The test is simple. Could someone say yes or no to the line, or choose among the options it names? If not, it is not yet ready for an owner.
The third is the right to agree, which is where the framework is most often spoiled. Most of the people who matter to a decision should be heard. Very few should be able to stop it. A useful rule is that the right to agree belongs only to those whose objection would rest on law, on a written policy, or on a harm they are responsible for preventing: the food safety lead, the person accountable for a regulator, the owner of a budget the decision would spend. Everyone else gives input, and their input is weighed, and the owner says how. A line that gives five people the right to agree has rebuilt the committee.
The fourth is the direction in which to ask. The head of commercial asked upward first, and that is usually right. A person who names herself as owner without being sure she is entitled to has made a claim the others may resist. A person who asks the person above whose decision it is, and is told, has a fact. When the answer comes, write it down where the people who will be affected can see it, so that nobody has to take a corridor's word for it.
The fifth is the decision that seems too large for one person. It is common to answer this by naming a committee, and it is almost always a mistake. A large decision can have many people who recommend, many who are heard and several who must agree. It still needs one person whose job is to close it on a date. If the decision is so large that no single person below the chief executive can be trusted with it, the owner is the chief executive, and the others recommend.
The sixth is what the owner owes. She owes the people heard a real hearing before the date, and a sentence afterwards saying what she decided and why their input did or did not change it. She owes the people who will carry it out the decision in plain words, with the date it takes effect. And she owes everyone a rule for reopening: a decision is reopened when new information arrives that was not available on the day, and not when someone who lost the argument raises it again. Without that rule, every owned decision drifts back into the garbage can.
And the decision that most changes the result is what the senior person does the first time an owned decision reaches her as a complaint. If she reopens it herself, the line at the top of the page means nothing from that day. If she sends the complaint to the owner and asks what new information it contains, the line holds. Chapter 6 places the naming of owners in the first weeks of a ninety-day plan, and Chapter 7 gives the signs of an unowned decision and the decision line on a page each.
The fourth meeting ends at twenty-five past four, as the others did, but it began at ten to and it is over. On the agenda the words for decision are printed, and nobody has crossed them out. The last slide says Decided: five days from the first of December. Owner: head of commercial. Review: the second week of January.
Through the window the vans are being hosed down in the bay, the same two, in the same low sun. The head of commercial puts the agenda in her folder and notices that her jaw is where jaws usually are.
On Thursday morning the drivers' rota for December goes up on the board by the bay, with five days on it.
Chapter Two
A decision sent upward waits in a queue like any other work, and usually arrives where less is known about it.
Three Signatures for a Pump
Haute Lumière · Who Decides · Chapter Two
It is ten past seven on a Tuesday, and the shift engineer at a water treatment works is standing in the pump house with her hand flat on the casing of the number two dosing pump. It is vibrating in a way it was not vibrating a month ago. The light through the high windows is the thin grey of early November, and the hum of the room is slightly off, the way a familiar voice sounds with a cold. In her other hand is a printed requisition for a replacement, raised nine days ago. At the bottom of the page are three boxes for signatures. The first, the maintenance manager's, is signed. The second, the site manager's, is signed. The third belongs to the regional operations director, who covers four sites and is at the other end of the region this week.
She is a composite, drawn from the way such mornings are described in the research below and in the house's own practice, and no works is meant. She knows this pump better than anyone who will sign for it. She has replaced its seals twice, she keeps the log of its flow readings, and she can tell from the sound alone that it has days rather than weeks. Her shoulders have been up since Friday, when she rang the regional office and was told the requisition was in the queue.
On the following Monday the pump fails. A hired unit is brought in that afternoon at a weekly rate which, over the month it takes the new pump to arrive, costs the works more than the new pump did. The requisition is signed on the Wednesday, by a director who approved it in under a minute, because there was nothing in it she could have judged better than the engineer already had.
Ammanuel Santa Anna's canon, writing on the pathologies of concentrated authority, describes the requisition's nine days in a passage that could be pinned above the regional office:
When all significant decisions must route through a small number of positions, those positions become bottlenecks. Decisions queue. Response times lengthen. Opportunities expire while waiting for approval. The organization operates at the speed of its most overloaded decision-maker rather than at the speed the situation demands.7
This chapter makes one claim. A decision sent up a hierarchy waits in a queue, exactly as work does, and it is usually sent up not because the person above knows more but because the person below has never been allowed to decide. Where a decision belongs is a question with an answer, and for most of the decisions a firm makes every week, the answer is lower than the chain it travels.
In 1976 Henry Mintzberg, Duru Raisinghani and André Théorêt published a study of decisions that nobody had a procedure for.8 They followed twenty-five strategic decision processes in organisations, from the moment a need was first recognised to the moment a choice was authorised, and they described the structure that such apparently unstructured processes share: phases of identifying the problem, developing options and selecting among them, and around those phases a set of interruptions, delays and returns to earlier steps. The decisions they studied were large ones, and most took a long time. What the study made visible was how much of that time was not spent deciding. It was spent waiting for something: a person, an approval, a moment when the matter could be raised.
The study is a set of field cases, rich and carefully coded, and its text was not reopened for this book; its phases are given as the decision literature has long summarised them. It should be read as a description of how strategic decisions move, not as a measure of how long a requisition takes. But the requisition has the same anatomy at a smaller scale. Recognition took a morning, when the engineer heard the pump. Development took an hour, when she priced a replacement. Selection took nobody any time at all. Authorisation took nine days.
The arithmetic of that waiting is already familiar from operations. A director's inbox is a queue like any other, and Where the Flow Stops, the house's first book on organisations, shows how [an order holding fifty minutes of work can spend six days inside a building]({{BOOK:orgdev-flow}}) for no reason except the number of things waiting beside it. A requisition on the desk of a person who covers four sites waits behind every other requisition from four sites. Nobody is slow. The director is busy, competent and conscientious, and the queue is long because everything is routed to her.
It is worth looking at the queue from her side of the desk, because from there it does not look like a queue at all. On a Monday morning she has something over forty items waiting for her signature from four sites: requisitions, overtime claims, a contractor's variation, a change to a shift pattern. Each is short, each is reasonable and each takes a minute to read. What takes the time is not reading them. It is the travelling, the meetings at each site, the calls, the incidents that come first, so that the minute for each item arrives days after the item does. Chapter 1 drew on Aghion and Tirole's observation that overload hands real authority to whoever proposes; the director's week shows the other half of the same arithmetic. Her signature has become formal authority with almost no real authority left in it, and it costs nine days to collect.
Santa Anna's book on resources describes what the chain does to a firm over time, in a sentence about a company that grew quickly: "Equipment requests languished in approval queues while consultants were retained at premium rates." The pump is the same story at the scale of one works. The money spent on the hired unit was spent precisely because the money for the new one waited for a signature.
Who knows about pumps
Haute Lumière · Who Decides · Chapter Two
It would be easy to conclude that the approval chain should simply be abolished. The economics of hierarchy suggests something more useful than that.
In 2000 the economist Luis Garicano published a model of how organisations arrange knowledge.9 His abstract describes what he called a knowledge-based hierarchy, "a natural way to organize the acquisition of knowledge when matching problems with those who know how to solve them is costly." In such an organisation, "production workers acquire knowledge about the most common or easiest problems confronted, and specialized problem solvers deal with the more exceptional or harder problems." Hierarchy, on this reading, is not a ladder of importance. It is a referral system. A problem goes up when the person facing it cannot solve it, to someone who has seen more of its kind.
Read the requisition through Garicano's model and the fault is plain. A worn dosing pump is not an exceptional problem at a water treatment works. It is one of the commonest problems on the site, and the person who knows most about it is the engineer with her hand on the casing. The chain routes it upward as if it were rare, to a person who knows less about it, who then approves it without being able to add anything. The hierarchy is doing the opposite of what Garicano's hierarchy is for. It is sending the easy problem up and keeping the person who could solve it waiting.
The canon asks the same question in plainer words, and it puts it before every decision rather than after the fact: "Where does the intelligence for this decision actually live?" Its answer for most operational decisions is the one Garicano's model implies. "Routing it upward through management layers does not add intelligence. It adds delay, distortion, and the demotivating message that frontline judgment is not trusted."
The canon is careful not to make a creed of this, and it should not be read as one. Some decisions need a wider view: a choice that commits two sites, a risk the region would carry, a change to what the works is for. Those belong higher, because the knowledge they need lives higher. And sometimes a decision has to go up because the person facing it is genuinely stuck. The canon treats that kind of escalation as a strength of a living system, not a failure of it: "This is not a failure of distribution. It is a feature of a well-designed system." The fault is not that the chain exists. It is that the commonest decisions on the site travel it.
If the engineer knows most, why is she not allowed to decide? The research suggests that the answer has less to do with her than with the person above her, and with what that person believes.
In 1986 Carrie Leana studied delegation in nineteen branch offices of a large insurance company, among forty-four supervisors and the hundred and ninety-eight claims adjusters who reported to them.10 She was asking what predicted whether a supervisor delegated, and what followed when one did. Her full text was not reopened for this book, and its findings are given here only in the shape the delegation literature has long cited: supervisors delegated more where they judged their staff competent and trustworthy, and where their own workload was heavy. It is one study, in one company, three decades old. Its value is that it locates the decision to delegate where the canon locates it, inside the supervisor's view of the person below.
A much larger study found the same pattern across whole economies. In 2012 Nicholas Bloom, Raffaella Sadun and John Van Reenen collected data on the decentralisation of investment, hiring, production and sales decisions from headquarters to plant managers "in almost 4,000 firms in the United States, Europe, and Asia."11 They found, in their abstract's words, that "firms headquartered in high-trust regions are significantly more likely to decentralize," and they offered evidence that trust raises productivity in part because chief executives "can decentralize more decisions." The study measures trust at the level of regions and countries, not of one director and one engineer, and its causal claims rest on careful but indirect methods. What it suggests, at a scale no single firm could reach, is that the height at which decisions are made follows trust, and that the cost of low trust is paid partly in decisions that travel too far.
The canon names the trap this creates, and it is unusually blunt. A firm that keeps decisions at the top tells itself it does so because people below are not ready, and the canon's answer is that they are not ready because they have never been allowed to decide: "Over time, hierarchical authority produces exactly the incompetence it uses to justify its own existence". The engineer has never signed for a pump. She will never be trusted to sign for one until she has.
Trust, in a works, is not built from conversation. It is built from records: the maintenance log the engineer keeps, the flow readings she takes, the history of the seals she has replaced and when. The evidence the director would need to trust the engineer's judgement already exists, written in the engineer's own hand, in a folder in the pump house. Nobody has ever been asked to read it for that purpose. The canon's description of what travelling up a hierarchy does to information is relevant here, because it runs in both directions: "By the time a frontline reality reaches the C-suite, it may be unrecognizable." The engineer's competence is a frontline reality. By the time it reaches the regional office it has become three boxes on a form.
It is worth staying with the person at the top of the chain, because the chain is usually not her design, and she is usually the one person who can shorten it.
The regional director would say, if asked, that she would be glad to sign fewer requisitions. She means it. And yet the canon is right that something happens in a person when a decision she has held passes to someone else, and its book on management describes it with a precision that most leaders will recognise in their own bodies:
When you delegate a high-stakes project to someone you're not sure is ready, something happens in your body. Your chest might tighten. Your jaw might clench. Your breath might shallow. There's a buzz of anxiety in your belly — the nervous system's ancient warning that control is slipping, that the outcome is now in someone else's hands.
The same book follows the sensation to its usual result. A manager who acts from it takes the decision back, adds a check, asks to be copied, and the person below learns that the authority was lent, not given. "The team that's always rescued never learns to rescue itself." The canon tells the case of a founder who knew all of this, had read the books on delegation and hired strong leaders for the purpose, and found that it did not change what happened on her desk: "And still, somehow, the decisions kept flowing back to her desk."
Santa Anna's book on resources adds the uncomfortable half of the explanation, and it is uncomfortable because it is kind. The person who holds an approval, it says, "experiences each act of gatekeeping as a micro-dose of safety." Each signature is a small confirmation that nothing on the site will go wrong without her knowing. The book's verdict on this is not a reproach: "This is not malice. It is survival behavior wearing a business suit." A director who understands why the third box exists in her own nervous system is better placed to remove it than one who believes it exists only on the form.
Moving the decision down
Haute Lumière · Who Decides · Chapter Two
What changes this is not a new policy issued from the regional office. It is a conversation, held once, about a stack of paper.
Picture the regional director and the site manager three weeks after the pump failed, at the table in the site office, with the last twenty requisitions and decisions that went up the chain from the works. The director has asked for them on purpose; she wants to see what she has been signing. They take each one and ask two questions of it. Who, on the day it was raised, knew most about whether it was right? And what did the signature above that person add? For most of the twenty the answers are the same. The person who raised it knew most. The signature added a date. For three of them the answers are different: a choice of contractor that would commit two sites, a change to the chemical dosing regime that the regulator would want to see, a capital item large enough that the region would carry the risk. Those three belonged above the site, and the conversation says so.
Then they agree what changes, and the canon's book on governance supplies the shape. It describes authority being moved in steps rather than all at once: "starting with advisory authority ("Make a recommendation and implement it unless someone objects"), moving to provisional authority ("Make the decision and let us know — we'll review for three months"), and arriving at full domain authority ("This is your domain. You decide.")." For the commonest decisions on the site, the replacement of equipment that has failed or is failing, the works moves straight to the second step. The engineer decides, records the decision and its reason in the same log as before, and the director is told the same day. The canon is exact about what that last step is for: "Not for approval — the domain holder has authority — but for coherence." The director needs to know what has been decided across her four sites. She no longer needs to decide it.
The conversation has a second effect, which the director notices only afterwards. Going through the twenty, she finds four decisions she had signed without any memory of signing them, and one she would have refused if she had read it closely, a contractor's variation that should have gone out to tender. The chain had not been protecting the works from bad decisions. It had been giving her the feeling of protection while her attention went elsewhere. Moving the commonest decisions down returns her attention to the three kinds that need it, and those three now get more of her time than they did when they arrived in a pile of forty.
Three further habits make the change hold. Every decision that does go up carries a date it was raised and a date it was decided, so that the waiting is visible to both ends of the chain. Every decision that went up and did not need to is noted, without blame, at the director's monthly visit, so the line can be redrawn. And the director does one thing that costs her more than it seems: when the engineer makes a decision she would have made differently, she says so in a sentence and lets it stand. That is the moment the authority stops being lent.
The engineer has her own adjustment to make, and the canon is honest that it is not trivial. "Many people in organizations have been thoroughly trained — by years of hierarchical experience — to defer, to wait for permission, to check with the boss before acting. Being told "You have authority in this domain" does not immediately undo that conditioning." The first time she orders a pump on her own judgement she will check the order twice and ring the site manager to tell him, which is not required, and the site manager should say, in a sentence, that it was not needed and that he is glad to know. The second time she will check it once. Santa Anna's book on management puts the stakes of these small moments plainly: "Every delegation decision is a growth opportunity or a growth suppression." The works is not only shortening a queue. It is growing an engineer who decides.
There are predictable ways this goes wrong, and Chapter 7 carries them with the run sheet. The director keeps a shadow approval, asking to see every decision before it is acted on. The site manager, newly trusted, becomes the new bottleneck himself. The class of decisions moved down is drawn so narrowly that nothing real moves. The limits that remain are the organisation's own to set, and they are set by the people who carry the risk; this issue does not suggest any. What it can say is that the conversation about the stack of paper is usually the first time anyone has looked at what the chain is for.
The gain in this is real and it can be found without inventing its size. Every decision that no longer waits is equipment that does not fail in the queue, hire that is not paid, time an engineer does not spend on the telephone to a regional office. Those costs are in the works' own records, in the hire invoices and the maintenance log, and that is where they should be counted. The pump is a single case; what a firm's late decisions cost across a quarter, and how to count it honestly (Chapter 5), is a question with an answer of its own.
The chapter followed one works through one conversation. The book can go further, into the finer points of running the same sort across a whole organisation.
Begin with the right stack. Take the last twenty or thirty decisions that travelled upward from one site, team or function: requisitions, approvals, variations, exceptions, anything that needed a signature or a meeting above the person who raised it. Twenty is enough to see the pattern and few enough to read in an afternoon. Choose ordinary ones. The decision everyone remembers, because it went badly, will mislead, for the same reason an extreme quarter misleads a measure.
Then put two questions to each, and write the answers down. Who, on the day it was raised, knew most about whether it was right? And what did the approval above that person add: information, a wider view, the carrying of a risk, or only a date? Sort the stack into three piles by the answers. The first pile is decisions where the person who raised it knew most and the approval added a date. Those belong at the level where they were raised. The second pile is decisions where the approval added something real, a view across sites, a risk the wider organisation carries, a regulator's interest. Those belong above, and the chapter's three examples are typical. The third pile is decisions that went up because the person who raised them was stuck: unsure of the rules, caught between two departments, short of a fact. Those need an escalation path, which is a different thing from an approval. An escalation is a request for help with a hard case. An approval is a toll on an easy one.
The second finer point is to move authority in steps, as the canon describes, and to name the step for each class of decision. Some will move straight to decide and tell. Some will begin at recommend and act unless someone objects within two days. A few will move to full authority only after a quarter at the earlier step. Write down which step each class is on, and the date on which it will be reviewed. The canon's provisional step carries its own review period, and the review is what makes the step safe to take.
The third is what is told, to whom and when. A decision moved down still needs to be visible, for the reason the canon gives: coherence, not approval. The simplest form is the log the works already kept, with two added columns, decided by and told, and the rule that the person above reads it weekly. If she finds something she would have decided differently, she says so to the person who decided it, in a sentence, and the decision stands unless it breaks a law or a policy.
The fourth is the limits. Every class of decision moved down will have edges: a size, a risk or a kind of commitment beyond which it goes up again. Those edges are the organisation's own, set by the people who carry the risk and written where everyone can see them. This book does not suggest any. What it can say is that an edge set very low has moved nothing, and that the way to find the right edge is to look at the stack, see where the approvals stopped adding anything, and draw the line there.
The fifth is the person at the top. Moving decisions down will feel, to the person who used to sign them, like losing sight of her own operation, and the chapter's account of what happens in the body is not exaggerated. The practical answer is to give her something better to see than the stack: the log, read weekly, and the dates in it. She will know more about her operation from a column of decision dates than she ever learned from forty signatures.
And the decision that most changes the result is what happens the first time a decision made at the lower level goes wrong, as some will. If it is taken back up, permanently, the organisation learns that authority below is conditional on never being wrong, which nobody can meet. If it is reviewed, the edge redrawn if need be, and the authority left where it was, the organisation learns that decisions below are held to the same standard as decisions above. Chapter 6 places the sort in the second month of a ninety-day plan, and Chapter 7 gives the sort and the steps of authority as a page each.
It is ten past seven on a Tuesday in February, and the shift engineer is standing in the pump house with her hand on the casing of the number two dosing pump. It is the new one. The hum of the room is the right hum.
On the clipboard by the door is the log of decisions for the works. The last entry is hers, from Friday: a worn seal kit on the number three pump, ordered the same morning, reason given in a line. Beside it, under told, is the regional director's name and the time the message went.
There is one box at the bottom of the page now. She signed it herself.
Chapter Three
A room with an owner and enough information can still fail to decide, for want of a rule for when it is finished.
The Meeting That Could Not Close
Haute Lumière · Who Decides · Chapter Three
It is ten past six on a Thursday, and the executive team of a software company has been in the long room on the second floor since half past four. On the whiteboard, in four colours, are the options for the old scheduling product, written up in March: A, retire it at the end of the year; B, retire it next summer; C, keep it on maintenance only; D, sell the customer base to a partner. Under each is a list of points that has grown at every meeting. There are thirty customers still on the old product. Two engineers spend most of their week keeping it running. Everyone in the room has said, at one meeting or another, that A or B is the right answer.
The chief technology officer has just asked for one more piece of analysis, on the cost of migrating the last eight customers. It is a reasonable request. It was also, in a different form, the request at the end of the third meeting.
The head of customer success has the thirty names in a spreadsheet on her laptop. Her stomach has been tight since five o'clock, because she knows that every week without an answer is a week in which she cannot tell those customers anything true. She is a composite, drawn from the way these evenings are described in the research below and in the house's own practice, and no firm is meant. The chief executive, who owns the decision and knows it, looks round the room and says that she wants everyone to be comfortable before they commit. There is a murmur of agreement, and a fifth meeting goes into the diary.
Ammanuel Santa Anna's book on decision-making names what has happened in the long room, and it does so by describing the rule the team is using without knowing it:
Traditional consensus asks: "Does everyone agree this is the best option?" This is an extraordinarily high bar that often leads to watered-down compromises, endless deliberation, or the tyranny of the most persistent objector.12
This chapter makes one claim. A meeting that cannot close usually has an owner and usually has enough information; what it lacks is a rule for when it is finished. The research on fast decisions does not find teams that know less or argue less. It finds teams that decide how they will decide before they begin, and that close a question on a different test from the one the long room is using.
In 1989 Kathleen Eisenhardt published a study of how top management teams made strategic decisions in what she called high-velocity environments, markets in which technology, demand and competitors change quickly and information is often out of date.13 She studied eight microcomputer firms in the 1980s, interviewing their executives and following particular decisions through. Some of the teams were fast and some were slow, and the differences between them were not the ones the slow teams would have predicted.
The fast teams, in the study's best-known finding, used more information, not less, and it was real-time information about current operations rather than forecasts. They considered more alternatives at once, side by side, rather than one at a time. They relied on what Eisenhardt described as a two-tier advice process, in which an experienced counsellor, often an older executive outside the line of the particular decision, was consulted closely. And they resolved conflict by a rule she called consensus with qualification: the team sought agreement, and if agreement did not come, the executive responsible decided, with the others' views heard. In her sample, the faster decision makers also performed better.
The grade of this evidence should be stated plainly. It is an inductive study of eight firms in one industry and one decade, and its findings are propositions built from cases, not a test across many firms; the text was not reopened for this book, and the findings are given as the decision literature has long summarised them. A larger study, by Robert Baum and Stefan Wally in 2003, gathered measures from three hundred and eighteen chief executives between 1996 and 2000 and found, in their abstract's words, that "fast strategic decision-making predicts subsequent firm growth and profit."14 It is survey evidence, and it cannot show that speed caused the growth rather than travelling with whatever else those firms did well. Taken together, the two studies say something modest and useful. Speed in strategic decisions is not the enemy of quality, and in fast markets it seems to go with it.
Read the long room through Eisenhardt's findings and the difference is not that this team is short of information. It has a whiteboard full of it. It is short of a rule for what happens when the information is enough and the agreement is not complete.
Eisenhardt's second finding, the alternatives considered side by side, also has a direct reading in the long room. Its whiteboard does hold four options, but the meetings have treated them one at a time: a meeting largely about A, then one about whether C was safer, then one about whether a partner would buy the base. Each option has been examined in isolation and found to have a weakness, which every option has. Held side by side, the four would have been compared with each other rather than with an imaginary option that had none, and the comparison would have been quick. A team that weighs alternatives in sequence is always one meeting away from the option it has not yet discredited. The chief executive's wish that everyone be comfortable is consensus without the qualification. It gives every person in the room, as Santa Anna's book puts it in a sentence of its own, the thing that keeps the whiteboard full: "It grants veto power to every individual."
A trap on the other side
Haute Lumière · Who Decides · Chapter Three
The research also contains a warning, and a team that has just discovered the cost of slow decisions should hear it before it overcorrects.
In 2002 Leslie Perlow, Gerardo Okhuysen and Nelson Repenning published a study of an internet start-up under intense pressure to move fast, and called what they found the speed trap.15 As the decision literature has summarised it, and the text was not reopened for this book, the firm's habit of deciding quickly created the conditions that demanded still quicker decisions: choices made without enough thought produced problems that had to be answered at once, and the pace fed on itself until the firm could no longer step back. Fast was not the fault. A single speed for everything was.
Santa Anna's canon describes a case of exactly this shape, a technology company whose founders had built their culture on deciding within hours, and its diagnosis is the clearest sentence in this part of the literature: "The problem was not that the company was moving too fast. It was that it was moving at one speed for all decisions." The case records the chief technology officer's verdict once the firm had learned to tell its decisions apart: "We used to make decisions in an hour and spend six months cleaning up the mess. Now we spend a few days making the decision and the mess doesn't happen."
The canon's answer is to give a decision a speed before it is discussed, and it names three. Swift decisions are those that are "clear, time-sensitive, and reversible", and for them the canon is unsentimental: "Spending two hours in committee deliberation on a decision that could be made in ten minutes by a competent individual is not conscious governance. It is bureaucratic performance." Deep decisions are the consequential ones that are hard to reverse, and they deserve the full weight of deliberation. And some questions are not yet ripe, and the canon allows a group to say so out loud: "We don't yet have the conditions for a good decision on this. Let us hold it, continue gathering intelligence, and return to it when the field is more clear." It is careful to separate this from the long room's habit. Holding a question on purpose, with a date to return to it, "is deferral from discernment". Holding it because nobody will close it is something else.
The long room's agenda shows the cost of a single speed in both directions at once. The same team that has held the scheduling product for four meetings has spent parts of those meetings on matters that should never have reached it: the choice of a conference stand, the wording of a holiday notice, a request from one engineer to work from another city for a month. Each was swift by the canon's definition, clear and reversible and someone's to decide, and each took ten minutes of seven people's evening. The team was not slow on everything. It was slow on the one decision that mattered and thorough on the ones that did not, which is what a single speed produces.
The scheduling product is not a swift decision, and it is not unripe. It is deep, and it has been ready since June. What the team has been doing is treating a deep decision as if it were permanently unripe.
There is an old alternative to consensus, and it changes one word of the question.
Sociocracy, the governance method the Dutch engineer Gerard Endenburg developed in the 1970s from older Quaker and educational practice, decides by consent. The house's Luminous 100, its reading of the frameworks every consultancy sells, describes it in a line: a "Decision-making method where decisions stand unless someone has a "paramount objection" (reasoned, based on the organization's aim)." Its verdict on the method's balance is as good as any in the literature: it "Elegantly balances efficiency (not everyone has to agree) with inclusion (everyone has the power to object)."
The canon's book on governance spells out what consent asks of a person in the room, and the difference from consensus is the whole of the practice. "You don't need to love the proposal. You don't even need to think it's the best option. You need only to be able to say: "I can work with this. It falls within the range of tolerance for me. I have no objection that would make this decision harmful to the organization's purpose."" Its book on decision-making puts the test as two questions a convenor can ask aloud: "Can everyone live with this decision? Does anyone have an objection that signals this decision would cause harm or move us backward?"
Consent is not a rubber stamp, and the house is careful to say why. A room can arrive at a false consent as easily as a false consensus; Irving Janis's account of groupthink, in Victims of Groupthink (1972), describes teams that mistook their own silence for agreement and made disastrous choices in good spirits.16 The canon's guard against this is a sentence every convenor of a consent round should carry: "Inclusive decision-making requires learning to read silence — not as consent, but as information that needs investigation." A consent round in which nobody objects because objecting is dangerous has decided nothing that a vote would not have decided worse.
The order in which people speak matters as much as the question asked. The canon's book on decision-making is direct about it: "When leaders speak first, they anchor the conversation in their perspective." In the long room the chief executive has spoken first at every meeting, warmly and in favour of B, and every person since has framed what they said as a response to her. A consent round that begins with the newest member of the team and ends with the owner hears more objections, and better ones, than one that begins with the owner and asks whether anyone disagrees. The canon notes, too, what a simple vote would have hidden: "A 51-49 vote looks the same as a 99-1 vote in the outcome, but they represent vastly different levels of collective alignment." A round that asks each person in turn shows the difference before the decision is taken.
The canon adds one more caution, in the context of the body's own signals, and it applies to the head of customer success's stomach as much as to anyone's: "Somatic consent is not a veto power for discomfort." A tight stomach in a meeting is information. It may be a forecast about the thirty customers. It may be fear of the conversation she will have to have with them. A consent round asks her to say which, and if it is a forecast, to say what harm she sees, so that the decision can be changed to meet it.
The long room's evenings are expensive in ways the company's accounts can see and in ways they cannot.
The visible cost is in the engineering time. Two engineers have spent most of their working weeks since March keeping the old product alive, and every month of delay is a month of that. The thirty customers are paying, but each is also waiting to hear what will happen, and some are planning around a silence. Those figures exist in the company's own records, in its time-sheets and its renewal dates, and that is where they should be read; this chapter does not put a number on them.
The less visible cost is in the team. Santa Anna's canon, writing about the shadow emotions that governance cultures leave unnamed, gives the most accurate description of the chief technology officer's last request, and it is not an accusation: it describes "the paralyzing "analysis paralysis" that is often fear disguised as thoroughness." The extra analysis is a real question. It is also, quite possibly, the form his doubt has taken because there is no other place in the meeting for it to go. A room with no rule for closing turns every doubt into a request for more data, because a request for data is the one form of objection that cannot be refused.
The thirty customers are paying a cost of their own, and it is the one that will be remembered longest. Several have asked their account managers whether the product has a future, and have been told, truthfully, that a decision is expected soon. That answer has now been given four times. A customer who is told something will be decided soon, four times, does not conclude that the company is careful. She concludes that it does not know, and she begins, quietly, to look at alternatives of her own. The company's renewal records will show this in a year, and nobody will connect it to a whiteboard in March.
The habit also spreads. A team that watches its leadership take four meetings over a decision everyone already agrees on learns that decisions here are slow, and plans accordingly, by building in delay, by asking for decisions earlier than they need them, and by holding their own decisions until they are sure nobody above will reopen them. Where the Flow Stops reads [the meeting after the meeting]({{BOOK:orgdev-flow}}), where a firm's real assessment of a project is held in the lift; the long room has a corridor of its own, and in it the engineers have been saying for months what the whiteboard says in four colours.
Closing the room
Haute Lumière · Who Decides · Chapter Three
What changes this is a rule stated at the start of the meeting, and the discipline to keep it at the end.
Picture the fifth meeting. The chief executive opens it with three sentences she has written down. This is a deep decision, and it is a decision for today. We will look for agreement until half past five. If we have it, good; if we do not, I will decide, and each of you will be asked one question: can you live with it, and if not, what harm do you see? She has done one more thing beforehand. She has asked the company's longest-serving board member, who sold a product line of her own a decade ago and has no stake in this one, to read the whiteboard and tell her privately what she would do. That is Eisenhardt's counsellor, and the conversation took twenty minutes on the telephone. And she has gone through the agenda for the next month and marked every item with its speed. Three of them were swift decisions that had been brought to the executive team only because nobody was sure they could be taken anywhere else, and she has sent each back to the person who raised it with a note saying it is theirs to make.
The chief technology officer's analysis has arrived, and it says what everyone expected: the last eight customers will cost more to migrate than the others, and the cost is manageable. The discussion that follows is short, because everyone knows when it ends. At twenty past five the chief executive decides: option B, the product retired next summer, the last eight customers migrated first with an engineer assigned to each. Then she runs the round. The chief technology officer has no objection. The head of finance has none. The head of customer success has one, and states it as a harm: two of the thirty are hospitals with a regulatory audit in the summer, and a migration in the same month would put them at risk. The decision is changed in the room to meet it. The two hospitals move in the spring.
The order of what the chief executive did matters more than the rule itself. She named the speed of the decision before anyone spoke, so nobody had to argue that it was ready. She set the time at which agreement would stop being sought, which made agreement easier to reach before it. She heard an outside counsellor before the meeting rather than during it, so the advice informed her without becoming one more voice at the whiteboard. And in the round she asked for harms, not preferences, which gave the head of customer success a way to say what her stomach had been telling her that was neither a veto nor a complaint.
There are predictable ways this goes wrong, and Chapter 7 carries them with the run sheet for the round. The owner decides before the hearing and runs the round as a formality, and the room learns that consent means compliance. An objection is answered with an argument rather than a change, and the next one is not raised. The rule is used for swift decisions that should never have come to the meeting at all, and the meetings grow longer.
A closed meeting raises a question of its own, which is what the four earlier meetings had cost, and how a firm can count it (Chapter 5) without inventing a number. The canon's three speeds carry one more lesson the long room has not yet needed. A deep decision, once made, is also the hardest kind to revisit, and a firm that cannot bring itself to stop what it has started (Chapter 4) has a closing problem of a different shape.
The chapter closed one meeting. The book can go further, into the finer points of closing every meeting that should close, and leaving open the few that should not.
Begin with the agenda. Before a leadership meeting, the person who chairs it marks every item with one of the canon's three speeds. Swift items are clear, reversible and someone's to decide; they come off the agenda and go back to that person with a note, and the meeting hears only the decision. Deep items are consequential and hard to reverse; they stay, with a named owner, a time by which agreement will be sought and the rule for what happens if it is not. Patient items are not yet ripe; they are listed with the date on which they will return and the information that is being gathered in the meantime. An item with no speed marked is not ready for the meeting.
The second finer point is the page the room decides from. Eisenhardt's fast teams held their alternatives side by side, and the practical form is a single page with the options in columns and the same few questions in rows: what it would cost, what it would risk, what it would need, what it would make impossible. A team that has the options in columns compares them with each other. A team that discusses them one by one, at successive meetings, compares each with a perfect option that does not exist.
The third is the counsellor. For a deep decision, the owner speaks before the meeting with one experienced person who has no stake in the outcome: a board member, a former colleague, a senior person from another part of the firm. The conversation is short and private, and its purpose is not approval. It is to hear, once, how the decision looks to someone who has made one like it.
The fourth is the round itself. The owner states the decision she is minded to take. Then each person is asked, in turn, beginning with the newest member and ending with the most senior, one question: can you live with this, and if not, what harm do you see? There are three permitted answers. No objection. An objection, which must name a harm to the organisation's purpose, not a preference. Or a question, which is answered before the round goes on. Nobody is asked whether they agree, because agreement is not what the round is testing.
The fifth is what is done with an objection. There are three honest responses, and the owner should choose one aloud. The decision can be changed to meet the harm, as the hospitals were moved to the spring. The objection can be tested, by a named person by a date, if it rests on something that can be checked. Or it can be answered, with a reason, if the owner judges that the harm is real but smaller than the cost of meeting it; in which case the person who raised it is thanked, and the reason is recorded. An objection that is simply overruled without one of these three teaches the room not to raise the next.
The sixth is silence. A round in which every answer is no objection, quickly, should make an owner more careful, not less. The canon's sentence on silence is the rule: read it as information. The simplest check is to ask, after the round, one person who said nothing whether there is something she would want the owner to know before the decision is announced, and to ask it privately, afterwards, so that it costs her nothing.
And the decision that most changes the result is what the owner does when the round produces an objection she did not expect from someone she did not expect it from. If her face hardens, the round is over before it has begun, whatever is said afterwards. If she says tell me more about the harm, the next objection comes sooner and better. Chapter 6 places the three speeds and the round in the first month of a ninety-day plan, and Chapter 7 gives the speed of a decision and the round as a page each.
It is ten past six on a Thursday, and the long room on the second floor has been empty for forty minutes. On the whiteboard the four options are still there in four colours, and someone has drawn a ring round B and written a date beside it, and under it, smaller, hospitals in spring.
The head of customer success is at her desk with the spreadsheet open. She has written the first of the thirty letters. It says what will happen, and when, and who will help.
Tomorrow she will send the other twenty-nine.
Chapter Four
Once a course is begun, the decision to continue is inherited rather than taken, and the remedy is written at the start.
Too Far In to Stop
Haute Lumière · Who Decides · Chapter Four
It is nine o'clock on a Sunday evening, and the programme director of a housing association is at her kitchen table with the steering paper for Tuesday. The overhead light is on, and the rest of the house is dark. The programme is a new system for scheduling repairs across eleven thousand homes, two years in, on its fifth forecast. The template the programme office uses has a box on its first page headed Recommendation, and in every paper since the first, the word in it has been Continue. She has typed it again. The cursor is blinking after it.
She proposed the system. She stood in front of the board two years ago with a slide of a repairs visit booked in one call instead of three, and she believed it, and in part she still does. The supplier has missed two releases. The tenants' panel has stopped asking when it will arrive. The operatives in the depots have built their own spreadsheet to work round the parts that do arrive, which is the most reliable thing the programme has produced. Her chest is tight in a way that is not about the deadline. She is a composite, drawn from the way these evenings are described in the research below and in the house's own practice, and no organisation is meant.
What she cannot quite see from the kitchen table is that nobody has ever decided to continue. The programme was decided once, two years ago, by the board. Since then the steering group has received a paper every quarter, and every paper has said Continue, and every meeting has noted it. The decision to stop has never been on any agenda, because nobody is responsible for putting it there.
The house's Luminous 100, its reading of the frameworks every consultancy sells, describes the purpose of the stage-gate process, the oldest formal answer to this problem, in a single plain sentence: it "Prevents the common failure of carrying weak projects too long because of sunk cost attachment."17 The housing association has no gate. It has a box with a word already typed in it.
This chapter makes one claim. Once a firm has committed to a course, the decision to continue is usually not taken at all. It is inherited, quarter by quarter, by a review that has no stopping signal agreed in advance and is chaired by the people whose names are on the start. Escalation is less a failure of nerve than a decision nobody has been given, and it is answered by writing the conditions for stopping before they are needed.
In 1976 the organisational psychologist Barry Staw published a study with a title that has become shorthand for the whole subject: Knee-Deep in the Big Muddy.18 He gave business-school students the role of a corporate financial officer allocating research money between two divisions of a company, and then told them how their first choice had turned out. Those who had made a choice that failed, and who felt personally responsible for it, put more money into the same division the second time, not less. The study's text was not reopened for this book, and its design is given as the escalation literature has long described it. It is a laboratory study with students, and its authors and successors have been careful about how far it travels.
Five years later Staw set out the general problem in a review, and its abstract is still the best short statement of the mechanism: "Because it is often possible for persons who have suffered a setback to recoup their losses through an even greater commitment of resources to the same course of action, a cycle of escalating commitment can be produced."19 A setback creates the hope of recovery, and the hope of recovery justifies the next commitment, and each commitment raises the cost of admitting the first was wrong.
In 1985 Hal Arkes and Catherine Blumer gave the mechanism its everyday name in a paper called The Psychology of Sunk Cost.20 Their best-known demonstration, as it is usually reported, was a field experiment with a university theatre's season tickets: people who had been sold tickets at the full price went to more plays in the first part of the season than people who had been given a discount, as though the money already spent obliged them to attend. The paper's text was not reopened for this book, and the experiment is given in outline. What it shows, in a small and humane setting, is that a cost already paid and unrecoverable still pulls on what people do next, even when they know it should not.
None of this says that persistence is foolish. Many projects that are rescued after a setback should have been rescued. What the research shows is that the decision to continue is made under a particular distortion, one that pushes in one direction, and that the people most subject to it are the ones who made the first decision.
Read the steering paper through Staw's abstract and its fifth forecast has a familiar shape. Each forecast since the first has followed a setback, a missed release or a supplier change, and each has proposed a recovery: one more release, a new module, a further quarter. Each recovery, considered alone, is plausible. What the paper never shows is the series. A reader who saw only the fifth forecast would judge it on its merits. A reader who saw all five side by side would see a cycle, and would ask the question none of the forecasts contains: if the next recovery fails as the last four did, what then? The template has no place for that question, because it was designed to report progress, and a programme reports progress until it stops.
A fair in Vancouver
Haute Lumière · Who Decides · Chapter Four
Laboratory findings invite the question of whether real organisations behave the same way, and in 1986 Jerry Ross and Barry Staw answered it with a case.21 They studied the decision of the government of British Columbia to hold a world's fair, Expo 86, in Vancouver. Their abstract gives the arc in one sentence: "Despite rapidly increasing deficit projections (from a $6-million projected loss in 1978 to over a $300-million projected loss in 1985), the provincial government remained steadfast in its plans to hold Expo."
Their account of why is the part of the paper that matters to the housing association. The case, the abstract says, "not only illustrates the frequently studied processes of self-justification and biased information processing but also highlights the potential importance of institutional explanations of escalation." And it proposes that "escalation starts with project and psychological forces but can evolve over time into a more structurally determined phenomenon." The fair began as a project its sponsors believed in. By the end it was a commitment held in place by contracts, reputations, public statements and the simple fact that every institution around it had organised itself on the assumption that it would go ahead. Ross and Staw later studied a nuclear power plant on Long Island, Shoreham, that was abandoned only after many years, as a case of how organisations do eventually exit.22
A meta-analysis published in 2012 by Dustin Sleesman and colleagues gathered more than three decades of this research and compared the many explanations offered for escalation, from self-justification to the framing of losses to the social and structural pressures Ross and Staw described.23 Its abstract notes that "little effort has been made to systematically investigate these various accounts," which is what the authors set out to do. The broad lesson of the literature is that escalation has several causes at once, which is why no single remedy, such as telling people to ignore sunk costs, reliably works.
That lesson matters for anyone tempted to treat escalation as a personal weakness. The programme director is not a person who cannot let go. She is a person inside a structure that has made letting go the only act that nobody is assigned to perform. The supplier is contracted to deliver, the programme office to report, the steering group to receive the report and the board to be told. Each does its job. A stop is nobody's job, and so, as Ross and Staw found in Vancouver, the project continues not because anyone decided it should but because every institution around it is organised on the assumption that it will.
The repairs programme is somewhere between the two ends of Ross and Staw's arc. It began with a director's belief. It is now held up by a supplier contract, a board paper, a tenants' charter that promised the new system, and a programme office whose posts exist because the programme does. Each of those is reasonable. Together they make a structure in which Continue is the only word the template has room for.
Staw's students escalated most when they felt personally responsible for the first decision, and the housing association has arranged its review so that the person most responsible writes the paper.
This is not a criticism of the programme director. It is a description of the template. She is the person who knows the programme best, so she writes its paper. She is also the person whose judgement two years ago is on trial in every paper she writes, and she knows it, and her chest knows it. The steering group, which includes the director of operations who seconded the original proposal, reads the paper in the same light. Nobody in the room has a role in which recommending a stop would be an ordinary piece of work.
Santa Anna's canon, writing about organisations that cannot move on from what they have lost, describes the mechanism in terms that are usually reserved for grief, and they fit: "Decisions that would require revisiting past losses are unconsciously avoided." To stop the programme would be to lose two years, a supplier relationship and a public promise, and to lose also the version of herself that stood in front of the board. The canon notes elsewhere how organisations disguise such losses even from themselves: "They are framed as changes, decisions, strategic pivots, transitions — language that implies forward movement rather than the experience of something being taken away."
The Cairn, another of the house's books, reads [what happens when a goal is made to carry something it cannot hold](https://hautelumiere.life/editions/rsd-rsd-and-self-confidence/page/), and a programme that carries its sponsor's standing is a goal of exactly that kind. The canon's book on governance records the pattern at the scale of a whole organisation, a healthcare network whose strategic initiatives had each followed the same arc: "enthusiastic announcement, initial momentum, gradual erosion, and quiet abandonment." Quiet abandonment is the other way escalation ends. The project is not stopped. It is starved, and nobody has to say the word.
The director of operations, who seconded the proposal, carries a version of the same weight, and so does the board, which approved it in public and told the tenants' panel. Each of them would find it easier to stop the programme if someone else proposed it, and each assumes, reasonably, that the person best placed to propose it is the programme director. She assumes the same of them. It is the unowned decision of Chapter 1, arriving at the end of a commitment rather than the beginning.
The Luminous 100 offers a different word for stopping, and it is worth taking seriously because it changes what a stop means to the people in the room. In its reading of the stage-gate, "the "kill" decision is reframed from failure to composting". A stopped project is not nothing. It is the knowledge, the spreadsheet in the depots, the data cleaned for the migration and the people freed for other work, returned to the organisation as material for what comes next.
The research on escalation is more useful on prevention than on cure, and its best advice is about what to write down before anyone has anything to defend.
In 1993 Daniel Kahneman and Dan Lovallo described the pattern by which managers make timid choices and bold forecasts, and the remedy they argued for, taking an outside view of how projects like this one usually go rather than an inside view of how this one is planned, has become standard advice.24 Their text was not reopened for this book, and the argument is summarised. The practical version is a single question asked at the start: of programmes like ours, in organisations like ours, how many arrived on time, and what happened to the ones that did not?
A second instrument works on the same principle from the other direction. In 1989 Deborah Mitchell, Edward Russo and Nancy Pennington studied what they called prospective hindsight, which their abstract describes as "generating an explanation for a future event as if it had already happened."25 Explanations for events treated as certain, they found, were longer and contained more episodic reasons. The psychologist Gary Klein later built a practice on the idea, the premortem, in which a team imagines that its project has failed and writes down why, before it starts. The Cairn also shows how [an obstacle written down in advance arrives as an appointment rather than a verdict](https://hautelumiere.life/editions/rsd-rsd-and-self-confidence/page/), and a stopping signal is an obstacle written down for the whole organisation.
Asked at the housing association two years ago, the outside-view question would have been uncomfortable and useful. Large systems programmes in organisations of its size have a long public record of running late and over cost, and the board could have read some of that record in an afternoon, in the reports that public bodies are required to publish about their own programmes. Nobody would have concluded that the repairs system should not be built. They might have concluded that a programme which was expected to run late needed, from its first day, a clear statement of how late was too late.
Santa Anna's book on decision-making describes the same discipline as a property of the decision itself rather than of the people making it: "Building review and adaptation into the decision architecture — what we call learning loops — transforms decisions from static pronouncements into living, evolving commitments." Among the questions it asks of every loop is one the housing association's template has never contained: "What voices were right that were initially dismissed?"
The review with a door in it
Haute Lumière · Who Decides · Chapter Four
What changes the Sunday evening is a different template, and a different person reading it.
Picture the same housing association a year later, starting its next large commitment, a programme to fit new heating systems across its older homes. Before the board approves it, the sponsor writes a single page that will be read at every review. It says what the programme expects to see, and by when: the number of homes fitted by the end of the first year, the cost per home, the tenants' satisfaction measured the way the association already measures it. It says what would make the programme pause and be reviewed from the beginning: two missed milestones in a row, or a cost per home above an agreed line, or the tenants' panel withdrawing its support. It names the person who will read each review and recommend whether to continue, and that person is the director of finance, who did not propose the programme and whose standing does not rise or fall with it. And it states that a recommendation to pause or stop is an ordinary outcome of a review, to be thanked and acted on, not a failure to be explained.
The Luminous 100's description of the stage-gate gives the rule its older form: "Each gate has explicit criteria — projects that don't meet criteria are killed, freeing resources for stronger ones." The page does the same work in an organisation that has no gates. Its value is not that it predicts the future accurately; nobody's first page does. Its value is that it was written by people who had nothing yet to defend, at the one moment when the conditions for stopping could be stated without anyone hearing them as a verdict on themselves.
For the repairs programme, which began without such a page, the programme director does something harder. She rewrites the steering paper on the Monday. Its first page now has two columns: on the left, what the board was told two years ago that the programme would have achieved by now; on the right, what it has achieved. The recommendation box says: Pause new development for one quarter. Ask the director of finance to lead an independent review, including the depots' spreadsheet, and bring a recommendation to the board. She sends it before she can change her mind, and then, because she is honest, she sends a second email to the director of operations, saying that she knows this is not what either of them hoped to write.
The order of what she did matters. She put the original promise beside the present result, which made the gap a matter of record rather than of her opinion. She asked for a reader who had not started the programme. She named the depots' spreadsheet, which made the operatives' work count before the review began. And she proposed a pause rather than a stop, which made the decision reversible and so made it possible to take.
The steering group, when it meets on Tuesday, does something it has not done before. It reads the two columns in silence before anyone speaks, and then the chair asks the director of operations to speak first, because his name is on the original proposal and the room needs to hear that he will not defend it for its own sake. He says that the right-hand column is what he would have expected to see a year ago, and that he agrees with the pause. The meeting takes forty minutes. Nobody is asked to explain anything.
There are predictable ways this goes wrong, and Chapter 7 carries them with the review page and its run sheet. The stopping signal is quietly moved when it is reached. The independent reader is chosen by the sponsor. The person who recommends the stop is moved to other work, and nobody recommends one again. A firm that wants to know what its late stops cost will find that the answer lies in records it already keeps (Chapter 5), and that the counting has to be agreed before the next commitment starts.
The first two places a firm's decisions wait, the decision nobody owns (Chapter 1) and the approval queue, are about decisions not yet made. This one is about a decision made so firmly that nobody can unmake it, and it needs the same remedy at the other end: an owner, a date and a question that has a place to be asked.
The chapter followed one programme to its stop and one to its first review. The book can go further, into the finer points of writing the page that makes a stop possible.
Begin with the moment. The page is written before the commitment is approved, by the person proposing it, and approved with it. Written later, it is written by someone who already has something to defend. If the organisation has commitments already running without such a page, which most do, the page is written now, for each, by someone who did not start it, and the commitment is read against it from the next review.
The second finer point is what the page contains, and it is short. What the commitment is expected to show, by when, in measures the organisation already keeps. What would make it pause and be reviewed from the beginning: a small number of signals, each stated as something anyone could see in a record, never as a judgement. Who will read each review and recommend, and that person is not the sponsor and does not report to the sponsor. When the reviews will be held. And one sentence saying that a recommendation to pause or stop is an ordinary outcome of a review.
The third is the level of the signals, and here the book is deliberately silent on figures. A signal set too tight will stop good work at its first difficulty; a signal set too loose will never be reached. The people who carry the commitment's risk set them, from the organisation's own experience of commitments like this one. The outside view of Kahneman and Lovallo is the best help: before setting a signal, ask how commitments of this kind have usually gone, here and elsewhere, and set the signal where a commitment would be going clearly worse than the usual.
The fourth is the review itself. It begins with the two columns, what was expected by now and what has happened, read in silence before anyone speaks. The reader speaks first and recommends. The sponsor speaks next, and the question put to her is not why is it behind? but what do you know now that you did not know at the start? Then the steering group decides: continue, change course, pause for review, or stop. A pause is often the right first decision, because it is reversible, and a reversible decision is easier to take than a final one.
The fifth is what a stop releases. The canon's word for it, composting, is practical. When a commitment stops, write down what it leaves: the people freed and where they will go, the money not yet spent, the knowledge gained, the parts that work and can be kept, like the depots' spreadsheet. A stop that releases nothing visible will be remembered as a loss. A stop whose released resources are named, and put to use within a month, will be remembered as a decision.
The sixth is the person who recommends the stop. Her standing is the test of the whole practice. If she is thanked in public and her next assignment is a good one, the organisation has learned that recommending a stop is part of the work. If she is quietly moved aside, it has learned the opposite, and no page will ever be read honestly again.
And the decision that most changes the result is the first time a signal is reached. Someone will propose that the signal was set too tightly, and that it should be moved. Sometimes that is right. But it is decided by the reader, not the sponsor, with the reason written down, and the commitment is reviewed from the beginning either way. A signal moved quietly when it is reached was never a signal. Chapter 6 places the pages for running commitments in the second month of a ninety-day plan, and Chapter 7 gives the commitment page and the two-column review as a page each.
It is nine o'clock on a Sunday evening, and the programme director is at the kitchen table with a different paper. The repairs programme was paused in the spring and closed in the summer, on the director of finance's recommendation. The depots' spreadsheet was kept, and rebuilt properly, and it now books most repairs in one call.
The paper on the table is for the heating programme's first review. Its first page is the page written before it began, with the expected figures on the left and the actual ones on the right. Two lines are behind. None has reached its signal.
In the box headed Recommendation, the cursor is blinking on an empty line.
Chapter Five
The cost of a late decision can be counted, but only from records a firm already keeps and by rules agreed in advance.
What a Late Yes Costs
Haute Lumière · Who Decides · Chapter Five
It is half past five on the last Friday of March, and the chief operating officer of a regional insurer has the decision log open on the screen in the small meeting room by the claims floor. The blinds are half down against a low sun. The log is a spreadsheet the convenor of a working group has kept since the first week of January, and it has one row for every decision of any consequence that crossed the operations leadership in the quarter: a hundred and forty rows. Each says what was decided, when it was first raised, when it was decided, who decided it, and whether it has since been reopened. Nobody had kept such a list before.
Across the table the finance director has a printout of the same sheet and a pencil. She has drawn a line under the twenty longest waits. The longest is a change to the way small water-damage claims are approved, raised on the ninth of January and decided on the twenty-first of March, and in the column the convenor added for notes it says agreed in principle 16 Jan; waited for sign-off. The chief operating officer looks at that row and feels the same heat at the back of his neck he felt in January, when he first heard the change described and thought, then, that it was obviously right.
He is a composite, drawn from the way such reviews are described in the research below and in the house's own practice, and no insurer is meant. The finance director asks him the question he has been expecting. What is it worth? He is tempted, for a moment, to multiply something by something and give her a large number. He does not, because he knows she will ask him where each part of it came from.
Ammanuel Santa Anna's canon describes the cost he is looking at in a passage about trust that names its price without inventing one:
In organizations with low collective trust — which is to say, in most soul-depleted organizations — everything takes longer, costs more, and produces less.26
This chapter makes one claim, and it closes the argument of the four before it. The cost of slow and unclear decisions can be measured, but only by counting time from records the firm already keeps, pricing only what those records show, and agreeing before anything changes how the count will be read. A decision log kept for one quarter is both the first measure of where a firm's decisions wait and the baseline against which any gain from deciding better can honestly be counted.
The scholar who did most to describe how organisations actually decide was Herbert Simon, and his account begins with a limit. In 1955 he published A Behavioral Model of Rational Choice, which argued that real decision makers do not search all the alternatives and pick the best; they lack the information, the time and the capacity to compute it.27 The following year, in Rational Choice and the Structure of the Environment, he described what they do instead, the behaviour he called satisficing: searching until an option is found that is good enough, and stopping there.28 Both papers are summarised here from their long reception, and their text was not reopened for this book. Simon's work on bounded rationality earned him the Nobel memorial prize in economics in 1978. Before either paper, in Administrative Behavior (1947), he had made the organisation itself the subject: a firm, he argued, is a structure for dividing up decisions, deciding which premises each person decides from and which are decided for them by the people around them.29 Read that way, every row of the convenor's log is a small piece of the firm's design, made visible.
Simon's idea is often read as a description of human limitation, and it is that. It is also a description of time. A decision has a moment at which it is needed, and a choice that is optimal but late can be worse than a choice that is good enough and on time. The water-damage change was good enough on the sixteenth of January. Everything after that date was not deliberation. It was waiting.
That is the first thing a decision log shows, and it shows it without any arithmetic. Read down the convenor's notes column and the same phrases recur: agreed in principle, waiting for sign-off, needs to go to the next meeting, parked pending. Each marks a point at which a decision was, in Simon's sense, already made, and the firm went on holding it. The canon's book on governance gives the general form of what the log has caught: "Every hierarchical organization is, to some degree, making decisions based on distorted information." The same book notes where some of the waiting comes from, in a list of what low trust does to an ordinary week: "Meetings take longer because people are managing impressions rather than speaking truth." The log does not record impressions. It records dates. But a decision that took nine weeks from agreement to sign-off is, very often, a decision that waited for someone to feel safe enough to put a name to it.
The log adds a second sentence beside it. Every organisation is also, to some degree, making decisions late, and it rarely knows by how much.
Three hundred and fifty-eight decisions
Haute Lumière · Who Decides · Chapter Five
Speed is not the only thing worth counting, and the best-known study of decisions that went wrong is a warning against counting it alone.
The management scholar Paul Nutt spent much of his career collecting decisions made in real organisations and following what happened to them. In 1999 he summarised the work in an article whose title states its finding: Surprising but True: Half the Decisions in Organizations Fail.30 His abstract describes "Studies of 358 decisions in medium to large organizations in the U.S. and Canada," and traces the failures "to managers who impose solutions, limit the search for alternatives, and use power to implement their plans." Success went with the opposite habits: managers "who make the need for action clear at the outset, set objectives, carry out an unrestricted search for solutions, and get key people to participate". The tactics most prone to fail, the abstract adds, "were used in two of every three decisions that were studied."
Nutt's definition of failure, and how each decision was classified, is in the full work, which was not reopened for this book; the figure is his, and it should be read as one researcher's careful judgement across a large set of cases, not as a law. But its direction agrees with a longitudinal study published three years earlier. James Dean and Mark Sharfman followed fifty-two strategic decisions in twenty-four companies to see whether the way a decision was made affected how well it turned out, and their answer, as the literature has long summarised it, was that it did: decisions made with more careful gathering and analysis of information were more effective, and decisions shaped by political manoeuvring less so.31
Put those two studies beside the decision log and they say what the log must also count. A firm that measured only how long its decisions took would reward the habits Nutt found in failing decisions, the imposed solution and the narrowed search, because those are fast. Chapter 3 described the speed trap a team can fall into when every decision is treated as urgent, and a measure of speed alone would build the trap into the firm's own figures. The insurer's own log bears this out on a small scale. Of the decisions made fastest in the quarter, within a day of being raised, several were reopened within a month, because they had been made by one person on the spot without asking the people who would carry them out. Of the decisions that took longest, almost none were reopened, but that was because most of them had waited for nothing more useful than a signature. Neither end of the distribution is the target. The target is a decision made as soon as the people who know enough have been heard, by someone entitled to make it, and then left alone.
So the convenor's sheet has a column for reopening. A decision that is made quickly and unmade a month later has not been made quickly. Santa Anna's book on governance asks the same of every decision once it has been carried out: "Did the decision produce the intended results? Were there unintended consequences? What can be learned?"
The finance director's question deserves a careful answer, because the honest answer is smaller than the temptation and much more useful.
Some of what a late decision costs is in records the insurer already keeps. The water-damage change, once made, lets claims handlers settle small claims on the first call instead of sending each one for a second approval. The claims system records when every claim was opened and when it was settled; it has done so for years. It is therefore possible to see, from the firm's own data, how long small water-damage claims took before the change and how long they take after it, and to read, from the time each handler records against a claim, how much work the second approval used to take. The days between the sixteenth of January and the twenty-first of March are days on which that saving was available and not taken. Their cost is in the records, and it can be counted.
Other costs are real and cannot be counted from any record. The handlers who knew for two months that a simpler process had been agreed and was not being used. The customers whose claims took a week longer than they needed to, some of whom will not renew, for reasons the insurer will never be told. The team leader who raised the change and has quietly concluded that raising changes is not worth the trouble. The canon is right that these are the larger costs over time, and its book on performance is right about what follows from that: the answer is not to give up on counting, "Not by abandoning measurement — that would simply replace one form of blindness with another". The answer is to count what the records show, say plainly what they do not, and never put an estimate of the second kind into a figure that someone will be asked to sign.
Practitioners and consultancies have published figures for what slow decisions cost firms in general, and some of them are large. This chapter does not repeat any of them, because they are rarely traced to a study that measured them, and because a general figure is no use to the finance director. She does not need to know what decisions cost the average insurer. She needs to know what the water-damage change cost this one, and the claims system can tell her.
There is a discipline in that smaller answer that the chief operating officer comes to value. A figure built from the claims system can be checked by anyone with access to the system, including an auditor, a regulator or a sceptical colleague in another department. A figure built from estimates of goodwill lost can be checked by nobody, and so it will be believed only by the people who already agreed with it. The first kind of figure persuades. The second kind starts arguments, and the argument is usually about the person who produced it rather than about the decisions it was meant to describe.
All of this could be read as a reason for caution, as though no gain from deciding better could ever be measured well enough to share. The opposite is true, and the shape of an honest measure is not new.
The house's first book on organisations, Where the Flow Stops, sets out [how a gain is measured in a way both sides would sign]({{BOOK:orgdev-flow}}): a baseline locked at the start from records already kept, adjustments named in advance, a second measure that would expose gaming, a calculation checked by someone with no stake in the answer, and the result shown to the people whose work produced it. That shape applies to decisions as well as to scrap, and the decision log makes it practical.
The log's first quarter is the baseline. It shows, for each kind of decision, how long it took from being raised to being made, how often it was reopened, and who made it. Beside it sit the operational records the decisions touch, the claims times, the repair times, the approval times in the purchasing system, each of which the firm already keeps and has kept for years. The adjustments are named before the second quarter begins: a holiday period, a change of system, a surge of claims after a storm. The second measure, beside the time a decision takes, is the rate at which decisions are reopened, so that a firm cannot appear to decide faster by deciding worse. Each adjustment deserves a sentence written in advance, because each will be argued about if it is written afterwards. The storm that doubles the number of claims in a fortnight will slow every decision connected to claims, and the agreement says, before any storm, that weeks of that kind are either excluded or read separately. The new claims system due in the autumn will change how dates are recorded, and the agreement says that the old and new records will be run side by side for a month before any comparison is made across the change. None of this is sophisticated. All of it is the difference between a figure that survives its first sceptic and one that does not.
The figures are calculated by someone in finance who reports outside operations, and shown to the operations team in the same form finance sees them. Which records, which adjustments and which period are questions for each firm and the people it covers, and they are the substance of an engagement rather than a book. How a measured gain is then divided is not described here at all.
It is worth saying who this protects. A measure agreed in advance protects the people who changed how decisions are made, the team leader who raised the water-damage change, the handlers who now settle on the first call, from the scepticism of whoever reads the figure later. Without it, the gain depends on the story the chief operating officer tells. With it, the gain depends on what the claims system recorded, by rules everyone saw before they started.
The house's interest in this is stated once, plainly. Luminous Prosperity works with organisations to find the gains they already hold, where work, knowledge and decisions have stopped moving, and to share in what is found. Sharing is only honest when the measurement is, which is why this chapter has spent so much of itself on the part most discussions of decision-making leave out.
Reading the log
Haute Lumière · Who Decides · Chapter Five
What the chief operating officer and the finance director do next is the practice this chapter is about, and it takes an afternoon.
They sort the hundred and forty rows by the time each decision took, and look first at the slowest tenth, not the average, because the average hides the decisions that waited longest. For each of those fourteen they ask the same three questions. Who decided it, in the end? Who could have decided it, on the day it was raised, with what they knew? And what did the wait hold up, in a record that would show it? Nine of the fourteen were decided by the chief operating officer and could have been decided by the person who raised them. Three waited because nobody was sure who owned them. Two waited for good reason, on a regulator.
Then they look at the eleven decisions that were reopened. Seven were reopened because something the person deciding needed to know had not been asked for at the start; four because the person who decided was overruled later by someone who had not been in the room. The canon's book on governance poses the question beneath both lists: "Are there decisions being made at the wrong level of the system?" The log answers it, row by row, without anyone having to argue.
They agree three changes, and they write them on one page. Small claims decisions of the water-damage kind move to the claims team leaders, who tell the chief operating officer the same day. Every decision raised to the operations leadership carries an owner and a date on the day it is raised. And the two largest programmes on the books each get a page saying what would make them pause. Before they sign anything they do one more thing, which costs nothing and changes what the log means to the people who fill it. The chief operating officer writes to the claims team leader who raised the water-damage change in January. He tells her the date it was agreed, the date it was decided and the reason for the gap, which was his own sign-off, and he tells her that the decision is now hers to make for changes of that kind. Santa Anna's book on decision-making names the failure this note is meant to prevent: "Without clear accountability structures and feedback loops, the collective intelligence of the deliberation phase can be quietly overridden by the operational preferences of a few." The team leader's change was overridden by nothing more than a queue. The note says so, and says what has changed.
Then the finance director writes a fourth line, which is the one that will matter most next quarter: what will be counted, from which records, against which baseline, with which adjustments, calculated by whom. They both sign it.
There are predictable ways this goes wrong, and Chapter 7 carries the log, the review and the measurement conversation as run sheets. The log is kept for a month and then lapses. The slowest decisions are explained rather than examined. The figure is calculated after the fact by the person whose standing it serves. And the first quarter's figures are compared with the worst quarter rather than an ordinary one, which flatters the start and punishes everything after.
The chapter read one quarter of one log. The book can go further, into the finer points of keeping one, which is less simple than a spreadsheet makes it look.
Begin with what counts as a decision. The log is for decisions of consequence that cross a leadership team or travel upward: approvals, policy changes, commitments, exceptions, choices among options. It is not for the routine work of the operation, which has its own records. A useful test is whether, if the decision were made wrongly, someone senior would expect to hear about it. In a team of a few hundred people the log usually runs to a hundred or so rows a quarter, which is enough to read and few enough to keep.
The second finer point is the date a decision was raised, which is the hardest field to fill honestly. The rule is the first written record: the email, the paper, the ticket in which the choice was first put to someone who could make it. Not the date it reached the agenda, which flatters the queue, and not the date someone first thought of it, which nobody can check. The convenor fills it from the record, and when the record is unclear says so in the notes column rather than guessing.
The third is the other columns, and they are few. What was decided, as a sentence. Who decided it. The date it was decided. Who was heard. Whether it was reopened within three months, and why. What it held up while it waited, if a record shows it. A log with twenty columns is not kept for long. A log with seven is kept for years.
The fourth is who keeps it. The convenor, a person trusted to write down what happened rather than what the leadership would like to have happened, and given an hour or two a week to do it. The log is read by the leadership weekly, in a few minutes, and reviewed properly each quarter. It is visible to the people whose decisions are in it.
The fifth is how it is read. Look at the slowest tenth before the middle, because the middle hides the waits that matter. Look at the reopened decisions beside the fast ones, because a decision made fast and reopened was not made fast. Look at who decided, and ask, for the slowest and the reopened, who could have. And look for repeated phrases in the notes, waiting for sign-off, parked, next meeting, because each is a place where the firm holds decisions it has already made.
The sixth is the link to the measure. The first quarter of the log, and the operational records beside it, is the baseline, and the measurement conversation of this chapter is held before the second quarter begins, with the adjustments named in advance and the calculation done by someone outside operations. If any change has already begun before the log was kept, agree the measurement for it from now and do not count the earlier improvement; that is a real cost and a small one beside a figure nobody believes.
And the decision that most changes the result is what happens to the people in the log. If the slowest decisions are used to blame the people who held them, the log will be kept less honestly every quarter until it describes nothing. If they are used, as the chief operating officer used them, to see where the firm's design makes decisions wait, the log becomes the most useful page the leadership reads. Chapter 6 places the log at the start of a ninety-day plan, and Chapter 7 gives the log, the quarterly review and the measurement conversation as a page each.
It is half past five on the last Friday of June, and the chief operating officer has the decision log open in the small meeting room by the claims floor. There are a hundred and twelve rows this quarter. The slowest tenth is shorter than it was. Four decisions have been reopened.
On the table is the page the finance director signed in March. Beside it is the management accountant's sheet, which says what the claims system says: small water-damage claims are settled faster than they were, against the baseline, after the adjustment for the storm in May, and the rate of complaints has not risen. The figure is smaller than the one he was tempted to give in March. Nobody in the room will dispute it.
He closes the laptop, and the sun on the blinds goes on falling across the empty claims floor, where the last handler has just settled the last small claim of the day on the first call.
Chapter Six
Five practices set in order, from counting the decisions a firm already makes to measuring what deciding better is worth.
The First Ninety Days
Haute Lumière · Who Decides · Chapter Six
It is seven o'clock on a Monday, and the head of a division of an engineering firm, some five hundred people across three sites, is sitting in her car in the car park with the engine off. The rain has just stopped and the windscreen is still beaded. On the passenger seat is a notebook in which, over the weekend, she wrote one line: Nobody here knows who decides anything. She has read the first five chapters of a book like this one, or lived them, and she means to do something about it. The temptation, which she can feel in her shoulders, is to walk in at half past seven and announce a new decision framework by Friday.
She is a composite, drawn from the way these first weeks are described in the research the earlier chapters read and in the house's own practice, and no firm is meant. This chapter is written to her, and to anyone in her position: the person who runs something and wants to begin. It describes the first ninety days in order, with who is in the room, what each step needs, roughly how long it takes, the decisions that change the result, the mistakes made at each step and how to see them early, how to read what comes back, and what to do on the first Monday. The run sheets and working pages it refers to are gathered in Chapter 7.
Two things should be said before the plan. The first is that nothing here needs a programme, a framework with a name or a launch. Ammanuel Santa Anna's canon puts the reason in a sentence about governance that every first ninety days should begin under:32 "governance is not a structure. It is a practice." A leader who introduces decision roles as an initiative, with a slide and a title, will find that people respond to the initiative and not to the decisions. The second is that the order matters more than any single step. The chapters were written in the order a reader meets the problem. The ninety days run in the order the practices depend on each other.
The five practices of this book are the decision line (Chapter 1), the sort and the steps of authority (Chapter 2), the speeds and the consent round (Chapter 3), the commitment page and the two-column review (Chapter 4), and the decision log with its measurement conversation (Chapter 5). In the first ninety days they run in a different sequence: count, then name, then close, then move down, then review what is running, then measure.
The reasoning is simple. Every later practice changes who decides what, and people accept such changes when they can see why they are being made. So the first step changes nothing and counts everything: the log is started on day one, because it is the only record of how decisions moved before anything was changed, and it becomes the baseline Chapter 5 requires. Naming owners comes next, because it costs least and is most visible, and because nothing can be moved down until it is clear who held it. The speeds and the round follow, in the leader's own meetings, where she controls the conditions and everyone can watch. Moving decisions down comes only once the room has seen that an owned decision is respected, because the people receiving authority are watching for exactly that. The pages for running commitments come in the second month, once the leadership has the habit of reading a page before speaking. And the measurement conversation is held last, because it measures what the other steps have found, but its history is fixed in the first ten days, for the reason Chapter 5 gives: a baseline drawn after the changes have begun has nothing honest to measure against.
The canon's book on decision-making gives the principle behind the whole plan in one question: "the question is not "What is the best decision-making method?" but rather "What does this particular decision, in this particular context, with these particular people, actually require?"" The first ten days are spent finding out what the firm's decisions require, by looking at them.
Three roles carry the first ninety days, and they are named before anything else is done.
The sponsor is the most senior person whose decisions the practices will touch. In a small firm it is the chief executive; in a large one, the head of a division or a site, like the woman in the car. Her job is not to manage the work. It is to go first in her own meetings, to name herself as owner where she is the owner and someone else where she is not, and to protect the people who make decisions that used to be hers. Expect it to take three to five hours a week, most of it inside meetings she already holds.
The working group is small: four to six people. It includes someone who runs operations, someone from finance who will later hold the measurement conversation, two people from below the leadership whose decisions travel upward most often, and, on purpose, one person who joined recently, because newcomers still notice how long things take. It meets for an hour a week. Its job is to read the log, run the sort and bring back what it finds.
The convenor keeps the log of Chapter 5: every decision of consequence, raised and decided, with its owner and whether it was reopened. Often this is the operations member of the group. The convenor needs about two hours a week, and the one quality that matters most is that people trust the convenor to record what happened rather than what the sponsor would like to have happened.
One decision changes the result more than any other at this stage. The sponsor must decide, before day one, which decisions she herself will give away in the ninety days, and be willing to name them. A sponsor who intends to change how everyone else decides while keeping her own desk as it was will find, as the founder in Chapter 2 did, that the decisions keep flowing back to it.
Days one to ten: counting before changing
Haute Lumière · Who Decides · Chapter Six
The first ten days change nothing visible. They gather four inputs, and the convenor does most of the gathering.
The first is the log itself, started on day one and filled backwards for the previous quarter from the records that exist: the minutes, the approval system, the purchasing system's dates, the email trail of the larger decisions. It will be incomplete. That is expected, and the notes column says where.
The second is a list of the standing meetings in which decisions are taken, with who attends. The sponsor sits in on two or three she does not usually attend and, in each, counts: how many items were marked for decision, how many were decided, how many were moved to the next meeting, and how many had an owner anyone could name. Chapter 7 gives the signs that decisions are waiting as a single page, and she ticks them as she sees them.
The third is the decision inventory of Chapter 7: the fifteen or twenty most significant decisions of the last quarter, and for each, who decided in the end, who held the knowledge it needed, and whether they were the same person. The canon's book on governance describes the purpose of the exercise in a sentence: "The gaps between the two maps are your governance opportunities."
The fourth is the history the measurement conversation will read. Finance and the convenor list the operational records the firm's decisions touch, the lead times, the approval dates, the hire and overtime costs, the complaint and claim times, and copy out the last one to two years of each, dated and filed unaltered. Nothing is decided about it yet. It is fixed before anything could move it.
At the end of the ten days the working group meets for the first time and writes, on a single page, its first guess at where decisions wait: two or three kinds of decision that seem to have no owner, one or two that travel further up than they need to, one or two meetings that cannot close, and any commitment that seems to be continuing because nobody has asked whether it should. It is a guess, and labelled one. The page is kept and read again on day ninety.
What goes wrong at this stage. The commonest mistake is to announce. A sponsor who tells the division that she is starting a programme on decision rights will find that every manager starts copying her on everything, to be safe, and the log fills with decisions that were never hers. The early sign is a sudden rise in what reaches her desk. The second mistake is to begin with a framework, roles handed out across the whole organisation before anyone has looked at the decisions. The roles will be drawn around the organisation chart, which is exactly the map the canon warns is not the authority structure. The third is to judge. The log of the last quarter will show waits that reflect badly on particular people, sometimes on the sponsor herself. If the first use of the log is to blame someone, the log will be kept less honestly from that day.
In the third week the sponsor changes two of her own standing meetings, chosen because decisions are made there and because the first ten days showed them failing to close.
Before each meeting, she marks every item with its speed, as Chapter 3 describes. Swift items go back, with a note, to the person whose decision they are. Deep items carry a decision line at the top of the paper, as Chapter 1 describes: the choice in a sentence, the owner, the date, who is heard first and the few who must agree. Patient items carry the date they will return. In the meeting, each deep decision is closed by the owner at the time set, and the consent round of Chapter 3 is run before it is announced, beginning with the newest person and ending with the most senior. Chapter 7 gives the run sheet for both.
Alongside the meetings, the sponsor does one quiet piece of work each week: she takes two of the unowned decisions from the first-guess page and asks, upward or across, whose they are. When the answer comes, she writes it down where the people affected can see it.
The round adds about ten minutes to a meeting of eight people. The speeds usually save more than that, because the swift items leave the agenda. In the first weeks the round produces mostly no objection, quickly. The canon's guard against reading that as success is the right one, and the sponsor should expect three or four meetings before the first real objection arrives.
The decision that changes the result is what she does with the first real objection, and above all with the first one that turns out to be mistaken. She answers the harm, not the courage. She changes the decision, tests the objection or gives her reason, aloud. And she treats a mistaken objection exactly as warmly as a correct one. If she does those three things, the room learns quickly that objections are part of deciding. If she misses one, it learns that too.
What goes wrong. The sponsor decides before the round and runs it as a formality. The decision line names a committee. An owned decision is reopened by someone senior because they disliked it. The swift items come back to the meeting because their owners do not believe they are really theirs. The early signs are visible in the log: if decisions marked deep are still being carried to a second meeting in week four, the closing rule is not being kept; if decisions sent back as swift reappear, the people receiving them need to hear, in person, that they are trusted with them.
How to read what comes back. By day thirty the convenor's log should show a handful of decisions closed on the day set, each with a named owner, and fewer items carried between meetings. Read the objections raised in the round as the working group's best evidence of where the firm's knowledge sits: the people who raised harms nobody else saw are often the people who should own the next decision of that kind.
In the fifth week the working group runs the sort of Chapter 2 on one site or function, the one the first-guess page named as sending most upward. It gathers the last twenty or thirty decisions that travelled, puts the two questions to each, and sorts them into three piles: those that belong where they were raised, those that belong above, and those that went up because someone was stuck.
The sort takes an afternoon. Then the sponsor and the manager whose site or function it is agree, for each class of decision in the first pile, which step of authority it moves to and when that step will be reviewed: recommend and act unless someone objects, decide and tell, or full authority. They write the edges of each class, set by the people who carry its risk. For the third pile they agree an escalation path, which is a named person to ask for help with a hard case, and a time within which help will come.
The decisions that change the result are about order and evidence. Move the commonest decisions first, because they are where the waiting accumulates and where the people below already know most. Use the records that already show those people's judgement, the logs, the readings, the histories, as the evidence for the move, and say so, so that the authority is seen to rest on something. And move fewer classes than seems possible, so that each can be watched.
What goes wrong. The person above keeps a shadow approval, asking to see each decision before it is acted on. The edges are drawn so tightly that nothing real moves. A middle manager, newly trusted, becomes the new bottleneck. The first decision made below that goes wrong is taken back up for good. Each has an early sign in the log: decisions in the moved classes that still show two names, the decider and someone above; a moved class with no entries in a month; a new name appearing on most of the slow rows.
How to read what comes back. By day fifty the log should show the moved classes being decided where they were raised, with the time from raised to decided falling for those rows, and the person above told the same day. Read the reopened column beside it. If decisions in the moved classes are reopened more often than before, look first at what the people deciding were not told, before concluding that they were not ready.
Overlapping with the sort, the leadership lists every commitment of consequence now running: programmes, projects, pilots, contracts with years still to run. For each it asks whether there is a page, written at the start, saying what the commitment was expected to show and what would make it pause. Most will have none.
For each commitment without one, the working group finds a reader who did not start it and does not report to its sponsor, and asks that person to write the page now, with the sponsor's help but not under her direction: what the commitment should have shown by now, by the measures the organisation already keeps; what would make it pause; when it will next be reviewed. At the next review of each, the two columns of Chapter 4 are read in silence before anyone speaks, and the reader recommends first.
The decision that changes the result is how the first recommendation to pause is received. The sponsor of the division should be present, should thank the reader in a sentence, and should ask the commitment's own sponsor what she knows now that she did not know at the start. If the first pause is treated as an ordinary outcome, the next page will be written honestly.
What goes wrong. The reader is chosen by the sponsor. The signals are written so loosely that nothing could reach them. A signal is reached and quietly moved. The person who recommended a pause is moved aside. The early sign of each is the same: the right-hand column of the two-column paper begins to be written in adjectives rather than figures.
How to read what comes back. By day seventy each running commitment of consequence should have a page and a date for its next review, and at least one should have been read against its page. A leadership that finds every commitment on track against pages written in the same month has probably written the pages to fit. Read the first pages again at day ninety, beside the log.
Days sixty-one to ninety: measuring what was found
Haute Lumière · Who Decides · Chapter Six
In the last month the working group chooses one or two of the changes made, the ones whose effect is most likely to show in records the firm already keeps, and holds the measurement conversation of Chapter 5. Finance leads it. The operations member, the people whose decisions were moved and whoever will do the calculation are in the room. The sponsor attends and says little.
The conversation follows the order in Chapter 7: what has changed in how the decision is made, in plain words; what would show in existing records if it worked; what those records show over the history fixed in the first ten days; what moves the figure for reasons nobody controls, and how each will be allowed for; what second measure would show if decisions were merely being made faster and worse, and the reopening rate in the log is the natural one; who calculates, and how it can be checked; how the result will be shown to the people whose decisions produced it. The output is one page, signed by the people in the room. The first conversation usually takes two meetings of ninety minutes. Later ones take one.
By day ninety the first monthly figure should be on the page beside the log, read against the ordinary range of its history, in the same form finance sees it.
The decision that changes the result is timing, and most of it was taken on day ten, when the history was fixed. If a team began to change how it decided before day ten, agree the measurement for that change from now. The earlier improvement is not counted. That is a real cost and a small one beside a number nobody believes.
What goes wrong. The baseline is taken from the worst recent quarter, because it makes the gain look large. An adjustment is added after the results are in. Time is measured alone, without the reopening rate beside it. The calculation is done privately and announced. The early sign of each is the same: someone in the room, often one of the people whose decisions were moved, goes quiet. The sponsor's job at this step is to notice who has stopped speaking.
What this step does not settle is how a measured gain is then shared with the people who produced it: the proportions, the timing, the thresholds and the terms. Those depend on the organisation, its people and its history. They are decided with the people who will be covered, and for an organisation working with Luminous Prosperity they are the substance of the engagement, not of a book. What the first ninety days can do is make that later decision honest, by producing a gain that both sides already believe.
The working group meets on day ninety with five things on the table: the first-guess page from day ten, the log of the quarter beside the log of the quarter before, the sorted stack with its steps of authority, the pages for running commitments with their first reviews, and the signed measurement page with its first figure.
It asks a short list of questions, in order. Where was the first guess right, and where wrong? Which decisions now have an owner that had none? How has the time from raised to decided changed, in the slowest tenth as well as the middle, and has the reopening rate moved with it? Which classes of decision were moved down, and what happened to them? Which commitments were read against a page, and what was decided? Is there a gain both sides believe, and what does it show? And then the question that matters most for the next ninety days: who is now deciding who was not, and who has stopped?
What not to conclude is as important. Ninety days are too short to say that the culture of decision-making has changed; they are long enough to say whether the practices are being kept. One quarter's faster decisions are not yet a gain, for the reasons Chapter 5 gave. And a log with few changes in it is itself a finding: it usually means one of the earlier steps is not yet real, and the first place to look is whether owners are really being named.
The canon's book on governance, describing the first pilot of a redesigned decision, gives the right expectation for the whole quarter: "Remember: distributed authority is an iterative practice. The first design will not be perfect. The goal is to begin the learning cycle." The second ninety days begin from the log the first ninety days produced.
If the whole of this chapter is too much to begin with, three things can be done on the first Monday, and none of them needs anyone's permission.
Take the agenda of the next meeting you chair, and beside each item write whether it is swift, deep or not yet ripe. Send the swift ones back to the people whose decisions they are.
Choose one decision that has been on more than one agenda. Write it as a sentence with a choice in it, and ask the person above you, or beside you, whose it is. Write down the answer.
And start the log. A spreadsheet with seven columns, filled from this week's decisions: what was decided, raised when, decided when, by whom, who was heard, whether it was reopened, and what it held up while it waited.
The canon's book on governance closes one of its exercises with an instruction that fits the first Monday better than any plan: "Simply notice. The noticing itself begins to shift the pattern." A firm that writes down, for one quarter, when its decisions were raised and when they were made has already begun to decide differently.
It is seven o'clock on a Monday, and the head of the division is in the car park with the engine off. The windscreen is dry this time. On the passenger seat is the same notebook, and under the line she wrote ninety-one days ago she has written a second: Some of them do now.
On her laptop the log shows a hundred and thirty rows for the quarter. The two meetings she chairs closed every deep item on the day it was set. At one of the three sites, the maintenance team orders its own parts and tells her the same afternoon. In finance's drawer there is a signed page with a history, two adjustments and a first figure on it.
None of these has a name. Nobody has announced anything. She marks the first item on today's agenda swift, sends it back to the person whose decision it is, and goes in.
Chapter Seven
Fourteen working pages, the decision line, the sort, the round, the review and the log, ready to carry into a room.
The Field Manual
Haute Lumière · Who Decides · Chapter Seven
The chapters of this book describe five practices as people doing them.33 This chapter holds the working pages: the run sheets, the question sets, the logs and the one-page agreements that the earlier chapters point to and do not print. Each page is meant to be photocopied, filled in by hand or kept as a plain spreadsheet, and held in a folder that the convenor of Chapter 6 looks after. Nothing here needs software.
The pages follow the order of the first ninety days, not the order of the chapters. Each opens with a line saying where it comes from and when it is used. Where a page gives a script, the words are a starting point; say them in your own voice. Where it gives a time, the time is a planning guide from practice, not a finding. Santa Anna's book on management gives the reason the small entries on these pages matter, in a sentence worth keeping at the front of the folder: "Every delegation decision is a growth opportunity or a growth suppression."
Two rules govern the whole manual. Write down what was decided, by whom and when, not what everyone remembers. And close what you open: a decision named and not made, an objection raised and not answered, a page written and not read, teaches the opposite of what the practice is for.
What these pages do not contain is any figure for a limit, a signal or a share, nor the design of any arrangement for sharing a measured gain: proportions, thresholds, timing or terms. Those belong to each organisation and the people it covers, and they are decided separately.
From Chapters 1 to 3. Used in days one to ten, and again at day ninety.
Sit in on a meeting where decisions are taken, and read the last quarter's minutes. Tick what you see.
| Sign | Seen? | Where, and when |
|---|---|---|
| The same item has appeared on three or more agendas | ||
| Papers are marked for discussion when a choice is being asked for | ||
| Minutes say agreed and nobody can say who will act | ||
| People ask who is signing this off? outside the meeting | ||
| A decision is described by its topic rather than its choice | ||
| A common, low-risk decision needs two or more signatures above the person who raised it | ||
| Items are carried to the next meeting for one more piece of analysis | ||
| A decision is reopened by someone who was not in the room | ||
| A running commitment has reported continue at every review | ||
| Nobody can say what would make a running commitment stop |
Reading it. One or two ticks are ordinary. Three or more in the same meeting mean decisions there are waiting for something other than thought. Start the decision line (page 3) and the speeds (page 6) in that meeting.
From Chapter 6, after the canon's authority assessment. Used once in days one to ten.
List the fifteen to twenty most significant decisions taken in the last quarter. For each, fill one row.
| Decision (as a choice) | Who decided in the end | Who held the knowledge it needed | Same person? | Raised (date) | Decided (date) | Reopened? |
|---|---|---|---|---|---|---|
Underneath, answer three questions.
Reading it. The rows where the third and fourth columns differ are where to look first. Bring them to the first meeting of the working group with the first-guess page.
From Chapter 1, after Rogers and Blenko (2006). Used from day eleven, at the top of every paper that asks for a deep decision.
| Decision (a sentence with a choice in it: whether to, which of) | |
| Decided by (one person) | |
| By (date) | |
| Heard first (who, and on what) | |
| Must agree (only by law, written policy, or a harm they are responsible for preventing) | |
| Carries it out | |
| Told afterwards | |
| Reopened only if (new information not available on the day) |
Before writing it. Ask the person above, or beside, whose decision it is. Write down the answer and where it was given.
After the decision. The owner tells those heard what was decided and whether their input changed it, in a sentence each. She tells those who carry it out the decision in plain words, with the date it takes effect.
Mistakes to watch for. Naming a committee as owner. Giving five people the right to agree. Writing a topic instead of a choice. The owner, once named, going back to consulting. A senior person reopening an owned decision because it reached her as a complaint.
Signs it is working. Within a month, papers arrive with the line already filled in. Items stop being carried between meetings.
From Chapter 2, after Garicano (2000). Used once in days thirty-one to fifty, for one site or function; repeat each quarter.
Gather the last twenty to thirty decisions that travelled upward from one site, team or function. Choose ordinary ones. For each, ask two questions and mark one pile.
| Decision | Who knew most about it on the day it was raised? | What did the approval above add? (information · wider view · risk carried · a regulator's interest · only a date) | Pile |
|---|---|---|---|
The three piles.
Reading it. Most first sorts put most decisions in pile A. Look also for decisions the person above signed without being able to recall them; they show what the chain was protecting.
5 · Moving authority in steps
Haute Lumière · Who Decides · Chapter Seven
From Chapter 2, after the canon's graduated authority. Used in days thirty-one to fifty, once the sort is done.
For each class of decision in pile A, agree the step and the date it will be reviewed.
| Class of decision | Step (1 recommend and act unless someone objects · 2 decide and tell the same day · 3 full authority) | Edges (beyond which it goes up again, set by those who carry the risk) | Who is told, and how | Review date |
|---|---|---|---|---|
The person above. Reads the log weekly. When she would have decided differently, she says so to the person who decided, in a sentence, and the decision stands unless it breaks a law or a policy.
The person receiving the step. Records each decision and its reason in the log on the day. The first time, she will check twice and tell more people than she needs to. Whoever she tells says, in a sentence, that it was not required and is welcome.
Mistakes to watch for. A shadow approval, every decision seen before it is acted on. Edges drawn so tightly nothing moves. A newly trusted manager becoming the new bottleneck. The first decision that goes wrong taken back up for good, instead of reviewed and the edge redrawn.
From Chapter 3, after the canon's three temporal modes. Used before every leadership meeting from day eleven.
Mark every agenda item before the meeting.
| Item | Swift (clear, reversible, someone's to decide) | Deep (consequential, hard to reverse) | Not yet ripe | What happens to it |
|---|---|---|---|---|
| Swift: back to its owner with a note; the meeting hears the decision | ||||
| Deep: decision line (page 3), closing time, consent round (page 7) | ||||
| Not yet ripe: return date, and what is being gathered meanwhile | ||||
For a deep item, the options page. Put the options in columns and the same questions in rows, so they are compared with each other rather than one at a time.
| Option A | Option B | Option C | Option D | |
|---|---|---|---|---|
| What it would cost | ||||
| What it would risk | ||||
| What it would need | ||||
| What it would make impossible |
Reading it. An item with no speed marked is not ready for the meeting.
From Chapter 3, after Eisenhardt (1989) and sociocratic consent. Used for deep decisions from day eleven.
Before the meeting (twenty minutes, by telephone). The owner speaks privately with one experienced person who has no stake in the outcome, and hears how the decision looks to her.
At the start of the item (one minute). The owner says: This is a deep decision, and a decision for today. We will look for agreement until (time). If we have it, good; if not, I will decide, and each of you will be asked whether you can live with it, and if not, what harm you see.
At the closing time. The owner states the decision she is minded to take.
The round. In turn, newest member first, most senior last, each is asked: Can you live with this, and if not, what harm do you see? Three permitted answers:
| Answer | What happens |
|---|---|
| No objection | Note it; go on |
| An objection, naming a harm to the organisation's purpose | The owner chooses one, aloud: change the decision to meet it · test it (who, by when) · answer it with a reason, and record the reason |
| A question | Answer it, then go on |
After the round. Ask one person who said nothing, privately, whether there is anything she would want the owner to know before the decision is announced. Then announce it, with the date it takes effect and who carries it out.
Mistakes to watch for. Deciding before the hearing and running the round as a formality. Answering an objection with an argument instead of one of the three responses. The owner speaking first in the discussion. Bringing swift items to the round. Reading quick silence as agreement.
Signs it is working. By the fourth meeting at least one objection has changed a decision. Deep items close on the day set.
From Chapter 4, after Kahneman and Lovallo (1993) and the stage-gate. Written by the sponsor before any commitment of consequence is approved, and approved with it. For a commitment already running without one, written now by a reader who did not start it.
| The commitment | |
| What it is expected to show, by when, in measures already kept | |
| How commitments like this have usually gone (the outside view) | |
| What would make it pause and be reviewed from the beginning (signals anyone could see in a record) | |
| Who reads each review and recommends (not the sponsor, not reporting to the sponsor) | |
| Review dates | |
| What a stop would release (people, money not yet spent, knowledge, parts worth keeping) | |
| Agreed: A recommendation to pause or stop is an ordinary outcome of a review. | |
| Signed and dated |
Mistakes to watch for. Signals written as judgements rather than records. A reader chosen by the sponsor. A signal moved quietly when it is reached.
From Chapter 4. Used at every review of a commitment against its page.
| What the page said we would see by now | What we see |
|---|---|
The order of the meeting.
Afterwards. Thank the person who recommended a pause or a stop, in public. Watch her next assignment.
10 · The decision log
Haute Lumière · Who Decides · Chapter Seven
From Chapter 5. Started on day one, kept by the convenor, read weekly, reviewed quarterly.
| What was decided (a sentence) | Raised (first written record) | Decided | Decided by | Heard | Reopened within three months? Why? | What it held up while it waited (if a record shows it) | Notes |
|---|---|---|---|---|---|---|---|
What counts. Decisions of consequence that cross a leadership team or travel upward: approvals, policy changes, commitments, exceptions, choices among options. Test: if it were made wrongly, would someone senior expect to hear about it?
The date raised. The first written record in which the choice was put to someone who could make it. Not the date it reached an agenda. If the record is unclear, say so in the notes.
The weekly read (five minutes). Which decisions have waited longest? Which phrases recur in the notes, waiting for sign-off, parked, next meeting?
From Chapter 5, after Nutt (1999) and the measurement standard of Where the Flow Stops. The review is held each quarter; the measurement conversation in days sixty-one to ninety, reading the history fixed by day ten.
The review (an afternoon). Sort the log by time from raised to decided.
The measurement conversation. Who is in the room: the person responsible for the money, the person responsible for the operation, someone who speaks for the people whose decisions changed, and the person who will calculate.
The one-page agreement. Signed before the period it measures begins, or from today if a change has already begun.
| What has changed | |
| Main measure, and the record it comes from | |
| History used, and the ordinary range it shows | |
| What moves the figure that nobody controls, and how each is allowed for | |
| Balancing measure (reopening rate), and its record | |
| Who calculates; who checks | |
| How and how often the result is shown, and to whom | |
| What happens if a record or system changes | |
| Date agreed | |
| Signed |
Mistakes to watch for. A baseline from the worst recent quarter. An adjustment proposed after results are known. Time measured alone. A calculation done privately and announced. Someone in the room going quiet.
From Chapter 6.
| Days | Practice | Who | Time it takes | What it produces | Early sign it is going wrong |
|---|---|---|---|---|---|
| 1–10 | Count: start the log, sit in, the inventory (page 2), fix a dated copy of the history | Sponsor, convenor | Sponsor about 4 hours a week; convenor about 2 | A first-guess page: where decisions wait | A sudden rise in what reaches the sponsor's desk |
| 11–30 | Name and close: speeds (page 6), decision line (page 3), consent round (page 7) in two meetings | Sponsor | About 10 minutes a meeting | Deep items closed on the day set | Deep items still carried to a second meeting in week four |
| 31–50 | Move down: the sort (page 4), steps of authority (page 5) | Working group; sponsor and the manager concerned | An afternoon to sort; an hour to agree | Classes of decision decided where they are raised | Moved decisions still showing two names in the log |
| 41–70 | Pages for what is running (pages 8 and 9) | Leadership; a reader for each commitment | An hour per page; a review per commitment | Every running commitment with a page and a review date | Right-hand columns written in adjectives |
| 61–90 | The measurement conversation (page 11) | Finance leads; operations; those whose decisions moved; calculator | Two meetings of 90 minutes | A signed page and a first monthly figure | Someone in the room goes quiet |
From Chapter 6. A meeting of the working group and the sponsor, about two hours.
Bring the first-guess page, the log of this quarter and the quarter before, the sorted stack with its steps, the commitment pages with their reviews, and the signed measurement page.
Ask, in order:
What not to conclude. That the culture of decision-making has changed (ninety days show whether practices are kept, not whether a culture has moved). That one faster quarter is a gain. That a log with few changes means there were few problems; it usually means one of the earlier steps is not yet real, and the first place to look is whether owners are really being named.
From Chapter 6. For anyone who wants to begin with less.
A decision is often waiting not for an answer but for someone to know whose answer it is. Naming the owner costs one sentence.
Luminous Prosperity Inc. takes this work into organisations as a gainshare. The flow is opened from the inside, with the people who already hold what the firm knows. The gain is counted from records the firm already keeps, against a measure both sides sign before anyone knows the answer, and the house is paid as a share of what is found: between thirty and ten per cent of it.
The book sorts the decisions of one team. Which of yours wait longest, and what deciding sooner would be worth, is read from your own records and your own people. That begins with a conversation.