Decision-Making Framework

The Meeting Where Everyone Nodded

Lesson 3.3 ended with two questions: which of the Five Voices did you hear today, and who is authorized to tell you it was talking? If you could not answer the second one, I told you that you did not have a thinking problem, you had a design problem. This lesson is the design.

Picture the meeting. Someone presents a recommendation. There is a deck. The numbers support it, the room has questions but not objections, and after forty minutes the leader says let us do it. Everybody leaves feeling productive. Now count what actually happened in that room. One option was considered. The evidence presented was assembled by the person advocating the option. No one was assigned to argue against it. Nobody asked whether the decision could be undone. And no record was made of what the group expected to happen, so in a year there will be nothing to check the outcome against — only memories, which will have quietly rearranged themselves to fit whatever occurred.

That is not a bad meeting. That is a normal meeting, in a normal organization, run by competent people. Which is exactly why decision quality is such a strange thing in most institutions: it is treated as an attribute of the leader rather than a property of a process. We say he has good judgment the way we would say he has a good ear. And if judgment is a talent, then there is nothing to build — you either have it or you hire it.

I do not believe that, and neither does the evidence. Judgment has a large learnable component, and most of it is not raw intellect. It is structure. A leader with an ordinary mind running a disciplined process will beat a brilliant one improvising, and he will beat him consistently rather than occasionally, because the disciplined process performs the same way on the day you are tired, on the day you are angry, and on the day the answer you want is sitting right there in front of you.

So here is the structure. I call it the Seven Questions, and the point of naming them is that you can run them in your head in ninety seconds on a small decision and on paper over two weeks on a large one. What is the actual decision? What are the real options? What would have to be true, and what do we actually know? Which door is this — one-way or two-way? It is eighteen months from now and this failed; why? Who decides, by when, and what exactly did we decide? And finally: what did we learn about the process, separate from how it turned out?

Before we walk them, one warning about the whole enterprise. A structure you only run when the decision feels big is not a structure; it is a ritual you perform to feel careful. The decisions that damage organizations are frequently the ones that did not feel consequential at the time — a salary arrangement, a handshake commitment, a contract clause nobody read. The value of a repeatable process is precisely that it runs on decisions your intuition has already cleared.

Answering the Wrong Question Correctly

Question one: what is the actual decision? This is where most bad decisions are made, and almost nobody spends time here, because the decision arrives pre-framed and the frame is invisible. Someone walks in and asks should we do X, and the room starts arguing about X. Nobody asks who decided that X was the question.

I will give you mine, since I have already told you the story. Two employees, not enough budget for both at full salary. I framed the decision as how do I keep both of these people. Given that frame, my answer was actually pretty good — cut both salaries in half, get their consent, keep the team intact, everybody stays. The frame was the failure. The real decision was what am I obligated to do under the terms of the program funding these positions, and what am I actually permitted to do. I never asked that question, so I never got its answer, and a few months later I was paying back more than twenty thousand dollars out of my own pocket. I answered my question correctly. It was not the question.

Notice the shape of the error, because it is common and it is quiet. My frame was generated by what I wanted to be true — that I could be good to everybody and lose nothing. Frames come pre-loaded with a preferred answer more often than not. So test them. Ask what decision are we actually making, and what is the decision behind that one. Ask what would make this the wrong question. Ask who framed it and what they wanted. And ask what constraints are real versus assumed, because half the constraints in any room are somebody's memory of how it was done last time.

Question two: what are the real options? The most common failure in organizational decision-making is not choosing badly among options. It is having only one. Research on decision processes has repeatedly found that a large share of consequential organizational decisions are evaluated as a single yes-or-no proposition rather than a comparison among alternatives, and that decisions considering multiple genuine alternatives tend to fare better. Whatever the precise numbers, any leader who has sat in enough meetings recognizes the pattern immediately.

And watch for the counterfeit version, which is worse than the honest one. A leader who has been told to bring options walks in with three: the plan he wants, an obviously inadequate cheap version, and something absurd nobody would sanction. That is not an option set. That is one option with two chaperones. The standard I want you to hold is simple and demanding — every option on the list must be one you would actually be willing to execute. If you would not do it, it is not on the list; it is decoration, and its presence makes the process feel rigorous while doing none of the work.

So the minimum is three live options, including the honest do-nothing baseline. Do nothing is a real option with real consequences, and it belongs on the list explicitly, because when it is unstated it becomes the default winner of every delayed decision without ever being evaluated. Then assign each option an advocate — a named person responsible for making its strongest case. This is the structured dissent from the last lesson doing its work upstream, before a favorite has been declared and the political cost of arguing against it has been set.

What Would Have to Be True, and Which Door Is This

Question three: what would have to be true, and what do we actually know? I prefer that phrasing over what does the data say, because it flips the direction of inquiry. Instead of assembling evidence for a conclusion — which is the confirmation machine from Lesson 3.3 running at full power — you specify the conditions under which each option would be the right one, and then go find out whether those conditions hold.

Then be honest about the limits, and there are three worth naming. First, sort your evidence: what is measured, what is inferred, what is somebody's assertion, and what is a vendor's or an advocate's claim. Those get different weights and they routinely arrive in the same PowerPoint at the same font size. Second, causation. Most of what leaders are handed is correlational, and the interesting question is almost always causal — will this intervention produce that effect. Module 9 goes after this properly. Third, the evidence you do not have. Every dataset was collected by somebody with a purpose, and the loudest signal in most organizations comes from wherever measurement happens to be easiest, not from wherever the truth is densest.

Question four: which door is this? Jeff Bezos gave leaders the most useful distinction available on this point in his letters to Amazon shareholders. Some decisions are one-way doors — consequential and effectively irreversible. Walk through and you cannot come back. Those deserve slow, deliberate, heavily scrutinized process. Most decisions are two-way doors: if you are wrong you can reopen it, walk back through, and try something else. Those should be made fast, ideally by the people closest to the work, and the cost of treating them like one-way doors is enormous and almost entirely invisible on any report.

His observation about organizations is the part I want you to sit with. As institutions grow, they tend to apply heavyweight one-way-door process to everything, which produces slowness, risk aversion, and the steady departure of the people who wanted to build something. Meanwhile — and this is the failure mode nobody puts on a dashboard — the actual one-way doors get walked through casually, because they did not come with a big meeting attached.

Learn to spot the real one-way doors in your context. A key hire is close to one-way; unwinding a bad senior hire costs a year and reputational damage in both directions. Public commitments are one-way, because Lesson 3.3's commitment pressure makes reversal expensive to your identity as well as your credibility. Multi-year contracts, debt, and capital commitments are one-way by construction. Anything that damages trust is one-way and then some, because trust repairs slowly and never quite to its former level. And in public office the list is long: appointments, labor agreements, and capital projects routinely bind a successor who has not been elected yet, which is a stewardship question before it is a management one.

Two-way doors are the other half of the discipline and the half leaders neglect. A pilot in one district. A reversible pricing change. A trial with a stated end date. Part of skilled leadership is converting one-way doors into two-way doors before you decide — piloting rather than launching, leasing rather than buying, hiring on a defined term, writing an exit clause. If you can find a way to make the decision reversible, you have not just reduced risk. You have bought the right to learn from being wrong, which is the most valuable thing a leader can own.

It Is Eighteen Months From Now and This Failed

Question five is the pre-mortem, and Gary Klein gave us the cleanest version of it. Before the decision is final, gather the people involved and tell them this: it is eighteen months from now, we did this, and it was a disaster. Everyone has ten minutes to write down why. Not what could go wrong — what did go wrong. Past tense, stated as fact.

The grammar is doing real work, and this is why I want you to run it exactly as written. The underlying research is on prospective hindsight — Deborah Mitchell, J. Edward Russo, and Nancy Pennington found that imagining an event as having already occurred and explaining why improved people's ability to generate concrete reasons compared with reasoning about it as a possibility. Asking what might go wrong invites vague, hedged, socially safe answers. Asking why it failed, with the failure stipulated, gives everyone permission to be specific without being the person predicting doom.

That permission is the political function of the exercise, and it matters more than the psychology. In most rooms, raising a serious objection late is a costly act — you are the obstacle, the pessimist, the one who was not on board. The pre-mortem removes the cost by assigning the pessimism to everyone simultaneously. Klein's own framing is that it legitimizes doubt. I would put it more bluntly: it lets the person who has known for three weeks that this will not work say so without spending capital he cannot afford.

Run it properly. Everyone writes privately before anyone speaks, or the first speaker anchors the room. Go around and collect one item per person before discussion, which surfaces the quiet people. Cluster what you get, and for the top failure modes decide now what you will do — mitigate, monitor with a named trigger, or accept openly. And do it before the decision is final, not after, because a pre-mortem run on a decision that has already been announced is a therapy session.

Question six: decide, and record it. Someone owns the decision by name, there is a date, and the date is written down. Then write a decision record while it is fresh — what we decided, what we expected to happen and by when, what our main uncertainties were, and what evidence would tell us we were wrong. Half a page. The reason for the record is that human memory of past reasoning is unreliable in a specific direction: once you know how something turned out, you remember having expected it. Without a written record, your review in a year is a conversation between people who all now believe they saw it coming.

Question seven: review the process, not the outcome. This is where ego has to be left outside the room, and it is the step almost everyone skips. Aristotle drew the distinction we need: deliberation concerns what is within our power and uncertain, and the virtue involved is deliberating well — which is not the same as things turning out well. Bernard Williams and Thomas Nagel developed the modern version under the heading of moral luck, the uncomfortable observation that we judge people by results that were partly outside their control. Annie Duke gave the practical name for the error, which is resulting: evaluating the quality of a decision by the quality of its outcome. A good decision is a sound process run on the information available at the time. A good outcome is what luck and the world delivered. Reward only outcomes and you will teach your organization to gamble and to hide, because the people who ran a careful process and got hit by variance learn quickly that carefulness is not what pays here.

The Calendar Decides If You Don't

There is one more thing this structure has to survive, and it is not analytical. It is the pressure to not decide. The most common sentence in a room that has already understood the problem is we need more data. Sometimes that is right. More often it is avoidance wearing the language of rigor, and you can tell the difference with one question: what specific information, obtainable by what date, would change which option we pick? If nobody can answer that, more data is not a research plan. It is a way to postpone owning a hard call while sounding responsible.

The reason this matters so much in executive work is that delay is not neutral. Delay is a decision — usually a decision for the status quo, made without evaluation, by default. And delay consumes reversibility. The longer you wait, the fewer options remain, and the ones that remain are more expensive and more likely to be one-way doors. This is the single most important operational point in the lesson, so let me say it plainly: deciding early is usually deciding cheaply. Deciding late is usually deciding under duress, and duress is where the Five Voices are loudest.

New York learned this at scale. Through the late 1960s and early 1970s the city financed operating deficits with short-term borrowing and accounting practices that made the books look survivable — recognizing uncollected revenue, moving expenses into the capital budget, rolling notes forward and calling the result a cash-flow issue. Nobody in that period had to make a catastrophic decision. That is precisely the point. The catastrophic decision was the accumulated absence of one. Then in the spring of 1975 the banks stopped underwriting the city's notes, and a problem that had been framed for years as liquidity revealed itself as structural. The state created the Municipal Assistance Corporation that June and an Emergency Financial Control Board that September, which moved effective fiscal control out of City Hall. Union pension funds bought city-backed bonds. Federal help came only later. And the options that remained were all one-way doors: tens of thousands of city workers laid off, deferred maintenance that the subway system paid for through the following decade, a fare increase, and the end of CUNY's long-standing free tuition. Mayor Beame, the city's comptroller before he was its mayor, did not face those choices because he chose badly among alternatives. He faced them because by 1975 the alternatives were gone.

Luke 14:28-30 puts the same discipline in a sentence, and Jesus applies it to counting the cost before you begin: 'For which one of you, when he wants to build a tower, does not first sit down and calculate the cost to see if he has enough to complete it? Otherwise, when he has laid a foundation and is not able to finish, all who observe it begin to ridicule him, saying, This man began to build and was not able to finish.' Note what He is not warning about. He is not warning about failure. He is warning about starting a thing whose full cost you never priced — which is the framing question and the reversibility question fused into one image.

And Proverbs 15:22, in the NASB, reads: 'Without consultation, plans are frustrated, But with many counselors they succeed.' I want you to see that wording, because the rendering people usually quote — plans fail for lack of counsel — comes from other translations, and the NASB's 'without consultation' is doing something slightly different and more useful. Consultation is an act you perform, not a resource you happen to have around. The verse is not saying that having advisers helps. It is saying that the process of consulting is what keeps plans from collapsing, which means a leader surrounded by capable people he never actually consults is described by the first half of that verse, not the second.

So put this on your calendar rather than in your intentions, because a leader who runs a good process only when the pressure is low does not have a process. A mayor's year has a fixed spine — a preliminary budget in the winter, an executive budget in the spring, and a budget adopted before the fiscal year begins on July 1 — and every one of those dates is a forcing function that will make decisions for you if you have not made them yourself. Know also which decisions are not yours to make: a mayor does not run the MTA, and school governance depends on Albany, so on transit and schools the real decision is almost never what will I do but what will I fund, negotiate, and publicly press for. Framing that honestly is question one applied to your actual authority, and getting it wrong in public is how candidates make promises they cannot keep. That is not merely bad politics. It is a false representation of your own power to people counting on it.

Here is the question I want you to carry into Lesson 3.5, where we start looking at systems rather than single choices. Take the decision in front of you right now. Which of the Seven Questions have you actually run — and which one are you skipping because you already know what you want the answer to be?