The canon
The method
Twenty entries, following the fifteen chapters of the book. Each gives the definition, the mechanics, a worked example drawn from operations, and the failure mode that most often breaks it. Nothing here is gated, abridged, or held back.
Read in order, or take what you need
Part IFind the Real Decision
- 01The Analysis Is Not the ProductChapter 1The deliverable is a challenger rule that beats the incumbent, gets installed, and can be stopped, not an explanation of what happened.The founding claim. Every decision you are asked to analyze already has a rule running, and the analysis competes with it. The product is a challenger that beats that incumbent, gets installed where it ran, and carries a way to tell whether it was wrong. Five questions define it, and the answers fit on one page.
- 02Every Decision Already Has a RuleChapter 2The incumbent is the rule that actually fires, not the one in the manual. Every recommendation is a comparison against it, whether or not the write-up says so.Before you analyze anything, describe the rule that is already deciding. The incumbent is the documented rule plus the spreadsheet, the pad, the old policy, the default, the habit, and the person who just knows. It sets the baseline, lists the commitments a challenger must keep, and is better read as information than as a defect.
- 03Why the Incumbent WinsChapter 3Being better on the merits is necessary and not sufficient; a challenger must be designed so the forces favoring the incumbent stop working against it.A challenger that is better on the page still loses more often than not, for reasons that have little to do with its quality. Defaults, loss aversion, the missing defender of change, incentives, timing, and diffuse ownership all favor the rule already running. A challenger that intends to win is designed against each of them.
- 04Name the ChoiceChapter 4A single sentence naming who will choose among which alternatives, by when, and to what end, knowing what will still be uncertain.Naming the incumbent says what a challenger must beat, not which decision it competes in. A complete choice is a sentence, not a topic. It carries a person who can commit, a date the choice goes stale, alternatives that are complete strategies of the same kind including the incumbent, and the uncertainty that will still be there when it is made.
- 05Problem, Decision, OutcomeChapter 4The separation of what is wrong, what someone can change, and what should improve as a consequence, joined by a stated mechanism that can be wrong.Three different objects that arrive mashed together in almost every brief, plus a fourth, the diagnosis, that hides among them. Splitting them inside the choice sentence and writing the mechanism that connects them is cheap, unglamorous, and resolves more confusion than any other single move.
- 06Refuse the First BriefChapter 5The practice of returning an underspecified request as a proposed choice the owner can correct, rather than absorbing it into analysis.Read the incoming request for the decision it contains and the decision it leaves out, get four answers before you spend serious time, and send back a better object the owner can accept, correct, or overrule. Refusal here means translation, not insubordination.
- 07Invent the ChallengersChapter 6Enlarging the set of challengers before ranking any of them, because a decision is usually more sensitive to the menu than to the mathematics that ranks it.The status quo on the menu is the incumbent rule; everything else is a challenger, and a challenger nobody invents cannot win. Alternatives are made, not given, and no amount of evidence rescues a list that never contained the commitment you should have considered.
Part IIEarn the Recommendation
- 08Work Backward from the ChoiceChapter 7Starting from a provisional answer and finding the cheapest evidence that could make the challenger lose to the incumbent.Write the candidate ranking before you open a notebook, then the claims that must hold for the challenger to beat the incumbent, then the cheapest test that could kill each one. Data last.
- 09Cut the Question Into PartsChapter 7Splitting a question into branches that neither overlap nor leave a gap, then dropping every branch that cannot change which alternative beats the incumbent.Draw the tree, check that the branches neither overlap nor leave a gap, size a quantity by partitioning it rather than by adding up its drivers, then mark every branch that cannot change the ranking and leave it alone. The pruning is the point.
- 10Economics Before MetricsChapter 8Ranking the incumbent and its challengers by the cost of being wrong rather than by movement in an internal measure.Put consequences in units that bite. House metrics are instruments, not consequences, and a metric treated as an objective will quietly decide the ranking for you. When the units will not reduce, put them on the same clock and state the exchange rate as a sentence someone can contest.
- 11Who Undoes This?Chapter 8The check that a recommendation survives the responses of the people it makes worse off, and the second round of the system it lands in.Locally optimal, globally stupid. Evidence priced correctly for one function can still recommend a rule a neighbor will undo. Write the second round into each alternative, find the adjacent team that will dislike the challenger if it works as designed, and go ask them what they will do about it.
- 12Complexity Must Pay RentChapter 9The principle that added method must be justified by a specific claim the simpler method could not settle, in the conditions where the rule will actually run.A ladder from a conversation on the floor to an optimization model. Start at the bottom and climb only when a lower rung cannot settle a claim the ranking depends on. A method pays rent only if it moves the ranking, its risk, or the cost of running the rule. Heuristics are first-class, and a heuristic facing tightly coupled constraints has to prove it is feasible.
- 13What Would Make the Challenger Lose?Chapter 10The identification of the specific conditions under which you would recommend a different alternative, each labeled by when it can be resolved and who holds it.The question that does more work than any framework, because it cannot be answered with a chart. Name the worlds in which the challenger loses to the incumbent or to another challenger, mark the threshold for each, and say whether it is a reason to stop analyzing, a trigger to stop the rule, or a judgment left for the owner.
Part IIIGet a Commitment
- 14Writing the DecisionChapter 11Presenting a decision in the order a reader can act on, which is the reverse of the order the work was done.The write-up is how a challenger asks to replace a rule that never had to be written down. Five answers on one page, the pick first, an opening that puts the question in the reader's head, headings that assert, and reasons checked for whether they are independent support or premises that must all hold.
- 15Diagnosis, Then a RecommendationChapter 11A stated causal mechanism followed by an actual commitment among the alternatives, written so that an owner could reject it.Ask what machine would produce these observations as its normal output, then commit to one of the named alternatives in a sentence that can be refused. A diagnosis makes a prediction; a story does not. The approach without the actions is a slogan, and recommending the incumbent deserves the same sentence as recommending a change.
- 16Interests, Not PositionsChapter 12A stated demand is a position; the concern it protects is the interest. Rank on the interest, in units of consequence, and keep the demand off the constraints.A position is what someone says they want. An interest is what they are actually protecting, often a piece of the incumbent rule. Arguing against a position that stands in for an unstated interest cannot succeed, and while positions deadlock, the incumbent wins by default.
- 17What They Do If You FailChapter 12What each party will do if the decision produces no commitment, treated as an alternative on the menu, with the incumbent as the room's shared fallback.Every party in the room has a fallback, whether or not they have said it, and the owner's is usually the incumbent rule. Those fallbacks are the alternatives that remain if your work produces nothing, and a ranking that ignores them ranks a meeting rather than a company.
- 18Disagreement Is Not a Request for Another NotebookChapter 13Handling objection by classifying and resolving the actual disagreement rather than by producing additional evidence that cannot change the ranking.When the room pushes back, the instinct is to go do more analysis. That instinct is usually wrong and always expensive, because the incumbent keeps running while you work. Name the fight, agree the criteria, give a disagreement about standards to its owner, and let people change their minds cheaply.
Part IVMake It Survive
- 19Design the RuleChapter 14A repeating rule specified as five parts plus a default, a priced exception path, and a closed loop, so following the challenger is the easiest thing to do.A challenger has to inherit the advantages the incumbent holds. That takes five named parts (the state it can see, the decision it may take, the information that will arrive, the transition to the next state, and what it accumulates) and three behavioral settings: a default it owns, exceptions that cost a sentence, and feedback the operator can see.
- 20Install the RuleChapter 15The specification and install design that make a challenger the rule the process runs next cycle, and keep it running after its designers leave.The incumbent does not stop when a meeting adopts its replacement. It fires again on the next run until something specific makes it stop. A decision is finished when the challenger is what fires, with an owner, a priced exception path, a staged install with a kill switch, a scorecard under which the pad stops paying, and alarms for the day the rule decays.
How to use this
The entries follow the book's order, which is the order the work runs in: find the real decision, earn the recommendation, get a commitment, and make it survive. Read straight through the first time.
If you have one live brief to fix this week, the order is different. Start with the rule already running and the choice sentence, then the candidate ranking and its kill conditions, then the flip point. Those account for most of the available improvement on any single piece of work, and all of them happen before you open a notebook.
If the same problem keeps recurring across projects, it is usually the brief you accepted or the rule you never specified. Those are the two ends nobody owns, and they are where good work is most often lost.
The toolkit turns the entries into worksheets you can fill in, and the diagnostic will tell you which of the three parts is missing from your own work.
Beyond reading
Reading the method and installing it are different problems
Coaching works on your live documents rather than on exercises. Team engagements work on the system that produces them.