In progress. Expected 2027.
The Decision Product
Turning Analysis into Decisions the Organization Can Run
The book is being written in the open. Subscribers get chapters as they are drafted and are told first when it is available.
The Decision Product
Turning Analysis into Decisions the Organization Can Run
The argument, as a pyramid
What the book claims, and what holds it up
A book about structured argument should be willing to show its own structure before you buy it. This is the whole thesis, with its supports.
Governing thought
Analysis that cannot change which alternative is chosen is decoration, and accurate decoration is still decoration.
01
A decision has three parts, and a study is what you get when one of them is missing.
You specify the choice: the alternatives including the status quo, who can commit, by when, and whether this fires once or repeats. You assemble the evidence that could change the ranking of those alternatives. You specify the rule: what the operating process does now, who runs it, and when it is revisited.
02
The work runs backward and the presentation runs forward, and technical teams routinely reverse both.
Backward means starting from the choice and finding the cheapest evidence that could change your mind. Forward means opening with the recommendation. The default is chronological in both directions: extract, clean, model, caveat, and conclude if there is time.
03
The test is whether the organization can make the decision again without you in the room.
The work is not finished when the analysis is correct. It is finished when a named owner runs a specified rule on a stated cadence, with a priced exception path and a date to revisit.
Fit
Who this is for, and who it is not
Written for
- Operations researchers and decision scientists whose work is supposed to change what the company does
- Supply chain and logistics analysts who own a rule somebody runs on an ordinary Tuesday
- Data scientists and machine learning practitioners tired of watching correct work die in a meeting
- Technical leads who have to coach that work, and product managers close enough to the math to see when it is decorative
Not written for
- Anyone learning SQL, Python, or statistics. There is almost no notation in it
- Readers wanting the mathematics of sequential decisions under uncertainty, or formulation tricks for mixed-integer programs. Those are different books, including ones the author has already written
- Readers looking for a visualization manual, slide design, or a textbook of decision theory
- Readers who want reassurance that being right should be sufficient. The book argues it is one term in a product, and does not soften that
Outline
Table of contents
Three parts, eighteen chapters, following the method: name the choice, earn the evidence, get it running. Every chapter links to its free method entry. The entry carries the mechanics; the chapter carries the cases, the objections, and the organizational politics a reference page cannot hold.
Part I
Name the Choice
Most requests arrive as a topic with a deadline attached. Before any analysis is worth starting, the topic has to become a sentence somebody can accept or reject, and the menu of alternatives has to be invented rather than inherited.
- 01
The Analysis Is Not the Product
What the default week actually produces, why labor gets mistaken for rigor, and what changes when the unit of work is a decision rather than a study.
- 02
What Decision Are We Actually Making?
A complete choice is a sentence, not a topic. Alternatives of the same kind, an owner rather than a requester, a freeze date rather than a meeting time, and the uncertainty you will still have when you decide.
- 03
Problems, Decisions, and Outcomes
The KPI sandwich, why observations are not diagnoses, and what splitting the three apart buys you.
- 04
Invent the Alternatives
The email is a first draft of the menu. Four habits that reliably enlarge a list that was too small, why you kill most of them on purpose, and why the missing option is usually a combination.
- 05
Refuse the First Brief
Four questions to ask before you spend serious time, when not to analyze at all, and how to refuse a brief without becoming the problem.
Part II
Earn the Evidence
Work runs backward, from the choice to the cheapest test that could change your mind. Evidence that cannot change the ranking of the alternatives is decoration, and accurate decoration is still decoration.
- 06
Work Backward from the Choice
Write the candidate ranking first, then the claims that could kill it. Data last. The ways you will fool yourself, and the answer to "we might miss something."
- 07
Cut the Question Into Parts
The tree, the two properties that make a cut usable, how each of them fails, and why pruning matters more than completeness.
- 08
Complexity Must Pay Rent
The ladder from a count to an optimization model. Climb only when a lower rung cannot settle a load-bearing claim. Heuristics are first-class.
- 09
Economics Before Metrics
Price the error. Which house metrics are pretending to be constraints, when a technical metric earns its keep, and how to rank when the units will not reduce.
- 10
What Would Change the Ranking?
Decision-relevant uncertainty, ranges and flip points, value of information as a delay tax, and what the choice costs to undo.
- 11
Locally Optimal, Globally Stupid
Local scoreboards, feedback you will pretend not to see, why the system oscillates, and the question of who undoes your recommendation if it works as designed.
Part III
Get It Running
The presentation runs forward, from the recommendation. Then the harder half: the people who can undo it, what each of them does if you fail, and the five parts of a rule that has to fire on an ordinary Tuesday.
- 12
Diagnosis, Then a Recommendation
From observation to mechanism, why teams refuse to recommend, and the legitimate case for recommending the status quo.
- 13
Writing the Decision
Lead with the pick. The opening and the pyramid are one structure. Headings that carry the argument, two shapes of support, layers rather than dumbed-down copies, and what to cut.
- 14
Interests, Not Positions
Positions feel like requirements. When the position is real, when interests genuinely conflict, what trust is made of, and why being right is not enough.
- 15
What They Do If You Fail
Every party has a walk-away, whether or not they have said it. No-deal is an alternative, it belongs on the menu, and a ranking that ignores it is a ranking of a meeting rather than of a company.
- 16
Disagreement Is Not a Request for Another Notebook
Name the fight, agree the criteria before arguing the answer, let people change their minds cheaply, and handle the objection you cannot answer.
- 17
What a Repeating Rule Is Made Of
Five moving parts: the state, the decision, the information that will arrive, the transition, and the objective. Most analysis specifies two of them and discovers the other three in production.
- 18
A Decision That Cannot Operate Is Unfinished
What has to be specified, what the rule can know at the moment it fires, how rules die, and why legibility is a feature rather than a concession.
Back matter
After the argument
- —
The Job
What you were actually hired to do, and the three failure signatures.
- —
Appendix A: The Working Sequence
Fourteen steps, in order, collected so you can find them on a Monday.
- —
Appendix B: The Write-up
The eight-section list. If a section is empty, you are not ready to present.
- —
Appendix C: Meridian Motors Parts
A completed example, worked end to end.
- —
Further Reading
Annotated. They influenced the cooking; they are not the meal.
This is a working manuscript and it is not for circulation, quotation, or sale. Chapters are being reordered as the material is tested against real engagements, and the through-line case is a composite built from patterns across automotive, logistics, retail, and consulting work rather than an account of any single employer or client.
Source: Working manuscript, September 2026.
Before the book
Everything structural in the book is already free on this site
The eighteen method entries, the seven worksheets, and the diagnostic cost nothing and are not abridged. The book adds the cases, the objections answered at length, and the running Meridian Motors example worked end to end.