Writing the Decision

The incumbent rule never has to be written up; it runs because nobody stopped it. A challenger runs only once someone with standing commits to it, on the strength of a document read in a few minutes. The write-up is therefore part of the decision product, and a challenger that cannot be read quickly will lose to a rule that does not need to be read at all.

Every written argument has a shape, whether or not the writer chose one. If you do not supply the hierarchy, the reader builds one, and the one they build will not be yours. Most bad technical documents are not wrong; every sentence in them appears equally important, so the reader grades the effort instead of the argument.

The page holds five answers

The one-page decision is the five questions answered in the order a reader needs them. The recommendation answers the third question, what could replace the rule, and it comes first, because the commitment is what the page exists to obtain. The first two answers, the incumbent as it runs and who can change it by when, travel in the header and the opening sentences. The alternatives sit beneath the pick, each a complete strategy over the same horizon: at Meridian Motors, keep the current rule; take every item down 20 percent by the freeze and then run the current rule at targets 20 percent lower; or replace the rule with classes. The fourth answer, what would make the challenger lose, supplies the reasons and the flip conditions. The fifth supplies the rule: owner, cadence, exceptions, kill switch, revisit date. Without the rule, the process runs on the day after the meeting what it ran the day before.

The shape that works

One governing thought at the top, a small number of ideas that support it, and beneath each of those the ideas that support them, down to queries and code. An idea at any level must be a genuine summary of the ideas grouped beneath it, not a label for the pile, and ideas in a group must be the same kind of idea, in a deliberate order.

The thought at the top must be a decision, not a finding, and that restriction does real work. “Inventory is too high because of A, B, and C” is a complete thought, and it still cannot be installed at the freeze. “Adopt a three-class stocking rule at the April freeze” provokes a different question, why this alternative rather than the others, and only reasons that discriminate between alternatives can answer it. Facts that are merely true stop qualifying for the page.

Opening so that the pick lands

Leading with the pick does not mean beginning cold. The pick has to arrive as the answer to a question the reader is already holding, and the opening paragraph, three or four sentences, exists to put that question there. Begin with the situation, something you and the reader already agree on. In a decision document it is usually the incumbent: Meridian Motors runs a single replenishment rule across all parts on a common service target. Then name the complication: finance has asked for a 20 percent reduction in parts working capital, dealer operations has asked for 98 percent fill, and the current rule cannot deliver both. Together those raise a question, and your recommendation is the answer.

Everything in the situation must be uncontroversial, because an arguable claim in the runway surfaces as an objection at the worst moment. Write the question out as an explicit sentence: it is where you discover that the evidence answers a different question.

Headings have to carry the argument

Write the headings before the prose. Full sentences, each an assertion.

“Benchmark results” is a filing label. “The misses are concentrated in 8 percent of items, and those items strand repair jobs” is a claim, and it does the reader’s work for them.

Then read the headings in sequence, ignoring the body. They must add up to the recommendation. If they read as a table of contents, rearrange them, because six sentences are cheap to reorder and six drafted sections are not.

The first level under the pick is the whole story

The reasons immediately beneath the recommendation are the argument. That is why two to four is the right number and seven is almost always a failure of grouping.

Put the flip conditions beside the reasons rather than in a closing list of caveats; caveats that cannot change the ranking belong in an appendix, or nowhere. Read the first level aloud without the numbers. If it sounds like a case, you have a pyramid. If it sounds like an agenda, keep grouping.

Reasons that look parallel may all have to hold

Independent reasons each favor the recommendation over its rivals on their own, and together make the case stronger than any one does. Meridian Motors’ decision not to rebuild the forecasting platform this cycle has this shape: even a large accuracy gain would not move the items that hold the cash across class boundaries; the platform could not be delivered before the April freeze; and the people who would build it are the people who must install the class rule. Any one of those, true while the others are false, still argues against the rebuild. Prefer this arrangement wherever the evidence allows, because it survives one reason turning out weaker than it first appeared.

Jointly necessary premises look the same on the page and behave nothing alike, because independence cannot be established by layout. The reasons for the class rule itself are of this kind. The cash and the misses sit in different items; the current rule cannot distinguish them; bypass of the protected class can be closed. Set side by side they read like three votes in favor. They are the conditions that must all hold for classes to win, and losing any one does not weaken the recommendation; it removes it. Each premise needs its own evidence, because none can borrow strength from the others, and each reappears among the flip conditions. The write-up should say plainly that the case rests on all of them, rather than present three votes and overstate its margin.

A chain runs step to step, with the recommendation as the last link. It reads as a proof when it holds, and an objection to any link takes down everything after it. Know which link is weakest, because that is the one that will be attacked.

The test takes a minute. Remove each reason in turn and ask whether the argument still stands. If it does not, say so on the page, naming the premise and what would show it false, or go find independent support. Most real write-ups mix the arrangements, which is fine provided you know which one is in use at each point.

Exhibit 1Delete the second reason and watch what happens

One decision at every length

The same pyramid is read at different depths by different readers, and each depth must imply the same ranking. Thirty seconds is the recommendation sentence. Three minutes adds the alternatives, the reasons, and the flip conditions. Fifteen minutes adds the economics, the work deliberately skipped, and the rule in operating detail. If the vice president, the finance partner, the planner, and the modeler would leave with different commitments, the work contains two decisions, and it has to be fixed before the document can be.

Walk it around the building first

Do not let the meeting be the first time the people with standing hear the ranking. Take the one-page version to the owner, the finance partner, and the people who will run the rule, separately, while there is still time to change it, and find out whether any objection is a flip condition you missed. Show the operators the blanks while they can still amend them. A parameter nobody helped specify gets padded on Tuesday. You still own the ranking. They own Tuesday.

What to cut

Cut any sentence that cannot change the ranking, the risk of the ranking, or the cost of running the rule. That removes the exploratory cuts that went nowhere, the chronology of the project, and the chart that is in the deck because it was made.

A third independent reason can be cut if two already carry the ranking. A premise the case jointly rests on never can. If the page is empty after cutting, the work has not yet produced a decision, and the remedy is to go back to the choice rather than to draft again.