Walk them. Maturity assessment: positions the company on a scale against peers, terminates nothing — a score is compatible with any subsequent action, which is why it is safe to produce. Opportunity register: twenty-odd initiatives, sized and ranked; feels like a decision because it has numbers and is the most reliably useless of the four, because a ranked list moves the decision downstream intact — somebody still has to say no to items four through twenty and nobody in the document did. Roadmap: sequences the register, and sequencing assumes the funding decision was already made, so the roadmap is downstream of a decision nobody took — which is why roadmaps get rebuilt every two quarters. Implementation plan: presupposes all three above. Not one contains a sentence of the form 'this option is not being pursued, for this reason, and here is what would have changed our answer'. Four hundred thousand dollars of genuinely competent work, none of it decision-grade.

Two forces, and neither is anyone acting badly. It's safer to supply: a decision-grade deliverable makes a falsifiable claim with the provider's name on it, an analysis-grade one makes none — if the client later funds the wrong thing, the assessment was still accurate. Providers optimise for defensibility over time and the equilibrium is description. And scope-of-work mechanics produce it: advisory work is scoped and priced on effort — weeks, workstreams, interviews — while a decision-grade output is frequently shorter, which looks like less value for the same fee. The commercial structure of the engagement selects against the thing you wanted. Macaulay's 1963 study found businessmen frequently don't plan exchange relationships completely and seldom use legal sanctions to settle disputes; planning is used when gains are thought to outweigh costs, and the decision whether to formalise is made by whoever's occupational role bears the exposure. Same structure: the party who bears the cost of an indecisive deliverable is not the party who scopes it.

Of the three parts of the term's test, the first two are easy to demand and get. The falsifier — what evidence would have reversed the call — is the one nobody writes, and it is the one that makes the other two worth anything. Without it you cannot distinguish a recommendation from a preference; both look identical on the page. Six months later, when the situation has moved, you cannot tell whether the call still holds because you never recorded the conditions it depended on — so the work has to be redone, which is why funds buy the same assessment twice. With it you get something that behaves like an asset: a dated claim with a stated dependency, which can be checked. That is the difference between a deliverable and a residual, and it is the property that lets a second portfolio company reuse the first one's work rather than commissioning it again. Demanding the falsifier costs nothing at scoping time and is nearly impossible to add afterwards.

The uncomfortable commercial fact, stated because it is the reason this is a hard sell in both directions. A decision-grade engagement is usually smaller, shorter and cheaper than the analysis-grade equivalent, because most of the effort in a large advisory engagement goes into breadth — covering the option space so no criticism lands. Terminating a decision requires depth on two or three options and the discipline to exclude the rest. Which means the buyer's procurement instinct — more weeks, more coverage, lower risk — reliably selects the deliverable that cannot decide anything, and the provider has no incentive to correct them.