The Two Questions That Do More Than Any Framework

Methods accumulate. Every few years a technical organization adopts a new one, produces templates, runs a workshop, and reverts within a quarter. The reversion is not a failure of discipline. It is what happens when a method adds vocabulary without adding a constraint.

Two questions survive that cycle, in my experience, because neither can be satisfied by producing more of what you were already producing.

What would change the ranking?

Most uncertainty in an analysis is irrelevant to the decision, and treating all of it as equally worth reducing is how a six-week project becomes a six-month one.

The uncertainty that matters is the uncertainty that could change which alternative wins. Everything else is scenery, however large the error bars.

Asked of a finished analysis, the question is a test. Asked of an unstarted one, it is a workplan, because the answer tells you exactly what to go measure and, more usefully, what not to.

The answer has a shape. A specific quantity, a specific threshold, and what it would cost to find out:

We recommend a class-based replenishment rule. That ranking survives if missed-job cost on the protected class is anywhere from roughly half to twice our base case. It breaks if dealers can bypass the system on the protected class, or if finance forbids item-level exceptions.

Notice what is absent: a fabricated eighty-seven percent confidence. Confidence belongs to claims, and even then as language rather than as jewelry. “We are willing to run this for a quarter” is a real sentence about operational confidence. A p-value completes a different thought, about a different object, for a different audience, and it is almost never the reason a vice president should change a rule.

A one-way table is usually enough. Rows are the inputs that actually argue, with the value marked at which the recommendation changes. It does something a paragraph cannot: it lets people argue about the inputs, and it lets them see, sometimes for the first time, that they are not arguing about your personality.

What does Tuesday do now?

The second question attacks the other end.

A recommendation that cannot be answered in terms of what an ordinary Tuesday does differently has not reached an operating process. It may be correct, it may have been agreed, and it will change nothing.

Asking it early is uncomfortable, because it frequently reveals that nobody knows who would own the resulting rule, which is a question people prefer to defer until after the analysis is impressive enough to carry it. Deferring does not make it easier. It makes it a surprise in week ten.

Asking it late is still worth doing, because it is the difference between a decision and a good meeting.

Why these two and not a framework

Both questions are cheap, which matters more than it sounds. A method that requires an hour of setup will be skipped under deadline. These take a sentence each and can be asked in a corridor.

Both are unanswerable with a chart, which is the property that makes them resistant to the usual response. “Let me pull some more data” does not answer either one. The first requires a threshold and a price. The second requires a name and a date.

And both are askable by people who are not the analyst. This is the part that compounds. A director who asks what would change the ranking, every time, changes the behavior of an entire function within about two quarters, without running a single workshop. A method that only works when the practitioner is disciplined is a method that works occasionally. A question that anyone in the room can ask is an institution.

The honest limitation

Neither question helps if the choice itself was never named. “What would change the ranking” presupposes a ranking, which presupposes alternatives, which presupposes somebody who can commit.

Asked of a brief that is still a topic with a deadline attached, the first question produces a blank stare, and the blank stare is the finding. It means the work has not started, whatever the project plan says.

That is a useful failure mode for a diagnostic question to have. It fails loudly and early, which is the only kind of failure worth designing for.