Shortest Path to Disproof

Kill the conclusion with the cheapest test, before you build on it

My conclusion is: [conclusion]. Do one thing only: find the shortest path to showing it doesn't hold. Requirements: 1) Give 1–3 of the cheapest tests, each doable within [time/budget]; 2) For each test, state what result would mean the conclusion fails; 3) Rank them by lowest cost and highest chance of a clean veto; 4) Name the test that, once run, makes the rest unnecessary; 5) If I can run only one, which one. Do not argue why it's true.

How to use

Write down what result would count as failure before you ask. Leave that blank and you get a generic checklist instead of a path.

Good for

You have one smooth-sounding conclusion and have to decide this week whether to keep funding it — half a day spent breaking it is cheaper than another round of validation.
Someone is winning you over with a polished industry report or a competitor's numbers, and you can't audit the logic but you can find out whether the conclusion collapses first.

Not for

The conclusion hasn't taken shape and you can't yet say what result would disprove it — you'll get a generic list of tests you can't act on.
You're already convinced it's true and need an execution plan and a timeline — this step only slows you down; go to plan review.

Example

Input

We're cutting the support team from 12 to 7 on the strength of a new AI ticket router that should absorb 60% of inbound volume. Budget is $400,000 over two months.

What you get

The shortest path isn't model accuracy; it's the share of existing tickets answered in one sentence — if only 35% qualify, a 60% deflection target has nothing to deflect and you'll be rehiring within two months. The second test is whether 7 people survive the three days around month-end billing, when volume runs 2.4× normal. Both finish inside two weeks, and if they fail, the $400,000 never gets spent.

Common mistakes

Reading “shortest path” as “find any counterexample”: one mild counterexample arrives and the conclusion gets declared broken, even though it sits inside normal variance.
Pasting the conclusion with no numbers and no constraints — it can't tell which test is cheapest, so the path it hands back costs more than the plan.

Try tweaking it like this

Swap “shortest path” for “cheapest single veto”: keep only the test that could end the project outright and delete the rest.
Add “list the three premises this conclusion depends on” — disproof becomes a line-by-line audit of the premises.