Judgment Playbook

Falsifiability

Turn a confident claim into one that names the evidence capable of proving it wrong.

5 minute readtaught in course chapter 4
EvidenceUncertainty

What it is

Falsifiability is a property of a claim: some result we could actually observe would count against it. A falsifiable claim rules things out in advance, so reality has a way of contradicting it. As a working method it turns a confident recommendation into a claim with a named exit: the observation that would break it, the threshold, the date, and the action that follows. It tests the shape of the claim, not yet its truth.

Where it comes from

Karl Popper (1902–1994) introduced falsifiability in Logik der Forschung, published in Vienna by Julius Springer in 1934 under a 1935 imprint. The English version, The Logic of Scientific Discovery, appeared in 1959. Popper wanted a line between empirical science and doctrines that only look empirical, and he found it in what a theory forbids. In the 1953 Cambridge lecture reprinted as the first chapter of Conjectures and Refutations (1963) he wrote: "A theory which is not refutable by any conceivable event is non-scientific. Irrefutability is not a virtue of a theory (as people often think) but a vice." The business use came much later and has no single author. Popper wrote about theories, not about strategy memos, and he insisted that falsifiability is a logical property, while an actual refutation always involves judgement.

What it corrects

The failure is not overconfidence. It is a claim built so that every outcome supports it. A team commits to a strategy on the belief that owning the customer relationship, or entering a market early, or hiring the best people, produces returns. Good quarters confirm the belief. Flat quarters mean the change needs longer. Bad quarters mean the company under-invested. Careful people make this worse rather than better, because they respond to doubt by gathering more evidence, and an elastic claim absorbs any evidence offered. Ordinary care raises the quality of the analysis while leaving the belief untouched. The bill arrives when the commitment is no longer reversible and the argument is still about interpretation.

How it works

  1. Write the central claim in one sentence with its mechanism: if we do X, Y changes because Z.
  2. State what the claim forbids. A claim that forbids nothing cannot be tested.
  3. Name the most credible rival explanation for the same result, and choose an observation that separates the two.
  4. Attach a number, a measurement and a horizon: what value, read how, by when.
  5. Prefer a source your own plan cannot flatter. A signal your team controls is not a test.
  6. Write the consequence beside the threshold now: stop, revise, or run a deeper test.

Worked example

General relativity, published by Einstein in 1915, predicted that starlight passing the edge of the Sun would bend by 1.75 arcseconds, about twice the 0.87 arcseconds a Newtonian calculation allowed. The prediction forbade the Newtonian value, and an eclipse fixed the date in advance. On 29 May 1919 two British expeditions photographed the eclipse, one at Sobral in northern Brazil, the other on the island of Príncipe. On 6 November 1919 Dyson, Eddington and Davidson reported 1.98 ± 0.12 arcseconds from the four-inch telescope at Sobral and 1.61 ± 0.31 from Príncipe. Both figures sat close to Einstein and far from Newton. The episode also shows the limit of the method: the larger 16-inch plates at Sobral were distorted by heat and set aside as unreliable, and historians still argue about that decision. A falsifier decides a question only as far as the measurement can carry it.

In a Business Case Weekly case

In the Nike case, the fork that needs this method is the June 2020 board meeting on Consumer Direct Acceleration. The plan rests on the claim that selling direct gives Nike control of demand and margin, and that claim can absorb almost any result: a strong quarter confirms it, a weak one means the transition needs more time. The case asks you for two tripwires measured outside Nike's own channels. Falsifiability is what turns those from reporting into tests, because each needs a threshold, a quarter in which it will be read, and a consequence agreed before the wholesale accounts are cut. Whether the plan deserves approval is yours to argue.

In your answer

  • "My recommendation rests on the claim that …, and I would be wrong if …"
  • "I would read … outside our own channels; below … by … I treat the claim as broken."
  • "A rival explanation for the same numbers is …; the observation that separates them is …"
  • "If that result appears I stop and renegotiate, rather than extend the timeline."

Common misuse

The counterfeit version is a falsifier that cannot fire in time or in fact. It arrives as a milestone: a metric the team can redefine, a review scheduled after the factory is built, a target so modest that any outcome clears it. One question separates them. Name the number that could appear next quarter and force you to change the plan, and say who reports it. If nobody can name the number, or only your own team can measure it, the memo has borrowed the vocabulary of a test without accepting one. The opposite error is treating a single bad quarter as proof. A result contradicts a claim only along with the assumptions behind it, so say which assumption you would give up first.

References

Opportunity cost & trade-offs

Next method

Make the recommendation name what it gives up, not only what it hopes to gain.

Open next tool