How to write falsifiable investment thesis conditions
A falsifiable investment thesis states an observable condition that would contradict a necessary assumption. It should tell you more than “review if results disappoint.” Define the business mechanism, the measurement and the time allowed for it to appear before choosing a numerical threshold.
Separate a warning from a failure
A warning asks for investigation. An invalidation condition says the original explanation no longer works on its stated terms. A share price decline alone cannot usually distinguish deteriorating operations from a change in market valuation. Likewise, a strong price does not repair a failed business assumption.
Start from a necessary condition: a distributor’s expansion requires new branches to retain customers after introductory discounts end. The test should therefore observe mature branches after the discount period, not group revenue during openings. The 10-K’s risk and operating discussion provide useful places to identify such dependencies; the invalidation rules remain your own documented choices.
Hypothetical test register
Assume five new branches have completed a twelve-month introductory period. Your original note expects repeat customers to support the next six months of activity. Use this register before their results arrive.
| Condition | Classification | Research response |
|---|---|---|
| A branch reports late | Data gap | Check the reporting boundary; do not score it as zero |
| Repeat sales weaken in one month | Warning | Examine seasonality and contract timing |
| Mature comparable branches lose repeat customers throughout the agreed six months | Invalidation candidate | Verify cohort continuity; revise the retention mechanism |
| New branches raise total group sales | Non-discriminating | Does not answer the mature-branch question |
The example’s six-month window is assumed, not an empirically established cutoff. A candidate becomes a confirmed failure only after checking that the cohort and measurement have not changed. Missing data may prevent a conclusion, but should not automatically grant the thesis another observation window.
Verify and handle exceptions
Write down allowable exceptions in advance. A planned closure can legitimately change a cohort; an undocumented exclusion added after weak results is a moving goalpost. Preserve both the originally reported and the adjusted series, including why an exclusion is justified.
Ask whether management could satisfy the metric while violating the mechanism. Revenue retention might improve through price increases even as customer counts fall. If loyalty is the necessary condition, track the customer population as well. If public disclosures cannot reveal it, the thesis is less falsifiable than your spreadsheet suggests.
This approach does not produce automatic trading instructions. A broken mechanism can coexist with an attractive valuation or a different defensible thesis, but that new thesis deserves a separate record. It is also less useful for broad statements such as “the brand will remain relevant” until those statements are narrowed into observable implications.
A further check on the example
Test the proposed rule with an imagined false alarm before adopting it. Suppose a branch reports weaker repeat sales because a large customer changes its ordering month, then catches up the following month. Would your rule distinguish that timing shift from permanent customer loss? If not, refine the observation rather than adding an exception after results arrive. Also identify who will verify the cohort and where the original threshold is stored, so the test can survive a change in researcher.
FAQ
Must every condition have a number?
No. Loss of a required license or termination of a named agreement can be an observable event. Specify the documentary evidence and its scope.
Can I change an invalidation condition?
Yes, when new information changes the question. Date the revision, preserve the old condition and explain whether it had already been breached.