Detect stale evidence before reusing a research thesis
Evidence becomes stale when the conditions making it relevant have changed, not merely when its publication date is old. Review the claim, population and mechanism before reusing a research note. A recent article can repeat an outdated observation, while an older contract may still govern current operations.
Give each observation a validity condition
Different evidence ages differently. A debt agreement remains relevant until amended, repaid or refinanced. A customer mix snapshot may become less representative after an acquisition. A production test can remain technically valid while saying little about a redesigned product.
The 10-K describes a specific fiscal period and its business context. Check both the date it became available and the period it covers. A newly published summary does not refresh the underlying measurement automatically.
Hypothetical stale-evidence register
Assume you reuse a thesis that a supplier has diversified demand. The supporting note is based on last year’s customer mix. The company has since acquired a business concentrated in one large customer.
| Evidence | Original use | Change that threatens validity | Refresh action |
|---|---|---|---|
| Prior customer mix | Supports diversified demand | Acquisition changes group composition | Recalculate latest comparable concentration |
| Existing supply agreement | Supports input availability | Expiry or amendment | Read current terms and updates |
| Old product reliability test | Supports low service cost | Product redesign | Seek relevant new-generation evidence |
| Reposted analyst summary | Appears recent | Underlying data unchanged | Trace to original observation date |
In this invented case, the old customer table remains a valid historical observation. It is the inference about the current group that has expired. Preserve the original and mark its current relevance as limited rather than deleting it.
Prioritize refresh work
Rank evidence by the importance of the supported assumption, the likelihood of changed conditions and the cost of rechecking. A necessary financing assumption deserves earlier attention than a minor descriptive detail. Create an event trigger for acquisitions, leadership changes, regulatory changes, product redesigns or reporting redefinitions relevant to the mechanism.
Use a refresh record: last verified date; underlying observation period; validity condition; change detected; replacement source; and conclusion impact. Avoid universal expiry rules such as “all sources older than ninety days are unusable.” Their appropriateness depends on the fact being checked.
Verify whether the latest document restates prior information or just supplies a new period. A new customer metric with a changed definition cannot be joined directly to the old series. You may need separate before-and-after analyses rather than a continuous trend.
This approach cannot guarantee that all material changes are public. It also does not mean older sources are inferior. Historical evidence is essential for base rates and process reviews; label the use so a reader understands whether the source supports a past fact or a current claim.
A further check on the example
For the acquisition example, record both the legal close date and the first reporting period containing the new business. A historical concentration table can remain the latest published table even after the economic group has changed. Until a new comparable breakdown appears, state that the diversification conclusion is pending refresh. Do not extend validity merely because the old table is still linked from the investor-relations page; website availability and current relevance are different tests.
FAQ
Should I delete an outdated note?
Usually preserve it with a status and replacement link. It can explain the prior reasoning and prevent the same outdated inference from being rediscovered later.
Is publication recency enough to rank sources?
No. Trace the date of the underlying observation and check whether its population, contract and mechanism still apply to the current question.