JSIFS Governance Group · 2026-03-04
An evidence label answers one question: what does this statement rest on? It does not answer whether the statement is important, whether the programme is going well, or whether a team is performing.
The most common failure is aggregation. Counting labels across a register and reporting a percentage converts a descriptive system into a score, and scores attract optimisation. Teams begin to avoid recording Hypothesis-level statements, which removes exactly the material the register exists to hold.
The second failure is silent upgrade. A claim moves from Data-Supported to Experimentally Supported only through a documented protocol with controls and artifact screening, reviewed by someone who did not run the experiment. If a label changes, the rationale and the reviewer change with it.
The third is deletion. Contradicted and Quarantined are register states, not removals. A programme that erases contradicted claims loses the ability to explain to a future reviewer, or to itself, why a direction was abandoned.