Quality teams rarely ask whether evidence is still relevant after the system changes. In a QI model, freshness may need to become a first-class property.
The hypothesis
Evidence freshness should be treated as a measurable quality attribute because stale evidence can create false confidence.
Why freshness matters
A test result can be technically valid and operationally irrelevant.
An architecture document may describe a dependency that no longer exists. A performance baseline may predate a major infrastructure change. An AI evaluation may use a prompt or model version that is no longer in production.
The evidence exists, but its relationship to the current system has weakened.
Can freshness be scored?
A simple freshness model might consider:
- age of the evidence;
- number of relevant changes since creation;
- whether the environment in which the evidence was generated still matches production;
- whether the source remains authoritative;
- and whether the evidence was revalidated after a major change.
Freshness is contextual
A one-year-old security test may still be valid if the component has not changed. A one-day-old AI evaluation may already be stale if the retrieval index or model changed overnight.
Time alone is not enough. Freshness must be related to change.
Experiment design
Choose a handful of evidence types: regression results, performance results, architecture mappings and AI evaluations.
For each, track age, relevant change events and whether the evidence remained predictive of actual behavior.
Then compare decisions made with fresh evidence versus decisions relying on stale evidence.
The goal
The purpose is not to invalidate old evidence automatically. It is to make uncertainty visible.
A Quality Intelligence system should be able to say:
“This evidence is technically valid but confidence in its current relevance is low.”
Freshness is part of provenance because evidence without time and change context can become misleading certainty.