Scenario selection and baseline
The gain depends on the scenario chosen. A support team's response preparation time, a new joiner's ramp-up and a field team's procedure lookup are measured differently, and each gets its own baseline.
- How many people, how many times, is answered per scenario
- Today's average time to find is measured
- Questions that go unanswered today are counted too
- A gain target is written per scenario
Content readiness and ownership
Outdated, contradictory and ownerless documents return poor results even from a good retrieval system. Content past its shelf life is excluded when the source set is chosen, and every document set gets an owner.
- Contradicting documents are identified and sent to their owner
- Content past its shelf life is left out of scope
- The document owner takes on responsibility for updates
- Scope starts narrow and widens on evidence
Trust and verification
A user who cannot see where an answer came from tries the system once and abandons it. Every answer shows its source, and where no source is found the system says so rather than guessing.
- Answers arrive with the document and section they rest on
- No source means no answer
- Users can flag an answer as wrong
- Flagged answers are routed to the document owner
Adoption and gain
The gain appears when the system is used, not when it is built. Usage, answer accuracy and time to find are tracked together; where adoption is low, the problem is usually content quality.
- Usage is measured per scenario and per team
- Time to find is compared against the baseline
- Unanswered questions reveal content gaps
- Content gaps are routed to document owners