Aglet

Triage query learning signals before changing search rules

Search improvement needs evidence about what people tried next, not only the first query. Triage the learning signal before changing rules. Define the session boundary, capture query and result state, and separate observable behavior from an interpretation about why someone left.

Establish what is happening

  1. Capture the session path

    Record original query, result count, filters, refinements, clicks, reformulations, exit, and time context for representative sessions. Include a successful path and a no-result path. Keep sampling and missing events visible.

  2. Classify the next action

    Separate reformulation, filter narrowing, result selection, backtracking, and abandonment. Compare what each action says about search behavior without claiming its motive. A click may indicate success, curiosity, or a wrong result that needs inspection.

  3. Set a learning boundary

    State query family, session definition, evidence quality, and the next question. Preserve an anecdotal query as an example, not a trend. Stop when another reviewer can tell what is observed and what remains an interpretation.

What to carry forward

The triage output is a bounded learning signal with session path, action classes, sample limits, and an open question. Stop when behavior and interpretation are distinct. If the session boundary is unreliable, route instrumentation or manual review before changing relevance.

Technical background: Elastic documentation.

Keep the decision with the work.

Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.

Create an account See the product workflow