Aglet

Investigate query learning signals with session comparisons

An investigation should distinguish a search failure from a successful refinement or an unrelated exit. Compare session paths with similar queries, keep result state attached, and test one hypothesis about vocabulary, facets, ranking, or scope without treating behavior as self-explanatory.

Build a useful investigation brief

  1. Build the session matrix

    List query, result count, selected filters, suggestion, click, backtrack, reformulation, exit, and session boundary for each sample. Include success and failure paths. Record missing events and preserve query family context.

  2. Compare competing explanations

    Group paths by result state, not only final action. Test whether reformulation follows zero results, irrelevant ranking, or a missing facet by replaying representative queries. Record a successful filter path and a false-positive click as controls.

  3. Write the learning brief

    State observed behavior, evidence limits, likely explanations, and one experiment that would separate them. Recommend a bounded review of rules, copy, facets, or vocabulary. Keep anonymous or sparse session data from supporting claims it cannot establish.

What to carry forward

The investigation is complete when session paths and result states support a bounded explanation or an explicit evidence gap. Stop with one experiment and a pass condition. Do not label abandonment as dissatisfaction without a comparison that links the exit to search behavior.

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