Aglet

Triage product search intent drift before changing behavior

A product query can begin as a shopping request and finish as a request for instructions, account help, or navigation. Triage the job behind the words before tuning ranking. Compare the submitted phrase with the result type chosen, the next query, and the context that led someone to search.

Establish what is happening

  1. Read the query in context

    Start with the original phrase, entry point, and selected scope. Pair it with the result type shown, the first useful click, and any immediate reformulation. Keep catalog, support, and account searches separate so a shared noun does not create a false intent match.

  2. Compare the journey after search

    Review whether the person refined, clicked, returned, or abandoned. A product noun followed by a setup article may indicate an informational job, while a category click may confirm commercial intent. Record these as observations and keep interpretation separate.

  3. Set an intent boundary

    Label the smallest supported intent set and list representative queries for each label. Note which signals are missing, especially when no click follows. Leave ambiguous phrases in a review queue rather than silently broadening product results.

What to carry forward

Return an intent brief with query examples, result types, follow-up behavior, and confidence. Stop when the team can distinguish shopping, support, and navigation cases from evidence. If the boundary remains unclear, preserve the query and request one targeted observation instead of changing ranking.

Technical background: Digital.gov guidance.

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