Skip to main content
  • Label: Optimizations
  • Description: Actions, made by your team or by the Layers learning system, that change how search behaves.
  • Default group key: action_type
Use the optimizations dataset to see how much tuning your search has received and who did it. Each row is one optimization: a semantic redirect, an intent modifier, a ranking rule, or a relevance threshold that Layers learned for a group of related queries. You can split the count between changes your team made and changes Layers applied on its own. The Searches Optimized card on the default dashboard is built on this dataset.
Layers rebuilds this dataset once a day from the optimizations that currently exist in your store. Keep the following in mind when you read the numbers:
  • Each optimization counts once, on the date it was created. A learned relevance threshold counts on the date Layers last updated it.
  • When you delete an optimization, it drops out of every time range at the next daily refresh, including past periods.
  • Editing an optimization does not create a new row or move it to a new date.
  • Changes you make today appear after the next refresh.

​
Metrics

  • optimizations (events)
    • Total optimization actions applied.
    • Example: COUNT_DISTINCT(optimizations)
  • autonomous_optimizations (events)
    • Optimizations the Layers learning system created with no person in the loop. See is_autonomous for what counts as autonomous.
    • Example: COUNT_DISTINCT(autonomous_optimizations)

​
Dimensions

  • action_type: The kind of optimization. One of:
    • redirect — A semantic redirect that sends matching queries to a page instead of a result list.
    • modifier — An intent modifier that promotes or demotes products when a query signals a preference, such as a color or material.
    • ranking_rule — A ranking rule that promotes, demotes, pins, filters, or sorts products for matching queries.
    • learned_floor — A minimum relevance score that Layers learned for a group of related queries. Products that score below it are left out of results for those queries.
  • is_autonomous: true when the Layers learning system created the optimization, false when a person did. Layers marks these as autonomous:
    • Redirects that Layers found by crawling your store’s pages.
    • Intent modifiers that Layers mined from shopper behavior or recovered from earlier learned data.
    • Visual ranking rules.
    • Every learned relevance threshold.
    Redirects you add by hand or accept from a suggestion, intent modifiers you create, and manual ranking rules count as false.
This dataset does not carry shopper or session context. The personalization dimensions, geography, marketing, and custom.* dimensions that other datasets share do not apply here.

​
Examples

Total optimizations compared to the previous period (last 30 days)
Optimizations by type (last 90 days)
Autonomous versus manual optimizations by type (last 90 days)
Weekly trend of autonomous optimizations (this quarter)

​
Next steps