Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.uselayers.com/llms.txt

Use this file to discover all available pages before exploring further.

Search quality is in Beta. Layers is still tuning how patterns are detected and how impact is reported. Optimizations are not deployed automatically — your team approves each change before it goes live.
Use Search quality to see what Layers is optimizing for your store. Each row is a query pattern Layers detected in recent shopper traffic and the optimization being tested for it. Examples include adding a synonym, adjusting an attribute weight, or rewriting an intent. Use this page when you want a plain-language view of search improvements in flight, without digging into raw query or ranking data.

​
Prerequisites

  • Your store must be connected to Layers.
  • Your catalog should be synced.
  • A few days of shopper search activity should exist so Layers can detect query patterns worth optimizing.

​
Open Search quality

  1. In the Layers dashboard, open Lab.
  2. Select Search quality.
  3. Review the table of optimizations.
If the table is empty, Layers has not detected enough recent search activity to suggest optimizations yet. Check back after a few more days of shopper traffic.

​
What each column means

  • Time - When Layers detected the pattern and proposed the optimization.
  • Query pattern - The group of related shopper queries the optimization targets. For example, color-led queries or queries that look for a specific product type.
  • Optimization - The type of change being tested:
    • Synonym expansion - Layers learned a new related term to broaden matching for the pattern.
    • Attribute boost - Layers is weighting a product attribute more strongly for the pattern.
    • Negative filter - Layers is filtering out a class of results that looked off-topic.
    • Query modifier - Layers is adjusting how the query is interpreted before ranking.
  • Source - The signal that drove the optimization, shown as the attribute or signal name followed by its type, for example color (Categorical Attribute) or product_type (Categorical Attribute). This tells you which part of your catalog the optimization is anchored to.
  • Expected impact - The expected change in low-quality results for shopper-visible positions. “8.4% fewer bad results” means Layers expects fewer weak matches in the top results after the change. “Pending” means the optimization has not been tested yet. “No change” means testing did not show a meaningful improvement.
  • Status - Where the optimization is in its lifecycle.

​
Optimization status

Each optimization moves through these states:
  1. Reviewing - Layers detected a pattern and proposed an optimization. It has not been tested yet.
  2. Tested - Layers replayed the affected queries with the proposed change and measured the impact.
  3. Queued - The optimization has been approved and is waiting to go live.
  4. Live - The optimization is active and affecting shopper search results.
Optimizations stay visible after they go live so you can see what Layers has changed for your store recently.

​
How Layers chooses what to optimize

Layers groups similar shopper searches into patterns, then reviews the top results for each pattern to decide which queries returned weak matches. When a pattern shows a consistent failure mode — for example, the wrong product type appearing for color-led searches — Layers proposes an optimization that targets that failure. Each proposal is tested against the same queries before any change goes live. You see the result of that process here; the underlying analysis stays inside Layers.

​
Troubleshooting

​
The table is empty

Layers needs a few days of shopper search activity to detect patterns worth optimizing. If your store is new or recently re-launched, check back after more traffic has accumulated.

​
An optimization stayed in Reviewing for a while

Optimizations are tested in batches. A new row may show Reviewing for a short period before moving to Tested.

​
Expected impact says “No change”

Layers tested the optimization and did not see a meaningful improvement on the affected queries. The optimization will not be queued for deployment.

​
Expected impact says “Pending”

The optimization has been proposed but not yet tested. The value updates once testing completes.

​
Next steps

  • Lab overview - See all Lab tools and recent engine activity.
  • Audit Log - Inspect individual AI search decisions Layers made.
  • Experiments - Review zero-result and weak-coverage searches.
  • Test text search - Replay a query to see current results.