Analyze query performance statistics

This page describes how to use the Query insights dashboard to detect and analyze query performance.

Query insights overview

Use the Query insights dashboard to monitor metrics-related queries. Based on the metrics, you can identify the most frequently used queries and queries with high latencies that might require optimization. Use the Query insights dashboard to help you with:

  • Query performance optimization: identify queries with high latencies and that might need optimization.
  • Query cost management: discover high-cost queries and optimize them to reduce costs.
  • Query stats monitoring: track query stats over time.
  • Index recommendations: identify missing or suboptimal composite indexes and create recommended indexes directly from the dashboard (Enterprise edition).

Query insights data

Query insights includes data from the following API methods:

You can view data about the queries that use these methods for a given project, database, and time duration ranging from 10 minutes to 30 days. Data for queries with equivalent structures is captured under a single normalized query.

Query insights returns the following information about a query:

Normalized query text The query structure represented in text.
Execution count Number of executions in the selected time window.
Error count Number of errors in the selected time window.
Average execution duration(ms) The average time in milliseconds it took the database to process the query.
Average number of results returned The number of results returned by the query. Results include documents, collection IDs, and aggregated buckets.
Average number of documents scanned The number of documents scanned in a query.
Average number of index entries scanned The number of index entries examined to execute the query.
Load by average time Data to help filter for the top queries based on average latency.
Load by total (billable) read operations Data to help filter for the top queries based on total billable read operations.
Index recommendations Recommended composite indexes, used indexes, and estimated workload impact for queries in Enterprise edition databases.

Data granularity and retention

Data granularity depends on the duration specified:

  • 10 minute granularity for intervals up to 4 days ago
  • 1 hour granularity for intervals up to 30 days ago

The maximum data retention for Query insights is 30 days. 10-minute data is stored for 4 days, and hourly data is stored for 30 days.

View Query insights

To view query insights for a Firestore database, open the database Query insights pane in the Google Cloud console.

  1. In the Google Cloud console, go to the Databases page.

    Go to Databases

  2. From the list of databases, select a database.

  3. In the navigation menu, click Query insights.

    The database usage dashboard in the console.

    Use the Load type drop-down to find the top queries by either latency or number of read operations.

Index recommendations

Index recommendations for Query insights automatically identify missing or suboptimal composite index configurations for frequently executed query shapes in Firestore Enterprise edition databases.

In Firestore Enterprise edition, queries can run even without matching composite indexes, which can result in slow full collection scans or in-memory sort operations. The Query insights dashboard analyzes historical query execution statistics and suggests targeted composite indexes to optimize query latency and reduce billable scanned documents.

How index recommendations work

An offline batch pipeline periodically evaluates historical top queries:

  • Query structure analysis: Extracts equality, sort, and range predicates from frequently executed queries.
  • Candidate index generation: Generates prospective composite index definitions using index-ordering heuristics (Equality > Sort > Range).
  • Optimizer evaluation: Injects candidate indexes into the query engine's cost model to compare baseline query execution costs against hypothetical indexed execution plans.
  • Workload impact ranking: Ranks recommended indexes into impact tiers based on total estimated workload cost savings:
    • HIGH: Converts full collection scans to selective index seeks, significantly reducing document reads and execution latency.
    • MEDIUM: Eliminates expensive operations such as in-memory sorting by providing pre-sorted index results.
    • LOW: Offers minor seek improvements.
To view index recommendations and create suggested indexes from the Google Cloud console:

  1. In the Google Cloud console, navigate to the Databases page.

    Go to Databases

  2. From the list of databases, select an Enterprise edition database.

  3. In the navigation menu, click Query insights.

  4. In the queries table, select a query shape to view its Recommendations card. The card displays:

    • The recommended composite index definition (collection, fields, sort orders, or array-contains properties).
    • The estimated workload impact tier (HIGH, MEDIUM, or LOW).
    • The indexes used, or an indication that a full collection scan is performed.

  5. Click Create index on the recommendation card. A dialog opens with the collection, query scope, and fields pre-populated.

  6. Click Create to begin building the index. You can monitor the index build progress on the Indexes page in the Google Cloud console.

Supported editions and modes

Index recommendations are supported for Firestore Enterprise edition databases, including:

  • Firestore in Native mode
  • Firestore with MongoDB compatibility

Standard edition databases aren't supported for index recommendations.

Limitations

  • Real-time listeners are not included in Query insights statistics.
  • Query insights data is delayed by one to two hours.
  • Index recommendations are generated periodically based on historical query data, not in real time.
  • Index recommendations suggest new composite indexes; they don't suggest modifying or deleting existing indexes.
  • Automatic index creation isn't supported; recommended indexes must be created manually.
  • Query normalization replaces literal values with placeholders. Index recommendations are based on normalized query templates rather than specific runtime values.
  • Cost evaluation uses heuristics rather than live data sampling. Recommended indexes might not improve performance in every scenario, and creating an index adds write and storage overhead.

Pricing

There's no additional cost for Query insights or index recommendations.

Required roles

To get the permissions that you need to view the Query insights dashboard and create recommended indexes, ask your administrator to grant you the following IAM roles on your project:

For more information about granting roles, see Manage access to projects, folders, and organizations.

These predefined roles contain the permissions required to view the Query insights dashboard and create recommended indexes. To see the exact permissions that are required, expand the Required permissions section:

Required permissions

The following permissions are required to view the Query insights dashboard and create recommended indexes:

  • View Query insights: datastore.insights.get
  • Create recommended indexes:
    • datastore.indexes.create
    • datastore.indexes.update

You might also be able to get these permissions with custom roles or other predefined roles.

What's next