persist_for (for models)

Stay organized with collections Save and categorize content based on your preferences.

This page refers to the persist_for parameter that is part of a model.

persist_for can also be used as part of an Explore, described on the persist_for (for Explores) parameter documentation page.

persist_for can also be used as part of a derived table, described on the persist_for (for derived tables) parameter documentation page.


persist_for: "5 hours"
Default Value
1 hour

A string containing an integer followed by a timeframe (seconds, minutes, or hours)


Consider instead using a datagroup and persist_with, as described on the Caching queries documentation page.

persist_for lets you modify the amount of time that cached query results are used for a given Explore. The default cache length in Looker is 1 hour. Cache results are stored in an encrypted file on your Looker instance.

The caching mechanism in Looker works as follows: once a user runs a specific query, the result of that query is cached. If someone runs precisely the same query again (everything must be the same, including minor things such as the row limit) in less than the interval specified by persist_for, then the cached results are returned. Otherwise, a new query is run against your database.

When the persist_for interval expires, data is deleted from the cache as long as the Instant Dashboards Labs feature is disabled. See the Caching queries documentation page for information on how long data is stored in the cache.

Explores also support persist_for. When an Explore and its model both have persist_for settings, the value set for the Explore will take priority for queries based on that Explore.

From an Explore, you can see whether a query was returned from cache or you can force new results to be generated from the database. See the Caching queries documentation page for more information.


Adjust the cache length to 2 hours:

persist_for: "2 hours"

Adjust the cache length to 30 minutes:

persist_for: "30 minutes"

Turn off caching so that users never get cached results for a query:

persist_for: "0 seconds"

Things to consider

Data is always written to the cache

When persist_for is set to 0 seconds, your users' queries will not retrieve data from the cache. However, Looker requires the disk cache for internal processes, so your encrypted data will always be written to the cache, even when persist_for is set to 0 seconds. Once written to the cache, the data will be flagged for deletion but may live up to 10 minutes on disk. See Caching queries documentation page for details.

persist_for does not necessarily align with your data import

Many companies have a daily data import to their analytics database. Sometimes, they reason that there is no purpose running fresh queries if the data isn't constantly updated anyway, so they set the cache length to 24 hours (like persist_for: 24 hours). However, this will not prevent users from getting data that is older than the most recent refresh.

For example, suppose a query is run at noon on January 1st, new data is imported the morning of January 2nd, and then the query is run again at noon on January 2nd. Since the query was run within the 24 hour window specified by persist_for, the data from January 1st will be returned, even though new data was loaded on January 2nd.

If you want your caching to be aligned with data imports, use datagroups and persist_with, as described on the Caching queries documentation page.

Scheduled Looks will cache results

When a scheduled Look is run, it will create a cached result set in the same way that a user run query would. If you want to precache a certain report, you may want to consider saving and scheduling it.