Dashboard Caching

A dashboard caches the result of every chart it draws. The first person to open a dashboard runs its queries against the data warehouse; everyone who opens it after that gets the stored result instead, which is why a second view is much faster than the first.

A cached result is discarded — and the next view re-queries the warehouse — in two situations: the table behind the chart is republished, or the result reaches the end of its cache lifetime.

Republishing a Table Clears Its Cached Results

When a workflow run republishes a table, every chart reading that table drops its cached results immediately. The next person to open the dashboard sees the refreshed numbers, without waiting and without forcing a refresh by hand.

This means a dashboard is never showing pre-refresh numbers because of caching. If a dashboard looks stale after a workflow run, the cause is the workflow — check that the run finished and that it republished the table you expect — not the cache.

Publish Under a Stable Name

Cached results are tracked per published table. If you rebuild a table under a different published name, the charts still pointing at the old name keep serving the old table’s data. Republish under the same name so the charts you already built stay attached to it.

Cache Lifetime

Every cached result also carries a lifetime, after which it expires on its own. The default is 1 hour.

The lifetime is not what keeps a dashboard current — republishing does that. So on data that changes only when a workflow runs, a short lifetime buys nothing and costs a great deal: any view after the last result expires re-queries the warehouse from scratch. Dashboards used in bursts — a morning review, a month-end close — pay that cost on almost every load.

The default was 5 minutes until August 2026. It was short because republishing did not reliably clear cached results, so expiry was the only thing keeping a dashboard current. Now that republishing clears them, an hour is both safe and far cheaper.

Raise it further on tables that change less often — a day suits one that reloads overnight — and the workflow that reloads the table still clears the cache the moment new data lands.

Caution

The lifetime is only safe to leave long for data that changes through PlaidCloud. If a chart reads a table that something outside PlaidCloud writes to, nothing signals the dashboard to drop its cache, and the lifetime is the only thing that refreshes it. Set a short lifetime on those datasets — the 1-hour default means they can now be up to an hour behind, where previously they were at most 5 minutes behind.

Changing the Cache Lifetime

Set the lifetime in seconds, either on a database or on an individual dataset. A value on the dataset wins over the value on its database, and the 1-hour default applies where you have set neither. Changing either needs the corresponding edit permission in the Dashboards area.

For every dataset on a database:

1. Go to Data → Databases in the Dashboards area
2. Click the edit icon on the database
3. Open the Advanced tab, then Performance
4. Enter a value in seconds in Chart cache timeout — for example, 3600 for one hour
5. Click Finish

For a single dataset:

1. Go to Data → Datasets
2. Click the edit icon on the dataset
3. Open the Settings tab
4. Enter a value in seconds in Cache timeout
5. Click Save

Two values have special meanings: 0 means never expire, so the result is only ever cleared by a republish, and -1 disables caching entirely for that database or dataset.

The change applies to results cached from that point on. Results already in the cache keep the lifetime they were stored with until they expire or the table is republished.

Refreshing a Dashboard by Hand

You rarely need to, but you can discard a dashboard’s cached results yourself: open the menu beside Edit dashboard and choose Force refresh dashboard. If a chart is serving an older result, a note in the upper right says how long ago the data was cached, and clicking it forces the same refresh.

Prefer either of these to reloading the browser page — a browser reload re-requests the charts but still receives the cached results.

Related

• Using Dashboards
• Learning About Dashboards