Analytics and Tracking

Data Extract

Also called cached snapshot, materialised source

A stored snapshot of a slow data source, queried in place of the live connection so a report loads quickly.

Quick facts: Data Extract

Category
Analytics and Tracking
Also called
cached snapshot, materialised source
Level
Intermediate
Affects
Report load time, connector quota, data freshness
Where to see it
Looker Studio (extract data source), Power BI, Tableau
In this article4
  1. How a data extract works
  2. Why data extracts matter
  3. Where data extracts go wrong
  4. How to act on it

How a data extract works

When a report is connected live, every chart sends a fresh request to the platform each time someone opens the page. That is accurate but slow, and it burns through whatever request allowance the connector has. An extract breaks the chain: the tool pulls the data once, stores it as its own table, and every chart then queries that stored copy instead of the platform.

The stored copy is deliberately narrow. You choose the fields, the date range and the level of detail, and everything outside that selection is left behind. That is the source of the speed — a small table with the columns a report actually uses answers far faster than a live query against an account’s full history. The copy is then rebuilt on a schedule you set, and between rebuilds the report shows the data as it stood at the last one.

Why data extracts matter

Speed is what people notice. A report that takes a long time to open does not get opened, and a client who waits for charts to appear stops trusting the whole thing. An extract turns a sluggish report into one that responds immediately, which is often the difference between a dashboard being used and being ignored.

The second reason is quota. Connectors limit how many requests an account can make, and a busy report with many charts can exhaust that allowance on its own, producing errors for everyone. Serving the same report from an extract collapses many live requests down to one scheduled pull.

On a slower or intermittent connection — a common reality for teams working outside major hosting regions — the difference is larger still, because a live report is repeating requests over the same link every time a control is changed.

Where data extracts go wrong

The main risk is staleness presented as truth. A report on an extract looks exactly like a live one, so a reader checking this morning’s spend can be looking at yesterday’s. If a report contains anything time-sensitive, say on the page when the data was last refreshed.

The second is the silent trim. Fields left out of the extract simply do not exist for the report, so a filter or a new chart that needs one returns nothing rather than warning you. The third is a failed scheduled refresh nobody notices — the report keeps loading happily on an increasingly old snapshot.

How to act on it

Extract only where the live connection is genuinely too slow or too close to its limits, and keep the live connection for anything that has to be current. Include the fields the report uses plus the ones you know are coming, set the refresh to match how the report is read rather than how often the data changes, and display the last refresh time somewhere a reader will see it. Check the refresh history occasionally — a stale dashboard that still loads quickly is the most convincing kind of wrong report, and it undermines every decision made from it.

Do and do not

Do

  • Show the last refresh time on the report itself
  • Include the fields a future chart will need, not just today’s
  • Check the refresh history for silent failures

Do not

  • Extract data that has to be current to be useful
  • Assume a fast-loading report is an up-to-date one
  • Leave out fields you plan to filter on later

Questions people ask about this

Is a data extract the same as a cache?

They are close cousins. A cache stores the answer to a query for a short while so the same question does not have to be asked twice. An extract is a deliberate, structured copy of chosen fields and rows, stored by the reporting tool and rebuilt on a schedule you set. The extract is something you design; a cache mostly happens on its own.

How often should an extract be refreshed?

Match the refresh to how the report is used, not to how quickly the underlying data changes. A monthly performance review does not need overnight rebuilds. A report someone checks each morning before spending decisions does. Whatever you choose, show the last refresh time on the report so nobody mistakes an old snapshot for live figures.

Will an extract fix a report that keeps hitting connector errors?

Usually, yes. Those errors normally come from too many live requests against a request allowance, and an extract replaces all of them with one scheduled pull. It will not help if the failure is caused by broken credentials, a revoked account permission or a field the connector no longer supplies — those need fixing at the source.

Related terms

Found this useful?

Share it, or ask an AI to summarise it

Back to the glossary

Knowing the term is the easy part

Applying it to your own site and budget is the work. Book a call and I will tell you what actually applies to you.