Looker Studio shows different numbers from GA4: which one is right.
- “looker studio not matching ga4”
- “looker studio numbers different from google analytics”
- “data studio vs ga4 discrepancy”
- “looker studio users higher than ga4”
- “looker studio sessions don't match ga4”
- “looker studio revenue different from ga4”
- “which is right ga4 or looker studio”
Baris Asa, tracking consultant, London · Last checked: 10 November 2026
When Looker Studio and GA4 disagree, the GA4 standard report is the reference, as long as both ask the same question: same metric, same dimension, same filter, same dates. Google’s own connector page says Looker Studio matches GA4 reporting, not explorations. So a gap usually means the tile asks something slightly different, reads cached data, or adds up users that GA4 counts once. Sampling, thresholds and the (other) row explain the rest, and they come last.
How big a gap should worry you? With matched fields and a closed date range, a small gap on users or sessions is expected: Google says both are estimates and gives no number. In my experience it stays within two or three per cent. A users total far above GA4 means rows were added up. Totals that match while the channel rows don’t usually mean the tile reads a different source scope. Anything else, work through the seven checks at the end.
One warning first. “Which one is right” only means “which one is right about GA4”. If GA4 disagrees with your store or your ad platform, a tile that matches GA4 perfectly is still wrong about your revenue. That’s a different file: GA4 revenue higher than Shopify, or GA4 and Google Ads numbers don’t match.
(Google renamed Looker Studio back to Data Studio in April 2026. Same product, same GA4 connector. Most dashboards I open still say Looker Studio, so I do too.)
Are you comparing with a report or an exploration?
Check this first. Google’s Data Studio help says the connector “will match the data, but not the explorations, in Google Analytics reporting.”
Google’s page on reports and explorations explains why the two differ. Some fields work in one and not the other. Exploration filters are case-sensitive. Explorations stop at your data retention setting. And the two are processed at different times.
So if the GA4 number you trust came from Explore, you’re comparing with the wrong screen. Rebuild the question in a standard report, with the same dimension and date range, and compare with that.
Is the tile asking the same question?
Most gaps I see start here. It’s what R-501 is about: one label, two definitions.
To see what a tile asks, open the report in edit mode and click the chart. The Setup panel names the exact dimension and metric, for example “Total users” or “Session source / medium”. Then look at the column header in the GA4 report you’re comparing with. The names have to match word for word.
| TILE FIELD | OFTEN CONFUSED WITH | WHAT DIFFERS |
|---|---|---|
| Total users | Active users | Counts people who didn’t engage |
| Source / medium | Session source / medium | Credit for the conversion, not the visit |
| Purchase revenue | Total revenue | Subscriptions and ad revenue |
| Page path + query string | Page path | Query kept, more rows |
- Users. Google’s user metrics help says Total users counts everyone who triggered any event, Active users counts those who engaged, and in reporting “you’ll only see active users”. A tile on Total users can run higher. Switch it to Active users.
- Source and channel. GA4 holds three versions. Session source is where the visit came from. First user source is where the person first came from. Plain source is, in Google’s words, “the source attributed to the key event”: the channel GA4 credits with the conversion. Traffic acquisition uses the session version. A tile on plain Source / medium spreads conversions across channels differently. Switch it to Session source / medium.
- Revenue. Purchase revenue is purchases minus refunds. Total revenue adds subscriptions and advertising revenue. Use whichever the GA4 report’s column header names.
- Page. Page path stops before the
?; the query string version keeps it. Used as a dimension, the long version splits one page into a row per URL variant. Again, match the report you compare with.
Then the filters. Google’s connector page says GA4 segments and comparisons aren’t available through the connector. A GA4 report with a comparison applied can’t be copied one for one. Rebuild it as a chart filter, with the same match type.
Is the tile showing old data?
Looker Studio caches. Google’s data freshness page lists 12 hours as the default for the Google Analytics connector. Its table also lists 1 and 4 hours, though the same page says Google Analytics refreshes every 12 hours. A tile can show what GA4 said this morning while GA4 has already revised the figure. Refreshing the data resets the cache.
GA4 also revises recent days. Google’s Data API page says data “might have changed due to ongoing processing” between two requests. Key events can shift for days as attribution settles: see why GA4 numbers change after a few days.
The rule I use: refresh the data, then compare a closed range that ends at least three days ago. If the gap disappears, it was timing.
Is something adding up users that GA4 counts once?
Users and sessions are counts of distinct people and visits. GA4 counts a person once however many rows they appear in. Add the rows up and they’re counted again.
- A user in two rows. Someone who came from Google Ads on Monday and email on Thursday is one user in GA4’s total, and one user in each channel row. A blend, a SUM in a calculated field, or adding the rows up yourself in a spreadsheet counts them twice. The usual result is “Looker shows more users than GA4”, and it’s a build problem, not a tracking one.
- Session fragments. Google’s connector page warns that Stacked Bar charts can show “double-counted” sessions when complex filters or segments meet certain date dimensions, “most commonly ISO Year ISO Week”. Its fixes: use Date or Month instead, simplify the chart’s filters or move them into a blended data source, or build the query as a view in BigQuery.
- Totals that go blank. Google says that when a table is filtered by a calculated field, a custom group or a bin, the grand total shows
null. That’s not a fault.
Neither tool counts users exactly anyway. Google’s Data API page says Active users and Sessions are approximations, so a small difference on a large property isn’t worth a meeting.
| STEP | TILE | GA4 REPORT | GAP |
|---|---|---|---|
| As found: Total users, blend | 58,960 | 41,820 | +41% |
| Active users | 52,310 | 41,820 | +25% |
| Blend removed | 42,140 | 41,820 | +0.8% |
| Refreshed, closed range | 41,790 | 41,820 | -0.1% |
The big jump was the blend, not the metric. That’s the pattern: when a users total is far above GA4, look for rows being added up before you look at anything else.
Has the API sampled, thresholded or grouped the rows?
Only now look at R-502 : same definition, same dates, still a gap. The Data API, the feed Looker Studio reads GA4 through, may have answered with less than the full table.
- Sampling. Google’s Data API page says the API “can return sampled data, especially for large datasets or complex queries”, with thresholds and algorithms that can differ from the GA4 interface. Google doesn’t publish the limit for the standard queries the connector runs. MeasureSchool reports that Looker Studio shows no sampling notice. Test it: compare an event count or revenue tile over one week, then over a year. If the week matches and the year doesn’t, sampling is the likely cause.
- Thresholds. The API can withhold rows with few users, as GA4 reports do. A tile broken down by age, gender or a narrow segment loses rows, and its total drifts. Triggers and fixes are in GA4 data thresholding.
- (other). When a dimension has too many distinct values, GA4 folds the rare ones into an (other) row, in reports and API responses alike (filed as R-502). Google’s Data API page adds that filters are applied after aggregation and “don’t look inside the (other) row”. Filter a tile for one campaign and any of its rows already folded into (other) are gone.
- Reporting identity. Google’s Data API page says user counts change with the reporting identity setting. If someone switched it between your two checks, the numbers moved for that reason alone.
A tile that shows an error instead of a number is a different problem: Google’s connector page says GA4 reports are subject to Data API quotas and may show an error when they exceed them. That’s a limit, not a mismatch.
The check, in order
- Compare with a standard report, not Explore.
- Match the metric: Active users, the same revenue field, the same key event.
- Match the scope: session, first user or key event source.
- Match the filter and match type; drop comparisons.
- Refresh the data and use a closed range ending three days ago.
- Remove blends, SUM fields and spreadsheet totals that add up users or sessions.
- Test a short range against a long one, and drop low-volume dimensions, to rule out sampling, thresholds and (other).
If the tile still disagrees after step 7, email me a screenshot of both, with the field names and date range. In the reports I’ve rebuilt, it’s usually step 2, 3 or 6. The fix that stops it coming back is writing down what each tile means: one definition, one source and one owner per number. That’s the metric register in reporting people actually trust. When the report needs rebuilding to that register rather than patching tile by tile, that’s Looker Studio reporting: the data checked first, then every tile built to its definition, in an account your organisation owns.
Common questions
Why does Looker Studio show different numbers from GA4?
Usually because the tile isn’t asking the same question as the GA4 screen you compare it with: a different metric (Total users against Active users), a different scope of source or channel, a different filter, a date range that includes recent days, or 12-hour cached data. Less often, the Data API the connector reads from has sampled, thresholded or folded rows into (other). Match the question first.
Which one is right, Looker Studio or GA4?
For the same metric, dimension, filter and dates, treat the GA4 standard report as the reference. Google’s connector page says Data Studio matches GA4 reporting, not explorations. If the tile still differs after you match the question and refresh the data, the tile is built differently. Neither number is right about your sales if GA4 disagrees with the store or the ad platform: that is a separate check.
How big a gap between Looker Studio and GA4 is normal?
With the same fields and a closed date range, a small gap on users or sessions is estimation or timing: Google says both are approximations, without giving a number. In my experience it stays within two or three per cent. A users total well above GA4 usually means rows were added up. Totals that match while channel rows differ usually means the tile uses a different source scope.
Why does Looker Studio show more users than GA4?
Two common reasons. The tile may use Total users, which also counts people who didn’t engage, while GA4 reports show active users. Or users were added up across rows: in a blend, in a SUM calculated field, or by hand in a spreadsheet. Someone who arrived through two channels is then counted twice. GA4 counts that person once.
Does Looker Studio match GA4 Explorations?
No. Google’s Data Studio help says the connector matches GA4 reporting, not explorations. Compare a Looker Studio tile with a standard report, never with a free-form exploration.
How often does Looker Studio refresh GA4 data?
Google’s data freshness page lists every 12 hours as the default for the Google Analytics connector. A tile can show numbers GA4 has since revised. Refresh the data, then compare a closed range that ends at least three days ago.
Is Looker Studio sampling my GA4 data?
It can be. The connector reads the GA4 Data API, and Google’s Data API page says the API can return sampled data for large datasets or complex queries, with thresholds that can differ from the GA4 interface. Google doesn’t publish the limit for the standard queries the connector runs. MeasureSchool reports that Looker Studio doesn’t flag sampling the way GA4 does. Test it: if a one-week range matches GA4 and a one-year range doesn’t, sampling is the likely cause.
Sources
Read on 10 November 2026:
- Data Studio Help, Connect to Google Analytics (matches GA4 reporting, not explorations; GA4 segments and comparisons not available; reports subject to Data API quotas and may show an error; Stacked Bar session fragments, most commonly with ISO Year ISO Week, and the three fixes; grand total shows
nullwith calculated-field, custom group or bin filters) - Data Studio Help, Manage data freshness (Google Analytics connector: every 1, 4 or 12 hours, 12 the default; refreshing resets freshness)
- Google for Developers, GA4 Data API: reporting data expectations (sampling with thresholds that can differ from the UI, thresholding, the (other) row, filters applied after aggregation, Active users and Sessions as HyperLogLog++ estimates, reporting identity, ongoing processing)
- Google for Developers, GA4 Data API schema (source, sessionSource and firstUserSource; purchaseRevenue and totalRevenue; pagePath and pagePathPlusQueryString)
- Google Analytics Help, Data differences between reports and explorations
- Google Analytics Help, Understand user metrics (Total users and Active users; reporting shows active users)
- Google Cloud blog, Looker Studio is Data Studio (11 April 2026)
- MeasureSchool, Data discrepancies between GA4 and Looker Studio (practitioner report: no sampling notice in Looker Studio)