You open a report, and instead of data there is a small orange icon and the phrase “Data thresholding applied.” Some rows are missing. Totals do not add up to the sum of what you can see.
Nothing is broken and no data was lost. GA4 has decided that showing you those rows could identify individual people, so it is withholding them at query time.
The confusion is that this looks exactly like sampling, and the fixes for sampling do nothing here.
Thresholding is not sampling
Both leave you with incomplete reports. They are unrelated mechanisms.
| Sampling | Thresholding | |
|---|---|---|
| Why it happens | query too large to compute exactly | too few users behind a row |
| What it does | estimates from a subset | withholds rows entirely |
| Effect on numbers | approximate | some rows simply absent |
| Triggered by | very large date ranges, complex queries | Google Signals + small segments |
| Fixed by | narrowing the query, GA4 360 | widening the range, disabling Signals |
Note the direction reverses. Sampling gets better when you narrow the query. Thresholding gets better when you widen it, because more users end up behind each row. Applying the sampling instinct here makes the problem worse.
What sets it off
The dominant cause is Google Signals being enabled while you view a report containing demographics, interests, age, gender, or cross-device data.
Google Signals stitches sessions across devices using data from signed-in Google users. That is personally identifiable territory, so Google applies a privacy floor: if fewer than roughly 50 users sit behind a row, the row is withheld.
This is why thresholding hits small sites, short date ranges, narrow segments and granular breakdowns hardest — precisely the reports where every row matters most.
Four ways out, in order of what they cost you
1. Widen the date range. Free, instant, no trade-off. Seven days becomes twenty-eight; more users sit behind each row; rows reappear. Try this before anything else.
2. Switch reporting identity to Device-based. Admin → Data display → Reporting identity → Device-based. This removes the Google Signals layer from report calculations and dramatically reduces thresholding. You lose cross-device stitching, so a person on phone and laptop counts as two users. For many sites that is an acceptable trade for reports that actually render.
3. Turn off Google Signals. Admin → Data collection and modification → Google Signals. Thresholding largely disappears. You permanently lose demographics, interests and cross-device reporting — and turning it back on does not backfill the gap. Consider this carefully rather than as a quick fix.
4. Use the BigQuery export. The raw event export has no thresholding, no sampling and no cardinality limits. It is the complete dataset and the only genuinely complete answer. The cost is SQL and a BigQuery project, and it is not retroactive: the export only contains data from the day you enabled it.
The decision underneath all four options is the same one: how much do you value demographic and cross-device reporting versus complete row-level reporting? GA4 will not give you both at small volumes, and no tool or workaround changes that.
What will not work
- Refreshing, or waiting. Thresholding is applied at query time, every time.
- A different report showing the same dimension. Same threshold, same result.
- Any third-party dashboard. Including ours — see below.
- Custom dimensions. They do not exempt anything from the privacy floor.
The honest limit for tools like ours
This is a case where it matters to say plainly what a third-party tool cannot do.
Thresholding is applied by Google, at query time, inside the API. GA4 Simple View reads the GA4 Data API, which means it receives exactly the same withheld results the GA4 interface does. We cannot show you rows Google has decided not to return, and any tool claiming to bypass thresholding is either using BigQuery underneath or is not telling you the truth.
What we do help with slightly: our default ranges are wider than GA4’s seven-day default, which incidentally reduces how often you hit it. That is a side effect, not a solution.
If you need genuinely complete data at low volumes, the honest answer is the BigQuery export — and it costs nothing to enable today so that the data exists when you eventually need it.
Questions people also ask
Each answer stands on its own, so it still makes sense quoted somewhere else.
What does data thresholding applied mean in GA4?
It means GA4 has withheld some rows from the report because too few users sit behind them and showing the data could identify individuals. The threshold is applied at report time, so the underlying data still exists and the same query over a longer period may return it.
Is thresholding the same as sampling in GA4?
No, and confusing the two sends you down the wrong path. Sampling estimates results from a subset of data when a query is too large. Thresholding withholds specific rows for privacy when the user count behind them is too small. Sampling makes numbers approximate; thresholding makes rows disappear entirely.
What triggers thresholding in GA4?
Most often Google Signals being enabled combined with a report that includes demographics, interests, age, gender or cross-device data. Any report where the user count behind a row falls below the privacy threshold, commonly cited as around 50 users, can trigger it.
How do I remove thresholding from GA4 reports?
Widen the date range so more users sit behind each row, switch reporting identity from Blended to Device-based, or disable Google Signals in Admin under Data collection. Each removes thresholding at a cost: you lose cross-device stitching and demographic reporting when Signals is off.
Does BigQuery export have thresholding?
No. The GA4 BigQuery export contains raw, unaggregated event data with no thresholding, no sampling and no cardinality limits. It is the only way to see the complete dataset, at the cost of writing SQL and running a BigQuery project.
Does thresholding affect the GA4 Data API and third-party tools?
Yes. Thresholding is applied by Google at query time, so any tool reading the GA4 Data API is subject to the same withholding as the GA4 interface. No third-party dashboard can show you rows Google has decided to withhold.
