How a threshold works
When a report is sliced finely enough, a row can end up describing a single person. City, device, age band and landing page together might leave exactly one visitor in a group, and anyone who knows that visitor could then read their behaviour out of a public report. To prevent this, analytics platforms apply a threshold: if the number of users behind a row is too small, the row is withheld rather than shown.
The platform does not tell you the exact cut-off, and deliberately so — publishing it would let people work backwards. In GA4 the withheld rows appear as gaps, and the report carries a small warning icon explaining that some data has been removed. Thresholding is triggered by the report’s reporting identity: it applies when demographic or signals-based data is in play, and disappears when the report is restricted to plain device-based data.
Why thresholds matter
The immediate effect is that totals stop reconciling. A summary row can be larger than the sum of the rows beneath it, because some of those rows were removed. People spend hours hunting a tracking fault that does not exist; the data was collected correctly and then suppressed at the reporting stage.
It hits smaller businesses hardest, which is most businesses in Nepal. A site with modest traffic, split across several cities and a handful of pages, reaches thin rows very quickly, so the more precisely you segment the less you can see. The tool is not failing — it is protecting the individuals in a small sample from being identified.
Where thresholds go wrong
The first mistake is treating a gap as a zero. A suppressed row is unknown, not empty, and treating it as nothing understates the channel or city it belongs to.
The second is fighting it with ever narrower filters. Each extra condition shrinks the group and makes suppression more likely, so the harder you dig the less you see. The third is exporting suppressed reports into a dashboard, where the warning icon does not travel and the missing rows silently become part of somebody’s chart.
How to act on it
When rows go missing, widen before you narrow. Extend the date range, combine cities into regions, group pages into sections, and drop the demographic breakdowns you do not actually need. Most suppression clears the moment the group behind each row is big enough.
If a report only needs behaviour rather than audience characteristics, switch its reporting identity to the device-based option and the thresholding usually lifts. For analysis that genuinely requires row-level detail, a raw export to a warehouse is the honest route, since thresholding is applied to the reporting interface rather than to the underlying export. Choosing the right identity setting is part of a considered GA4 setup, and it is worth understanding reporting identity before you change it.
