You open a GA4 report, and some of the data is simply gone, replaced by a small warning icon. That vanishing act is data thresholding, and it confuses more marketers than almost any other GA4 quirk. Here is what it is and how much it still matters.

What thresholding does

Data thresholding is a privacy mechanism. GA4 withholds rows when the underlying user counts are so small that an individual could, in theory, be identified – especially when demographic or Google-signals data is involved. Instead of showing you two users in a narrow segment, GA4 hides the row so no one can reverse-engineer who they were.

Why it happens less than it used to

The important update most old guides miss: on February 12, 2024, Google removed Google Signals data from GA4's reporting identity. Because signals-based identity was the main trigger for thresholding, that change eliminated it from a large share of standard reports. The advice you will still read – “switch your reporting identity to Device-based to kill thresholding” – is now mostly obsolete, because signals are no longer part of that setting at all.

When you'll still see it

Thresholding has not disappeared entirely. You are most likely to hit it in these cases:

  • Reports that include demographics and interests (age, gender, interests), which still depend on signals-derived data.
  • Very small segments or short date ranges where counts are tiny.
  • Reports where Google Signals data is explicitly in play.

How to work around it

When a row you need is being withheld, a few moves usually bring the data back:

  • Widen the date range so the counts clear the threshold.
  • Remove demographic dimensions from the report or exploration.
  • Use the BigQuery export, which delivers 100% of your events with no thresholding applied.

That last option is the durable one. Thresholding is a property of the GA4 reporting interface, not of your data, so raw event-level access in BigQuery sidesteps it completely.

You open a report and the data is gone – but now you know why, and how little of it survives the February 2024 change. Thresholding is a shrinking privacy guardrail, not a wall: widen your ranges, drop demographic dimensions when you can, and lean on BigQuery when you need every row. The icon that once hid so much of your reporting now stands in the way far less than it did.

Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.