Why Your Dashboard Numbers Keep Changing: GA4 Delays, Thresholds and Retention Explained

"The numbers were different yesterday." Anyone who looks after a website's analytics hears this a lot, usually after a client takes a screenshot of Monday's report, looks again on Wednesday and sees more sessions, a changed conversion count or a chart line that has drifted. That makes the data seem unreliable, but in most cases it's still arriving.

Google Analytics 4 and Search Console both change after the fact, and Google's own documentation explains why. Processing takes time, late events are dropped after a cut-off, small groups of users are hidden for privacy, older detail expires, and rare searches never show up in query reports. These changes are expected behaviour and don't mean anything is broken, and knowing the reasons lets you read any dashboard calmly. We've listed every reason we know of below, and each one comes from Google's own documentation.

The reasons at a glance

1. Processing delays

GA4 can take 24-48 hours to process data, according to Google, so figures in your reports may shift during that window. Realtime reports show activity within minutes. Standard reports fill in over the next several hours, though some data from the previous day can show up as much as 7 days late.

Google's data freshness page states: "Data processing can take 24-48 hours." Elsewhere it adds that "some data can arrive late, potentially up to 7 days delayed." Intraday data on standard properties typically updates within hours, but the complete daily figures take longer.

What to do: End the date range two or three days back when you compare periods, or mark the latest days as provisional. A "final" weekly report that includes Sunday shouldn't go to a client on Monday morning.

2. Late events are dropped

GA4 ignores events that arrive more than two calendar days, plus today, after they happened. This mostly affects apps that send data when they reconnect and server-side setups that queue events. Events sent too late never appear in reports.

The same page states that "Google Analytics ignores events that arrive more than 2 calendar days, plus today after the events are triggered." If your tracking sends events in batches, for example from an app that's often offline, make sure they go out quickly, or some of your activity will simply be missing.

3. Data thresholds

GA4 hides some rows to stop anyone working out the identity of individual users. Thresholds apply when a report includes demographic data or audiences built from it, and to some search query data when there aren't enough users. They're set by Google and can't be adjusted.

GA4 applies data thresholds "to prevent anyone viewing a report or exploration from inferring the identity or sensitive information of individual users." Google is clear that "data thresholds are system defined. You can't adjust them."

What to do: widen the date range so groups get bigger, remove demographic dimensions you don't need, or export raw data to BigQuery if you need unthresholded detail.

4. Data retention

Standard GA4 properties keep detailed user-level and event-level data for 2 or 14 months, depending on the setting. That setting limits explorations and funnel reports, while the standard aggregated reports aren't affected. Age, gender and interest data always lasts just 2 months.

In Google's data retention page, the choices for standard properties are "2 months" or "14 months", and longer periods exist only on Analytics 360.

What to do: Switch retention to 14 months now, and your explorations will reach far enough back to compare against the same period last year.

5. Consent and modelling

If consent is required and a visitor declines analytics cookies, GA4 can't track them normally. Behavioural modelling can fill in part of that activity, but only for properties that reach Google's minimums, and most small business sites fall short.

The prerequisites are in Google's behavioural modelling page: your property has to collect "at least 1,000 events per day with analytics_storage='denied' for at least 7 days" and also have "at least 1,000 daily users sending events with analytics_storage='granted' for at least 7 of the previous 28 days."

For a small business, those visitors usually just go uncounted. After you add or change a cookie banner, GA4 figures will probably fall, and that doesn't mean you lost traffic. Search Console doesn't rely on your site's tag or cookies, so it's the better way to tell whether search visibility actually changed.

6. Hidden Search Console queries

Google leaves anonymized queries out of the query report, which means queries that no more than a few dozen users issued across two to three months. On our own site, queries we couldn't see accounted for about 72% of clicks.

In our Search Console data study, over 90 days the page report counted 789 clicks, and the query report covered only 218. The rule is described in Google's performance data deep dive, so rely on page totals for your headline number and read the query list as a sample.

7. GA4 and Search Console never match

The two tools measure different things: Search Console records clicks on Google results, while GA4 records sessions once your tag has loaded, from every source, and relies on consent. Exact agreement between them isn't possible, and that's fine. Our post on GA4 vs Search Console goes through each difference, and it also covers a case where the GA4 organic sessions on our own site were almost double the clicks in Search Console, which is the reverse of what most people expect.

A simple rule set for reading any dashboard

1. Treat the last two or three days as provisional.
2. Compare equal periods, ideally the same weeks last year.
3. Watch the direction of the trend, not single-day spikes.
4. Use page-level Search Console totals, not the sum of queries.
5. Set GA4 retention to 14 months.
6. Write down every tracking or consent change with its date, so you have an explanation ready when the numbers suddenly move later.

Our guide to business analytics dashboards is there to read, and the free ONE Dashboard gives you a dashboard that already applies these rules. If your numbers suddenly fell and none of these reasons explains it, begin with our traffic drop checklist.

Sources

Each source is linked above: the Google Analytics Help pages on behavioural modelling, data retention, data thresholds and data freshness, Google's Search Central blog, and the Search Console figures for greatinflux.com.