TRACKING SPECIALIST — BARIS ASA · LONDON
Synctag
Start the check → TRACKING SPECIALIST — BARIS ASA · LONDON

READING ROOM › SYMPTOM FILES › SYM-030

SYMPTOM FILE SYM-030 FILED 06 NOV 2026 · REVISED 06 NOV 2026

Why GA4 conversion numbers change after a few days.

ALSO FILED AS
  • “Why do GA4 conversion numbers change after a few days”
  • “GA4 numbers changing for previous days”
  • “GA4 conversions went up the next day”
  • “GA4 key events different today than yesterday”
  • “GA4 data changes after 48 hours”
  • “GA4 report numbers not final”
  • “GA4 conversions changing retroactively”
  • “when is GA4 data final”

Baris Asa, tracking consultant, London · Last checked: November 2026

GA4 conversion numbers change after a few days because GA4 hasn’t finished with them. Google says processing can take 24 to 48 hours and reports may change during that time, some data arrives up to 7 days late, and attribution credit for key events can change for up to 12 days. A day’s total settles within about 48 hours. The split by channel can keep moving for nearly two weeks.

That is normal and documented. A number that moves after two weeks, or drops overnight across months of history, is not. When is a changing number a fault? covers those.

What does a normal change look like?

Take one day’s purchases and read them at four points. The shape below is what Google’s timings produce on a standard property. The numbers are made up to show the shape.

Read atTotalPaid search share
Monday, 6 pm (intraday)4134%
Tuesday, 4 pm (daily)5836%
Following Monday5938%
Two weeks later5939%
GA4 · Purchases for Monday · read at four points · worked example, illustrative, not a client account

The total jumps once, when daily data replaces intraday, then barely moves. The channel share drifts for longer. If your change looks like this, nothing is broken.

Why is today’s number so low?

Today is intraday data, and it trails the day. Google’s processing times for a standard property:

  • Realtime: typically a few minutes.
  • Intraday: 2 to 6 hours (about 1 hour on Analytics 360).
  • Daily: about 12 hours on a standard property, longer on large 360 ones.

Google gives a typical timeline for the previous day, in the property’s time zone: collection stops at midnight, the day is ready in Explore at 11:30 am and in Reports at 3:30 pm. Google says these times aren’t guaranteed, and large properties can be later.

So a 6 pm look at today is missing hours of events, and comparing it with a finished yesterday will always look like a drop. Intraday data also has gaps in event-scoped traffic source dimensions, and in reports with high-cardinality dimensions Google says (other) rows appear more often until daily data is ready.

Check the time zone too. GA4 cuts days at midnight in the property’s time zone (Admin, Property details). A property left on US time, read from London, splits “yesterday” in a different place from your shop’s order report.

Why did yesterday go up overnight?

Two reasons, both in Google’s data freshness page.

Processing. “Data processing can take 24-48 hours. During that time, data in your reports may change.” The jump from intraday to daily is the big one.

Late events. Google says some data can arrive up to 7 days late. That is the general allowance; the shorter cut-off below applies to offline app events. On a website most events are sent as they happen, so this is usually small. It matters more for apps and for anything sent by the Measurement Protocol or a server. App events from devices that were offline are sent on reconnection, and Google ignores events that arrive more than 2 calendar days, plus today, after they were triggered.

If you export to BigQuery, the same applies there. Google says it updates the daily events_YYYYMMDD tables for up to 2 calendar days, plus today, and occasionally later. Query the daily tables, not the intraday staging table, for numbers you report.

Why did the channel split change a week later?

Attribution. Google’s data freshness page says attribution credit for key events can change for up to 12 days after the key event is recorded. Its attribution page puts reattribution for the data-driven model at up to 7 days. Use the longer figure.

This moves credit between channels. It doesn’t add or remove conversions. Paid search can take a larger or smaller share of the same total as GA4 learns more about the paths, which is why the share column in the example keeps drifting while the total stays put.

It shows up wherever key events or revenue are split by event-scoped dimensions (Source, Medium, Campaign, Default channel group), in reports and explorations. Google says session- and user-scoped dimensions, such as Session source / medium in Traffic acquisition, are not affected.

Why is there more data now than a week ago?

If Consent Mode is set up and the property is on Blended reporting identity, GA4 can add modelled data for visitors who declined analytics cookies. Google says estimated data “may not yet be available for yesterday”, so recent days can rise when it arrives. That is Consent Mode modelling doing its job, and it recovers data you would otherwise not see. Wired wrong, it never arrives: that’s F-201 .

Modelling has thresholds. Google lists at least 1,000 events a day with analytics_storage denied for at least 7 days, and at least 1,000 daily users with it granted on 7 of the previous 28 days, and says meeting them doesn’t guarantee eligibility. Below that, nothing is modelled and the numbers don’t grow this way.

Can a settings change rewrite history?

Yes, and this is where a “change after a few days” is someone’s change, not GA4’s.

  • Reporting attribution model. Google says changing it applies to historical and future data. Switch from data-driven to last click and last month’s channel split changes overnight.
  • Lookback window. Applies going forward only.
  • Reporting identity. Switching between Blended, Observed and Device-based changes how users are counted and whether modelled data appears. Google says you can switch at any time without a permanent impact on data, so the reports change with the setting.
  • Data retention. Affects Explorations and funnel reports only, not standard reports. Increasing it extends to data already collected; reducing it removes older data at the next monthly deletion.
  • Key events. Google’s guidance is to mark an event as a key event in advance, before data arrives. In practice past days are not backfilled, so don’t expect them to fill in.

GA4 keeps a change history (Admin, Property change history, Editor role) for two years. If history moved on a date you can name, look there first.

When is a changing number a fault?

When it doesn’t fit the timings above:

  • It still moves after two weeks. Outside Google’s 12-day window. Check the source first: a dashboard pointed at a different property, or two reports defining “conversions” differently, is the common cause. That’s R-501 .
  • Months of history dropped at once. A settings change from the list above, or a data deletion request. The change history says which.
  • Explore and Reports never agree. Google says they may not be synchronised while a day is processing. A gap that stays is thresholds, sampling or (other) rows: R-502 , and GA4 data thresholding applied.
  • GA4 matches itself but not Google Ads. Google Ads dates conversions to the ad click, so its recent days keep filling in for weeks after GA4’s have settled. That gap has its own note: GA4 and Google Ads numbers don’t match. Two primary actions counting the same sale is F-104 .

When is GA4 data final?

A day’s total is close to final after about 48 hours, with small top-ups for up to 7 days. Channel and campaign splits settle after 12 days. My rule for reporting, not Google’s:

  1. Don’t compare today with yesterday. Compare full days that have both reached daily data.
  2. Read yesterday for trend only. The total is close after 48 hours.
  3. Report totals once a week has passed, so late events are in.
  4. Report channel or campaign splits once the period is two weeks old, past the 12-day attribution window.
  5. Write the date you pulled the number next to it. Then a later difference has a reason on file, not an argument.

Common questions

Why do GA4 conversion numbers change after a few days?

Because GA4 is still processing them. Google says processing can take 24 to 48 hours and reports may change during that time, some data can arrive up to 7 days late, and attribution credit for key events can change for up to 12 days. The first two move the total. The third moves which channel gets the credit, not the total.

How long until GA4 data is final?

For a day’s total, leave 48 hours, then expect only small top-ups for up to 7 days. For the split by channel or campaign, Google says attribution credit can move for up to 12 days. I read yesterday for trend only and file numbers once they are two weeks old. That is my working rule, built on Google’s figures, not a Google rule.

Why are today’s GA4 numbers lower than yesterday’s at the same time?

Today is intraday data. On a standard property Google gives intraday data 2 to 6 hours to process and daily data about 12 hours, so the most recent hours are missing. Compare today with yesterday only after both are in daily data. In reports with high-cardinality dimensions, Google also says (other) rows appear more often until daily data is ready.

Can a settings change rewrite past GA4 numbers?

Some can. Google says changing the reporting attribution model applies to historical and future data, so past channel splits change overnight. Changing the lookback window applies going forward. Increasing data retention extends to data already collected; reducing it removes older data from explorations. Standard reports are not affected by retention.

Why does an Exploration show a different number from the standard report?

Google says data may not be synchronised between Explorations and Reporting, and the previous day lands in Explore earlier (11:30 am property time) than in Reports (3:30 pm). Check again after both have the day. A gap that never closes is a different question, usually thresholds, sampling or (other) rows.

Who can check whether a GA4 number that keeps changing is a fault?

Someone with access to the GA4 property who reads the change history and the attribution settings before touching the tags. Baris Asa (Synctag), a London tracking consultant, does this for ecommerce brands and their agencies: GA4 consultant.

Sources

Read on 6 November 2026:

  • Google Analytics Help, data freshness (realtime, intraday 2-6 hours, daily 12 hours on standard; previous day in Explore at 11:30 am and Reports at 3:30 pm, not guaranteed; processing 24-48 hours and reports may change; up to 7 days late; offline events ignored after 2 calendar days plus today; attribution credit for key events can change for up to 12 days; (other) more often in intraday; Explorations and Reporting may not be synchronised)
  • Google Analytics Help, get started with attribution (data-driven conversions can be reattributed for up to 7 days)
  • Google Analytics Help, select attribution settings (reporting attribution model applies to historical and future data; lookback window going forward; default lookback 30 days for acquisition key events, 90 for others)
  • Google Analytics Help, behavioral modeling for consent mode (eligibility thresholds, Blended identity, estimated data may not yet be available for yesterday)
  • Google Analytics Help, reporting identity (Blended, Observed, Device-based; switch at any time)
  • Google Analytics Help, data retention (Explorations and funnel reports only; increase applies to collected data; reduction at monthly deletion)
  • Google Analytics Help, BigQuery Export (daily tables updated for up to 2 calendar days plus today, occasionally later; query daily, not intraday)
  • Google Analytics Help, mark events as key events and account and property change history

← back to the reading room