It is common to observe a gap between the sales recorded in your back office and the data displayed in Google Ads. This phenomenon, often wrongly interpreted as a tracking incident, is inherent to how Google's attribution models work.
This technical documentation details the latency mechanisms, in particular the impact of click-based attribution and consent-related modeling (Consent Mode).
Unlike traffic analysis tools (Google Analytics 4, CRM) which date the conversion at the time of the transaction, Google Ads attributes the conversion to the date of the ad click.
This is the main cause of data discrepancies: statistics are rewritten retroactively.
Example of an attribution flow:
Monday the 1st: The user clicks on an ad (cookie set).
Thursday the 4th: The user returns to the site and converts.
Google processing: The conversion is credited to Monday the 1st (day of the engagement) and not to Thursday the 4th.
Analysis impact: If you check Monday's statistics on Tuesday morning, this conversion is invisible. It will appear later by modifying past data.
Beyond attribution, the technical processing of signals adds an incompressible latency, particularly reinforced by regulations (GDPR / ePrivacy).
The time needed to validate server "logs" and reconcile identifiers (e.g. click on mobile, purchase on desktop) requires a delay.
Average delay: 3 to 48 hours.
When using a CMP (such as Sirdata), if a user refuses statistics/marketing cookies, Google Ads uses Consent Mode to model the lost conversions via AI.
This statistical modeling process is not performed in real time. Google must aggregate enough data to extrapolate the results accurately.
Observed delay: Modeled conversions can take up to 7 days to appear in the reports.
The delay also comes from the user's natural behavior. Depending on the sector (B2B, real estate, complex purchases), the decision window can extend over several weeks.
As long as the conversion has not taken place, no data is reported. Once it happens, it is subject to the retroactivity rule described in point 1.
"Same Device" conversion
3 to 24 hours
Standard processing
"Cross-Device" conversion
24 to 48 hours
Identifier reconciliation
Modeled conversion (Consent Mode)
5 to 7 days
Algorithmic computation (AI)
Offline import (CRM)
24h after upload
Processing of the imported file
For technical reference, here are the official Google Ads support documentations validating these behaviors:
On the date logic (Retroactivity):
"Conversions are reported based on the date and time of the ad click. [...] If a user clicks your ad on January 1 and makes a purchase on January 8, the conversion is counted on January 1." Source: About conversion tracking windows
On data latency:
"Most conversions are imported within 3 hours. However, other conversion types can take up to 24 hours or more." Source: About data freshness
On modeling and privacy:
"Conversion modeling uses Google AI to attribute conversions [...] when we cannot directly observe conversions (for example, due to user privacy settings)." Source: About modeled conversions
Mis à jour
