Divergence does not always mean loss of conversions.
Two systems can show different numbers because they use different rules. Before searching for missing FTDs, you need to check the event definitions and reporting period.
Start with the same window.
The most common reason is different time zones. If one system closes the day at UTC and the other closes the day at local time, the events around midnight will end up on different dates. Compare the same period up to the clock.
Checklist
| Verification | Question |
|---|---|
| Time. | Is it the same time zone and time period? |
| Event. | Is registration/FTD defined the same way? |
| Deduplication | Can a single system remove repeats? |
| Redirects | Does click_id still exist throughout the chain? |
| Attribution | Is the window the same as the last click rule? |
| Status. | Do you compare raw FTD to confirmed FTD? |
Look not for a general difference, but for specific lines.
The phrase "I have 120, you have 108" is almost useless. A much stronger list of twelve identifiers that are in one system and not in another. This allows you to check every case.
Separate technical and business logic
If it doesn’t happen at all, it’s a technical challenge. If it came, but received a different status, you need to look at the confirmation rules. Mixing these causes stretches the diagnosis.
Normal reconciliation process
- Choose one day and one source.
- Export events from both systems.
- Normalize time.
- Comparison by click_id/SubID.
- Separately analyze the missing and status differences.
- Only then can we conclude the scale of the problem.
Prevention
Stable circuit SubID and tested postback Reduce the number of situations where data needs to be retrieved retroactively.
Acceptable level of discrepancy should be understood in advance
Absolute coincidence of different systems does not always happen: one can take into account blockers, the other - only server events, the third - its own attribution window. It is important not to ignore the difference, but to know its normal range. A sharp exit from this range is a stronger signal than constant small differences.
Check the “event funnel,” not a single number.
If clicks match, registrations diverge and FTDs are close again, the problem is not where the "clicks diverge from the start" situation is. Comparison of several stages helps to localize the source of the discrepancy.
After the correction, make a check day.
Do not close the task immediately after the change. Select the next full day where both systems work with the new setup, and repeat the reconciliation. Otherwise, a correction may look successful only because some of the old events have not yet been processed.
Start the incident note.
Briefly record the date, cause, campaigns affected, and decision. If a similar discrepancy reappears, history will test the known cause of the former.
Create a table of sources of truth
For each metric, define the system that you consider to be the main one. For example, expenses – advertising cabinet, confirmed by FTD – partner system, internal transitions – web analytics. Then the team does not choose a convenient number from three different interfaces.
Differences in definitions are more important than interfaces
Two dashboards can call a conversion metric the same way, but count it by different dates: one by the date of the click, the other by the date of the event. In the documentation for the report, fix the definition, not just the name of the column.
Escalation to manager
Transfer a specific file with identifiers, period, time zone and expected status. The better a reconciliation is prepared on your side, the faster the other side can check their log.
