Write down three separate settings
The reporting period selects the results you are reviewing. The lookback window defines eligible earlier interactions. The attribution model allocates credit among eligible observations. Changing the window can change the set of touchpoints before a model allocates anything. It does not create new company revenue.
In GA4, Google documents different lookback defaults for acquisition events and other key events, plus a separate engaged-view setting. Changes apply going forward. Verify the actual property and report instead of assuming that an account uses the default. Record the settings with the export used in the trading meeting.
Relate the window to the shopping question
A considered occasionwear purchase may involve several visits, while an urgent replacement purchase may happen quickly. Use observed purchase timing to frame a question about the window. Avoid choosing a longer setting simply because it makes a campaign report more revenue.
For a review, write down the hypothesis before comparing numbers: perhaps an earlier collection launch interaction falls outside one report’s eligibility. Check what path evidence is actually available. An interaction that was never observed cannot be recovered just by extending a setting.
Compare reports with a settings register
For each source, record the purchase event, interaction type, window, attribution model, timezone and revenue basis. Also record whether the report is grouped by purchase date or interaction date. Highlight unknown settings instead of treating them as identical. These notes often explain why superficially similar totals should not be added together.
Keep a fixed order cohort when investigating a reporting difference where the available data permits it. Separate the effect of changed eligibility from differences in identity coverage or event collection. If the evidence cannot isolate a cause, record the remaining uncertainty and the next check.
Turn the difference into a reviewable decision
The output should be a short comparison explaining which question each report supports. A platform view may support creative review; an agreed order source supports the commercial total. Use the attribution pillar to connect those views without presenting one convenient window as the universal truth.
Before altering measurement settings, assign an owner and note the change date. Explain to the team how future comparisons will be annotated. A visible measurement change prevents a reporting shift from being mistakenly celebrated as a sudden improvement in the underlying business.
A purchase on day twenty
Illustrative path: a shopper clicks a collection ad on day zero and purchases on day twenty after an email on day nineteen. Under a simple seven-day lookback, the early ad is outside the eligible period; under thirty days, it can remain eligible. The model still determines how credit is allocated. This arithmetic example does not reproduce any platform’s complete identity or attribution logic.
Turn the guide into a useful review
- Record reporting dates, window and model separately.
- Check eligible interaction types and the purchase event.
- Keep revenue and date definitions visible in comparisons.
- Document setting changes and unresolved evidence gaps.
Reference material
Platform guidance checked 3 October 2026. Examples and working checklists are Faccelerate editorial illustrations.
Bring the question.
Keep the context.
Map the relevant sources and specialist workflow for your team.
Explore your stack
