Attribution windows can change the winner without changing the campaign

Two channels each spend $5,000. One reports 100 conversions and the other 150. Before moving budget, check how long each channel is allowed to claim credit after an interaction.

An attribution window defines an eligible interval between an advertising interaction and an outcome. A longer window can include outcomes excluded by a shorter one. That is a reporting rule, not evidence that delivery became more effective.

Google’s conversion-window guidance describes how these intervals work in Google Ads. Record the actual settings for every source in a comparison.

Separate interaction types

A click, an engaged video view, and an impression can have different eligibility rules. Do not combine their credited outcomes under a vague “conversions” heading without showing the composition.

Suppose a fictional report shows 80 click-attributed purchases and 40 view-attributed purchases. A second platform reports only click-attributed purchases. Comparing 120 with the second platform’s total would conceal a major definition difference.

Google’s view-through conversion documentation explains that viewing without interaction can contribute under its rules. That does not establish that the exposure caused the purchase.

Build a comparison sheet

For each report, record the conversion action, interaction types, window lengths, attribution model, reporting time basis, and whether modeled results are included.

Also note whether the same purchase can receive credit in more than one platform. Adding platform-attributed totals is not a reliable way to count unique business outcomes unless the reporting process explicitly deduplicates them.

The order system can establish recorded purchases. It does not, on its own, assign advertising causality.

Use a worked sensitivity check

Imagine 60 purchases occur within one day of a campaign click, 25 more within seven days, and 15 more within thirty days. The corresponding cumulative counts are 60, 85, and 100.

Those numbers describe the same underlying set of purchases under different windows. They are not three separate performance results to add together.

Where the platform permits comparable views, examine how the conclusion changes with the window. If a channel looks stronger only under a substantially longer rule, explain that dependence.

Avoid changing the rule to improve the score

Choose windows that fit the measurement question and business journey before judging results. If the definition changes, mark the effective date and distinguish periods where comparability is limited.

Do not assume a setting change rewrites all historical data in the same way across systems. Confirm the destination’s documented behavior.

For causal questions, use a suitable design such as the approach described in the incrementality-testing guide. Attribution is useful for operating campaigns, but it answers a different question.

A clear report can show several attribution views without declaring one universally correct. It should make the chosen basis visible enough that another reader can reproduce the comparison.