A calendar total can hide an uneven comparison.
Two sources may appear side by side in a report while their users have had different amounts of time to activate, purchase or return. A newer source can look weaker simply because its cohorts are less mature.
Adjust describes cohorts as groups of users sharing a characteristic, such as installation within a selected period, whose later behavior can be followed over time.
Set a consistent starting point.
Decide whether the analysis begins at install, reattribution or another agreed event. Keep new-user acquisition separate from re-engagement where their objectives and starting conditions differ.
Compare like with like before moving budget.
Agree the observation window.
Pick a window suited to the conversion journey. A first task, an approved account and a returned order can have different validation timelines. Do not treat an incomplete observation window as a finished result.
Check the comparison inputs.
- Population: are country, product and operating system comparable?
- Event: does the action have the same definition and acceptance rule?
- Maturity: have the cohorts had equal time to convert?
- Cost: is the same cost basis used for each source?
- Volume: is the sample sufficient to support the decision, or does the test need more time?
Turn the review into an operating decision.
Use the comparison to identify a specific next step: continue the test, adjust a message, investigate a source or expand a segment. Averages are useful for orientation; source and market context make them useful for action.
How Geostar applies the framework.
We agree the available measurement and review window with the advertiser, then use those definitions when discussing traffic quality and campaign changes. Cohort analysis depends on the data and reporting access available for the engagement.