Marketing attribution assigns credit to recorded interactions under a defined set of rules or an algorithm. It helps explain the evidence available to a reporting system. It does not reveal every influence on a customer or prove that one channel caused a sale.
Start with the business question
Decide whether you are investigating first discovery, an enquiry's immediate source, assistance during consideration or credit for completed business. These questions need different records and dimensions. Give reports names that describe the question rather than labelling every source field as attribution.
A customer might discover a business through a recommendation, read several pages and later follow an advert. Their own account of discovery and the platform's recorded source can both be useful without agreeing. Preserve those meanings instead of forcing one to overwrite the other.
Record the model and eligible channels
In GA4, the reporting attribution model governs credit for key events in relevant event-scoped reports. Rules-based and algorithmic models can distribute credit differently, including fractional credit. First-user and session dimensions describe separate scopes and are not changed simply by choosing another reporting attribution model.
Record the model, the report's dimensions and which channels can receive credit in the relevant configuration. Integrations with advertising products can affect which settings apply. Check the actual property and linked accounts before interpreting a total.
Understand windows and changes
A lookback window limits how far back an eligible interaction can be considered for attribution. A conversion window in an advertising product governs a different recording context. Neither is a complete map of a commercial relationship that developed over months.
Keep a dated record of setting changes. In GA4, changing the reporting model can affect relevant historical and future reports, whereas changing a lookback window applies going forward. A comparison can therefore change even when the underlying commercial activity did not.
Inspect collection and campaign handling
Document manual campaign tags, automatic tagging, redirects and connected systems. Consistent source and medium conventions help interpretation, but UTM tags alone do not describe every acquisition mechanism. Test the finished links and the actual reporting destination.
Review consent, device changes, identifier availability and missing journeys. Someone may phone after an unrecorded recommendation or return on another device. Show material limitations alongside the report rather than treating unknown activity as proof that a channel contributed nothing.
Reconcile operational outcomes
Agree how an enquiry becomes qualified and when it becomes completed business. Compare the website, advertising and CRM records using those definitions. Separate an event count, a unique user and a business enquiry where the systems measure different things.
Investigate sample records, including duplicates, cancellations and enquiries outside the service area. Identify who resolves mismatches. Avoid publishing a combined total that counts the same outcome separately in several systems.
Use attribution alongside other evidence
Read attributed outcomes with customer feedback, actual costs, operational capacity and suitable experiments. A model's credit allocation is different from a causal estimate of additional business generated by a campaign. State that distinction when deciding budgets.
For a long buying cycle, keep enquiry cohorts and later outcomes visible where possible. Allow time for genuine decisions to mature. Do not compare a fully developed cohort with a recent one as though both had the same opportunity to convert.
Reporting checklist
- Name the decision and acquisition scope.
- Record the attribution model and eligible channels.
- Document windows, dates and configuration changes.
- Test campaign links and connected collection.
- Reconcile meaningful outcomes with the CRM.
- Explain missing evidence and avoid causal claims without suitable support.
Our analytics validation guide covers event definitions. The paid search guide explains bidding goals and conversion settings.
