Attribution is a systems problem

Attribution is often treated as a reporting problem: choose a model, configure a dashboard and decide which channel gets credit. In practice, the quality of the answer is determined long before the report loads.

The chain begins with campaign structure and naming. It continues through consent, browser events, server events, lead capture, identity matching, CRM stages and commercial outcomes. A break anywhere in that chain changes what the dashboard can know.

Define the decision first

Different decisions need different evidence. Budget allocation needs a consistent view of incremental commercial value. Creative development needs fast directional feedback. Sales follow-up needs source context attached to an individual lead.

Trying to make one attribution number answer every question produces false precision. Define the decision, the acceptable latency and the level of certainty required.

Build a shared measurement contract

Marketing, sales and finance should agree what the important events mean. A qualified lead must not have three definitions. Revenue should have a clear source and treatment for refunds, tax and recurring value.

Document event names, required properties, identifiers and owners. Validate the chain from ad click to commercial outcome with test records. Monitor the joins, not only the event volume.

Use models as lenses

Platform reporting, last-click analytics, blended models and experiments each reveal something different. Compare them deliberately. Large disagreement is a diagnostic signal, not an invitation to average the numbers.

Attribution improves when the organisation treats measurement as infrastructure: owned, tested and connected to real decisions. The dashboard is the final surface of that system, not the system itself.