Attribution modelling starts with a better question

A customer sees a campaign, searches for the company a week later and eventually buys after receiving an email. Who gets the credit?

Before choosing an attribution model, I would ask another question: what decision are we trying to make? Budgets, agency relationships and people’s confidence in their work deserve a clearer basis than a competition for credit.

Start with the decision you can actually take

“Understand the customer journey” sounds reasonable, but it is a broad brief. “Decide whether to increase search spending next quarter” gives the analysis a purpose.

Write down the decision, the available options and the evidence that would change your mind. Include commercial constraints such as margin, sales capacity and the time between enquiry and purchase.

Consider a hypothetical business comparing two campaigns. One generates inexpensive enquiries. The other produces fewer enquiries, but those customers buy more valuable services. A report focused only on the cost per enquiry could encourage the wrong budget decision.

This is where the commercial conversation belongs: before the dashboard has made one number feel definitive.

Understand what the model is assigning

Attribution assigns credit for an outcome across the interactions a reporting system can observe.

Google’s GA4 documentation describes data-driven attribution, paid and organic last click, and Google paid channels last click. These approaches allocate credit differently. Changing the approach can therefore change the apparent contribution of a channel.

That allocation does not, by itself, establish how many purchases would have happened without the marketing. It describes credit under a model, while estimating additional sales requires a comparison with what would otherwise have happened.

A hypothetical customer might already intend to buy, then click a paid search advertisement shortly before ordering. The click is observable. The customer’s alternative behaviour is harder to establish.

Keep that distinction in the discussion. A precise-looking share of credit can still sit alongside uncertainty about incremental impact.

Use attribution alongside other evidence

I would want attribution reports beside sales quality, customer feedback, margin and changes in overall demand.

Where practical, a carefully designed experiment can help investigate incrementality. Where an experiment is unavailable, be explicit about the limits of the comparison and the other changes happening at the same time.

Agree reporting definitions with the people using the numbers. Check date ranges, conversion definitions and whether teams are comparing the same outcomes. Record important changes so that next month’s meeting does not begin with another argument about the baseline.

Attribution becomes useful when it helps a team make a better decision and understand the uncertainty attached to it.

The strongest meeting outcome may be a bounded next step: increase a budget modestly, define what success would look like and review the evidence on an agreed date. That is something the business can act on.