Attribution Model

An attribution model decides how credit for a sale or conversion is divided between the touchpoints a customer interacted with before buying. Because the model determines which channels appear to be working, it directly shapes where marketing budget goes.

Why the choice of model matters

The same set of orders can produce very different channel reports depending on the model applied. A last-click model will make branded search and retargeting look strong, because those are usually the final interaction before purchase. A first-click model will favour the channels that introduced the customer, such as paid social or display. Neither is wrong in itself, but each answers a different question, and reading one as though it answered the other is how budget gets misallocated.

Common attribution models

• Last-click — all credit to the final interaction. Simple and stable, but systematically overvalues bottom-of-funnel channels.
• First-click — all credit to the first interaction. Highlights discovery, ignores everything that closed the sale.
• Linear — credit split equally across every touchpoint. Balanced, but treats a passing impression as equal to a decisive click.
• Time-decay — more credit to interactions closer to the purchase. A middle ground, though the decay rate is an assumption rather than a measurement.
• Position-based — weights the first and last interactions most heavily, splitting the remainder across the middle.
• Data-driven — weights are derived from observed conversion patterns rather than a fixed rule.

The limits of attribution

Every attribution model shares a structural constraint: it can only divide credit among interactions that were actually recorded. Touchpoints that were never tracked, and demand that would have converted without any marketing at all, are invisible to it. Cookie restrictions, cross-device journeys, and platform-level reporting differences all shrink what attribution can see, and each ad platform's own model tends to favour its own channel.

That is why attribution is best read as a description of recorded journeys rather than proof of causation. Answering "did this spend cause incremental sales?" requires an experiment, not a division of credit.

How to use attribution alongside other methods

Attribution is most useful for understanding journey composition and for day-to-day channel monitoring. Pair it with methods that measure causality: incrementality testing establishes what a channel actually caused, and Marketing Mix Modeling measures full-mix contribution from aggregate data without depending on user-level tracking.

For a direct comparison of the two approaches, see our attribution versus MMM breakdown, or explore Dema's unified measurement stack.