Data & measurement

Attribution models: why last click is lying to you

An attribution model is the rule that decides which channel gets credit for a sale. Last click, the default in almost every platform, always rewards the end of the journey and hides whatever created the demand. The right model depends on the decision you are about to make with the number.

A graphite relay race where runners pass a baton and only the last one carries it in orange

What matters

  • Last click is not a neutral model: it hands 100% of the credit to the closing channel and zero to the one that created the demand weeks earlier.
  • No attribution model measures causality. They all split credit for something that already happened; knowing what caused the sale takes an incrementality test.
  • If every platform counts its own conversions, reported totals exceed the real sales of the business. Unifying the source comes before choosing the model.
  • Pick the model by the decision: moving budget across channels needs a multi-touch model; judging a creative is settled with the platform’s own data.

Your report says Google brought the sale. The content team knows that person showed up three weeks earlier through an article, followed along by email, and only searched your brand at the very end. Both are true: what changes is the attribution model, the rule that decides who gets credited with the goal. Picking it without thinking is how budget moves to the wrong channel while looking data-driven.

What each model rewards

An attribution model splits credit for a conversion across the touchpoints that preceded it. The ones you will find in any platform:

  • Last click. All credit to the final touch. Rewards closing channels: brand search, retargeting, cart emails. It is the default almost everywhere and the easiest one to misread.
  • First click. All credit to the first touch. Rewards discovery and hides whatever sustained the decision for weeks.
  • Linear. Splits evenly across every touch. Honest and not very informative: it treats an impression the same as a booked demo.
  • Position based. More weight to the first and last touch, the rest shared. A reasonable middle ground when the journey is short.
  • Data driven. The platform splits credit with its own model. Useful, with big fine print: it only sees what happened inside that platform.

The problem no model solves

They all split credit for something that already happened. None of them answers what would have happened without that channel, which is the only question that justifies a budget. That answer comes from a different measurement, built on control groups rather than sharing rules: it is the gap between attribution and incrementality.

And there is a problem that comes before the model. If every platform counts conversions by its own criteria, the reported total beats the real sales of the business and nobody knows which number to trust. Before arguing about models, every channel has to report against the same data, offline sales included. That is what a CDP is for.

How to choose without losing your mind

  1. Start from the decision, not the model. Write down what you would do differently depending on the result.
  2. If the decision is moving money across channels, use a multi-touch model and read the series over time, not one month’s snapshot.
  3. Unify the source: one definition of a conversion for every platform.
  4. Whatever moves the needle, validate it with an incrementality test before scaling it.

A well chosen attribution model organises the argument. What ends it is measuring on your own data, which is exactly the job of a data marketing agency.

Frequently asked questions

What is an attribution model in marketing?

It is the rule that splits credit for a conversion across the touchpoints that preceded it. If someone saw an ad, read an article and then bought from an email, the model decides how much of that sale each channel gets. It is not an objective fact: it is a criterion you choose, and it changes how every channel appears to perform.

Which attribution model is best?

There is no best model in the abstract: it depends on the decision. To split budget across channels, a multi-touch model that recognises the full journey works better. To compare two creatives inside one platform, that platform’s own data is enough. And to decide whether a channel earns its budget at all, no model is enough: that takes an incrementality test.

Why do platforms report more conversions than the business actually sold?

Because each one claims the sale if it had any touch inside its attribution window, with no idea what the others did. If three platforms touched the same buyer, all three count the same sale and the total triples. That is why the number that governs is the business number, with every channel reporting against a single source of truth.

Keep reading

Retention & lifecycle 6 min read

Customer retention: the three-stage framework we use

Retention does not get fixed with a reactivation campaign. It gets fixed by placing every customer in their lifecycle stage and measuring where the base stops leaking.

AI & automation 3 min read

What a CDP is and when your business needs one

A CDP unifies every customer interaction into a single actionable profile. What it actually does, how it differs from a CRM, and the signs your business already needs one.

All topics

×
Hi! 👋

How can we help you today?

Please write a message first.