Glossary · Attribuer

Attribution model

An attribution model is the rule that decides which touchpoint gets the credit for a sale, when the person went through several. Changing model does not change what happened: it changes who, in the report, receives the credit for it.

Relier une vente à ce qui l’a produite — et refuser d’inventer quand le fil manque. The forty terms of the trade

What it is for

Without a model, a four-touch journey has no way of being summed up in one row of a table. The model is what makes the report possible — and also what makes it debatable, since two models on the same data produce two different campaign rankings. Choosing it deliberately beats inheriting it by default.

How it is set

You display it next to the figure it produces, never anywhere else. Revenue per campaign read without knowing which model split it is not a figure: it is an opinion. The model is chosen once, written down, and not changed in the middle of a period-to-period comparison.

The classic mistake

Switching model because the new one flatters the campaign you are defending. That is called choosing your conclusion before your measurement, and it always shows: the total sales figure has not moved.

First click

Three rules, three rankings

First

The touchpoint that made the brand known takes everything.

Last

The touchpoint just before the purchase takes everything.

Linear

Every touchpoint in the journey gets an equal share.

The three models applied, on a real report

What Subtraq does with it

Subtraq offers three — first click, last click, linear — and first click is the reference model. In all three, attributed revenue plus unattributed revenue equals the total exactly, to the cent: a model is a reading, never a source.


Read next

Un terme seul ne sert à rien. Voici ceux qui le rendent utilisable.

Linear attribution

Read the definition

Marketing attribution, the pillar page

Open the page

Unattributed revenue

Read the definition

« Attribution model », une fois posé sur de vraies ventes

An attribution model is not understood by reading it: it is understood by watching two campaigns swap places when you change it. The demo opens a workspace filled with fictional data, labelled as such, where the three models are compared on the same sales.

Compare the three models