Marketing Attribution Models: Find Your True Revenue Drivers
Last-click attribution is lying to you. How to move to multi-touch attribution and allocate budget based on true revenue contribution.
Attribution doesn't tell you the truth. It tells you a version, and you pick which one.
Every model is wrong in a different way. The point isn't to find the correct one, it's to know how the one you use lies.
How each one lies
Last click — the default everywhere. Gives all credit to the final touch, which is almost always a brand search or a retargeting ad. Result: you undervalue everything that brings new people in and cut budget from exactly there.
First click — the inverse. Gives all credit to discovery and ignores what closed the sale. Useful as a secondary report, dangerous as the only one.
Linear — splits evenly. Simple, but pretends a banner seen once counts as much as a demo.
Data-driven — the platforms' model. Best in theory, but it's a black box that can't see channels outside its own platform.
What broke every model
Cookie windows shortened, and a significant part of the journey became invisible: apps, messaging, conversations, searches on another device.
In practice, 20–40% of a customer's path appears in no report. Not because it's badly configured, but because it can no longer be measured.
What to do instead of certainty
The switch-off test. Turn a channel off for two weeks and watch total sales, not channel-reported ones. It's the simplest and most honest method available.
Just ask. "How did you hear about us?" at checkout or on the first call. The answers are messy, but they catch exactly the channels tracking can't see.
Account-level correlation. Raised budget on a channel by 30%? Did total sales rise? It's crude, but cannibalisation can't fool it.
The practical rule
Use attribution for optimising within a channel — which campaign, which ad, which keyword. It's reliable there.
Don't use it for splitting budget between channels. For that you need switch-off tests and judgement.
The signal that your reporting is wrong
The sum of conversions reported by all platforms exceeds your actual order count. It happens almost everywhere, and it means each platform is claiming the same sale.
How many hours a week do you lose to repetitive work?
In 20 minutes on Zoom we work out which of them can run on their own. You leave with the answer, whether we work together or not.
Free call, no obligations. You leave with a concrete list of problems and automation opportunities, whether or not you choose to work with us.