Attribution Is Broken — Here's How We Actually Measure
Last-click is a lie and multi-touch isn't much better. We stopped trying to perfectly attribute and started trying to actually decide.
Attribution promises to tell you exactly which touch caused the sale. It can't, and chasing that promise has wasted more marketing hours than almost anything else. We stopped trying to attribute perfectly and started trying to decide correctly.
Why the models mislead
Last-click hands all the credit to the final harvesting touch and starves everything that created the demand. Multi-touch spreads credit with false precision based on tracking that privacy changes have quietly broken. Both produce confident numbers that move budget the wrong way.
Holdouts over models
The cleanest signal is a test: turn a channel off in some regions, leave it on in others, and measure the difference in actual revenue. Incrementality beats attribution because it answers the only question that matters — what would have happened anyway.
A north-star you trust
Pick one business-level metric — qualified pipeline, new customers, contribution margin — and judge marketing by whether it moves, not by the sum of channel-reported conversions (which always exceeds reality).
Triangulate, don't worship
- Platform data for in-channel optimisation.
- Holdout tests for true incrementality.
- A simple top-down model to sanity-check the blend.
Perfect attribution isn't coming back. Good decisions never required it. Let's build measurement you can act on.