What is attribution?

Attribution follows individual conversion paths. A customer sees a video ad, clicks a search ad, buys; an attribution rule decides how much of that purchase each touch gets. Last click gives it all to the search ad. Multi-touch rules spread it along the path. Platform attribution, inside Google Ads or Meta, credits the platform's own touches. Google Analytics does the same across the touches it can see.

Inside a platform, attribution is the best signal there is. The platform sees far more real-time data about its own auctions and audiences than any outside model, so bidding and campaign optimization belong to it. That is the half that should be said plainly.

Across platforms it breaks in three ways. It needs a tracked touch, so TV, radio, out-of-home, print and most of social's effect on demand are invisible. It credits touches, not causes, so a channel at the end of the journey claims sales that would have happened anyway, and branded search and retargeting look like the best channels in the account. And each platform counts in its own way, so the numbers cannot be added up or compared. Privacy rules and cookie loss make the tracked share smaller every year.

Marketing mix modeling answers the questions attribution cannot: how much to spend in total, how to split it across channels including offline, and what a change would return. It uses aggregated spend and sales, so it needs no tracking and sees every channel on the same terms. Odins does not replace attribution; it complements it. Platform data, together with the other knowledge the customer has, informs the model's priors, and day-to-day reporting tools stay where they are.

• Last-click attribution: All the credit to the final click. Simple, free, and systematically wrong about which channels create demand.
• Marketing mix modeling (MMM): The statistical method behind everything else on this list: spend in, sales out, every channel on equal terms.
• Triangulation: Combining MMM, attribution and experiments. The question is where: inside one model, or in a dashboard afterwards.
• Incremental sales: The sales a channel caused that would not have happened without it. The only kind of sales worth paying for.
• ROAS: Return on ad spend: revenue per unit of spend. Useful, as long as you know which ROAS you are looking at.

Go one level deeper: Why Google Analytics and Meta report different numbers. Triangulation: MMM, attribution and experiments. Or go back to all terms in the glossary.