What is marketing attribution?
A customer rarely buys after one exposure. They see a film on Instagram, search the brand a week later, click a paid listing, read a review, get a retargeting ad and finally buy from an email. Marketing attribution is the method for deciding which of those moments gets credit for the purchase. The simplest models give all the credit to the last click or the first touch. Multi-touch models split it across the journey, by rule or by statistical weight. Marketing mix modeling steps back further and estimates each channel's contribution from aggregate spend and sales data, without tracking any individual.
Attribution is both a technical setup and an opinion. The setup is tracking: pixels, tags, UTM parameters, conversion events and the analytics that stitch them together. The opinion is the model, because every model embeds a belief about how marketing works, and the choice changes which channels look good.
Why it matters
Why marketing attribution matters
Budget follows evidence, and attribution is the evidence. Without it, a brand funds the channel that is easiest to measure or the one its most persuasive manager runs. With it, a brand can see that the search campaign only converts people the brand film already reached, or that the retargeting ads are taking credit for sales that would have happened anyway. Those are expensive things not to know.
It matters for creative too. Attribution is how a brand learns that the vertical cut outperformed the horizontal one, that the second hook worked better than the first, that the content people saw before they bought was not the content the brand thought was selling. For a beauty brand running dozens of creative variations, that feedback is the difference between guessing and learning.
In practice
How marketing attribution works in practice
- 01
The foundation is tracking that works. Every campaign is tagged, every conversion is defined and recorded, and the analytics platform can connect an ad view or a click to an eventual action. Then a model is chosen and applied consistently, so the numbers mean the same thing month to month. Reports are built around the decisions they inform: which channels to grow, which to cut, which creative to iterate.
- 02
The practice has become harder. Privacy changes have reduced what can be tracked across sites and devices, and platforms each report in their own favor, so the sum of what every channel claims is usually more than the brand actually sold. Serious teams now triangulate: platform attribution for day-to-day optimization, marketing mix modeling for budget allocation, and controlled tests, such as holding a region out of a campaign, to check the models against reality.
At RIOT
Marketing attribution at RIOT
The two ends of the problem show up in our own work. The Meta ad engine we ran for Homebase from February to April 2025 lived inside a platform where attribution is native: each piece of creative had conversion data next to it within days, and the next round was made with that in view. The CTV spot "Real Life, Worry Free" for Bogg Bags sits at the other end. Connected television reaches people on the biggest screen in the house and leaves far less of a trail, which is where modeling and judgment have to carry what the pixel cannot.
Watch out
What people get wrong about marketing attribution
Believing the platform's number. Every ad platform attributes generously to itself, and a brand that adds up the dashboards will think it sold twice what it did. The second mistake is choosing last-click because it is easy, which starves every channel that works early in the journey, including the brand film that started it.
The third is treating attribution as the truth rather than an estimate. Models are opinions with math attached; the right use is to compare channels and inform decisions, not to settle arguments to the decimal. The fourth is ignoring what cannot be tracked. Brand work, word of mouth and a billboard outside the store all move sales and none of them carry a UTM, which is why mix modeling and tests belong alongside the pixel.
