Attribution
Attribution.
Done properly.
It’s been the beating heart of Ringside since day one.
Productising what we preach
Russell, our CEO, has spent twenty years thinking about attribution. He’s worked with hundreds of businesses, written courses that have trained thousands of marketers and spoken about it at conferences around the world. Google him and see.
His view has stayed pretty consistent: credit for a sale should come from what people actually did on the way to it, not from an arbitrary rule somebody chose years ago. If a tool gives you six attribution models and asks you to pick one, it hasn’t really answered the question.
Ringside is his thinking turned into a product. One model, trained on your own data, valuing each visit by what actually happened during it. And done in a way that’s explainable to anyone who asks.
Machine Learned Attribution
One model, zero fuss
Other tools and platforms promise the world, offer you handfuls of models and expect you to pick the right one.
Ringside offers a single machine learned attribution model: the culmination of everything Russell has spoken about his whole career. And of course, the standard first and last click models are always there if you want them.
The model is trained on your own data rather than on somebody else’s benchmark, and retrained every week as your business changes. It works across the whole journey with no lookback limit and no platform bias, so a sale six months after the first click still credits the visit that started it.
- Machine learned model based on actual user events.
- First and last click models still available.
- Understand the (traditionally undervalued) role of organic search.
- Correctly attribute the value of affiliates.
- No black boxes. We’ll explain whatever you want to know.
Showing our working
You can see how the model arrived at the numbers it gives you.
Credit comes from odds, not raw probability. That lets the model distinguish between two strong visits in the same journey, rather than flattening both as they approach certainty.
Time is part of the model. A visit three days before a sale is treated differently from one three months earlier, and you can see the effect that has on the result.
Every sale still adds up to one sale. Credit is normalised across the journey, so the total always reconciles back to the conversion that actually happened.
The coefficients are visible. The weights the model has learned for your data are written to a table. If you want to know why a visit received a particular amount of credit, we can trace it back through the model.
Platform numbers, alongside ours
There’s a page in the dashboard showing exactly what Google, Meta and the other platforms report: spend, clicks and conversions, pulled straight from their APIs and left untouched.
Those numbers sit alongside ours. The difference is simple: platforms report the conversions they influenced according to their own rules. Ringside counts what your business actually sold and attributes it across the journey.
Having both in one place means you can see where they agree, where they don’t, and why.
All powered by what happens in your business
The Ringside machine learning model integrates your marketing data and analytics data with the on-site events captured by the Ringside tracker on your website. These events are the primary factor in the model’s understanding of your marketing’s impact. On a live client model, the weighting breaks down roughly like the boxes here.
Customer history sits at zero on purpose. Existing customers are more likely to buy again, but that isn’t something a marketing channel should get credit for. We want the model to value what happened in the journey, not who the customer already was.
Understand the value of every visit in the journey
Last click, for the last time.
Ringside’s machine learned attribution model assesses the impact of each visit, showing how much influence each interaction had on the outcome. This approach ensures a precise evaluation of the value of investment across all visits and channels.
For instance, look at this fairly typical e-commerce journey taking place across five visits, with a voucher code affiliate at the end:
2026-04-01Paid Search
Keyword: “women’s sneakers”
Cost: £1.31
2026-04-03Organic Search
Landing page: /womens/sneakers/
2026-04-07Paid Social
Landing page: /selected/seasonal
Cost: £0.44
Bounced visit
2026-04-11Organic Search
Landing page: /pdp/run-fast-sneaker
2026-04-11Affiliate
Name: Honey Commission
Cost: £4.25
journey
We can then see how this compares in a traditional last-click model, versus the distribution of the sale in a Ringside machine learned model:
attributed view.A Ringside attributed view. Notice the conversions and revenue distributed throughout the customer’s journey, based on the impact of their actions during the visit.
| Visit #1 | Visit #2 | Visit #3 | Visit #4 | Visit #5 | |
|---|---|---|---|---|---|
| Conversions | 0 | 0 | 0 | 0 | 1 |
| Revenue | 0 | 0 | 0 | 0 | £76.50 |
| Media Cost | £1.31 | 0 | £0.44 | 0 | £4.25 |
| ROAS | 0% | No media cost | 0% | No media cost | 1800% |
| RSD Conversions | 0.2 | 0.15 | 0.05 | 0.4 | 0.2 |
| RSD Revenue | £15.30 | £11.50 | £3.80 | £30.60 | £15.30 |
| RSD ROAS | 1167% | No media cost | 863% | No media cost | 360% |
Visit #4 is the clearest example. On a last-click view it produced nothing. It’s the visit where the customer read the product page they went on to buy from, and it carries more attributed revenue than any other visit in the journey.