Alin Vlad (E-commerce Director, Otter): "Anyone who blames the network for overlap is also blaming maths"
The text below was written by Alin Vlad (E-commerce Director, Otter). With his permission, we have translated it (with AI) for our blog. E-commerce is a discipline that must account for many factors, and the role of affiliate marketing in this mix can bring significant benefits to online stores.
In April 2025, BusinessLeague by 2Performant reported 5,254 sales and RON 1.23M to me. GA4, on last non-direct click, credited me with 1,484 sales and RON 468K. The difference: ~3,770 orders. And no, nobody is... lying.
On those orders, the affiliate was somewhere in the customer journey, but the last meaningful touch came through a different channel. The network counts based on its own cookie, last-affiliate-click, with a 30–90 day window. It doesn't see that someone else closed the sale. It can't.
This isn't a flaw in any particular platform. Wherever you have 5–6 channels each placing their own cookie and claiming the same conversion, their attribution models will overlap. Guaranteed. Anyone who blames the network for overlap is also blaming mathematics.
I understand why this sparks debate. You're paying twice: the affiliate commission, plus the channel your model considers the last touch.
I know people who removed the channel entirely because of this. I didn't. I ran the numbers using the strictest counting method available.
On LNDC, where affiliate gets credit only where it was the last meaningful touch, the channel delivered RON 81K in spend, RON 468K in revenue, 41% net margin, and ~RON 21K in profit after COGS and commission. It pays for itself under the worst-case scenario. Everything it contributes in an assist position is a bonus.
What LNDC still doesn't tell me: how many of those sales would have happened anyway. Incrementality doesn't show up in any platform report — it can only be measured through a holdout test. I haven't done one yet.
Until then, I work with what I can control: my own tracking, clean, with data in one place. A single attribution model, applied identically across all channels. And decisions based on POAS and operating margin — not on what each platform reports about itself.
The decision changes if POAS drops below 1 or if the commission exceeds the operating margin. Until that point, what I need to do is clean up the program internally — not pause the channel. If you look at each platform's report in isolation, you'll always see inflated numbers. Without unified tracking, it's just noise.
A few notes on terminology I kept consistent with your usual style: LNDC (last non-direct click), POAS (profit on ad spend), COGS — all left as-is since they're standard industry acronyms your audience will recognize.