I Knew What Their ROAS Would Say Before They Ran a Single Ad

A few weeks ago I sat on a call to set up measurement for a retargeting partner that hadn't launched yet. Standard kickoff — how the data would flow, what we'd need from each side, when the model would have enough signal to score them fairly.

Somewhere in the middle, they mentioned their reporting setup almost in passing. Thirty days post-click. A short post-view window on top of it. Nobody reacted, because it's a completely normal way for a retargeting partner to report.

But I already knew two other things about this advertiser. Most of their revenue was organic — the business would keep selling at scale with every campaign switched off. And their customers took a couple of weeks, on average, between first interest and purchase.

Put those three facts together and you can forecast the vendor's dashboard before it exists. Their reported return was going to look spectacular. Their actual contribution was going to be a much smaller and much more interesting number. And I could say that before a single impression served.

A thirty-day post-click window means that if someone clicks a retargeting ad and buys any time in the following month, that sale lands in the vendor's column. When a business converts in roughly two weeks, thirty days doesn't capture the influence of the ad. It captures nearly every purchase that person was already going to make.

Now layer on the organic base. Retargeting, by construction, shows ads to people who already visited the site. These are the most likely buyers the business has. A large share of them were coming back regardless — that's what a high organic base means. The vendor is standing in the doorway of a store people had already decided to walk into, handing out flyers, and counting everyone who walks in.

None of this requires anyone to behave badly. The vendor didn't pick that window to inflate anything; it's their standard configuration, and they'd tell you so honestly. The number is real. It's just measuring proximity to the purchase rather than influence over it.

Here's where most marketers get this backwards, and it's the reason I wanted to write this one.

When you switch from platform-reported return to incremental return, the numbers get smaller. Sometimes dramatically. And the instinctive reaction is that the channel just got worse — that a partner reporting a big multiple who turns out to be delivering a modest one has been exposed as a failure.

That reaction is wrong, and it costs people money.

Incremental return above break-even is profit. Full stop. A partner delivering a modest incremental multiple is genuinely growing the business. It's not a disappointing version of the big reported number — it's a completely different measurement, on a different denominator, answering a different question. Comparing them is like being disappointed that your net income is lower than your revenue.

I've watched teams cut partners who were making them money, because the honest number looked small next to the dishonest one they'd gotten used to. And I've watched the reverse — teams protecting a partner reporting an enormous multiple that was, on inspection, almost entirely claiming credit for demand somebody else created.

The reported number tells you where the sale was observed. The incremental number tells you whether the sale happened because of the spend. Only one of those is a business decision.

Here's the thing you can act on this week, and it takes one email.

Ask every media partner what attribution window they report on. Post-click and post-view, separately. Most will tell you immediately — it isn't a secret, and the good ones are slightly relieved someone finally asked.

Then put those windows next to how long your customers actually take to buy. If a partner's window is meaningfully longer than your typical purchase cycle, their reported performance includes a lot of activity that was going to happen anyway. That doesn't make them a bad partner. It makes their number uncomparable to another partner reporting on a tighter window — and if you're allocating budget by ranking those numbers against each other, you're ranking the windows, not the performance.

The follow-up question is the one that really tells you something: ask whether they'd be willing to report on a shorter window alongside their standard one.

Watch what happens. Partners confident in their contribution generally shrug and say sure. Partners whose case depends on the long window will explain why the shorter one wouldn't be fair to them. Both responses are informative, and neither requires you to accuse anyone of anything.

You don't need a model to start. You need three numbers you probably already have: roughly what share of your revenue happens without paid media, roughly how long your customers take to buy, and the reporting window each partner uses.

With those three, you can rank your partners by how much of their reported performance is likely to be capture rather than creation — before you spend anything on measurement, before you have a fight with anyone, before the next planning cycle forces a decision.

You will not get a precise answer that way. You will get something more useful in the short run, which is a much better sense of which of your partners' numbers deserve to be believed.

A partner's reported return tells you how they count. Only incrementality tells you what they contributed.

Talk soon,
Jeff Greenfield
CEO Provalytics

Jeff is the co-founder and CEO of Provalytics. He'll mention it occasionally. He'll also tell you when it's not the right tool.