The number on the dashboard
Every retail media network reports some version of sales back to the advertiser. It might be labeled sales, attributed sales, online and in-store sales, halo sales or assisted sales, but it all measures the same thing: sales that followed someone seeing or clicking your ad. Divide by spend and you get ROAS.
For running campaigns day-to-day, it's a perfectly useful number. The trouble starts when it gets pulled into a bigger conversation, usually the one with finance, where the question is simple. If we hadn't spent this money, how much less would we have sold?
Attributed sales can't answer that. Incremental sales can, and the gap between the two is almost always bigger than teams expect. It also doesn't fall evenly across your campaigns, which is where it starts costing money.
What attributed sales actually counts
Attribution works by matching ad exposure to purchases inside a window, often a week or two. If a shopper clicked your ad and bought within that window, the sale counts. The rule never asks why they bought.
That's where the overcount comes from. Plenty of those shoppers were already on their way to checkout. They buy your brand every month. They searched for it by name. It was sitting in their cart from last week. The ad happened to be the last thing they touched, so it gets the credit for a sale that was going to happen regardless.
There's an undercount running the other direction, too. Someone who sees a display ad on Tuesday and buys in-store on Saturday usually doesn't show up at all. Neither does the shopper who sees your ad on one retailer and buys on another. Each network sees its own ads, its own site and its own window. In our cross-retailer halo research, roughly half of offsite media's total impact went uncounted for exactly this reason.
So attributed sales is a record of what happened near your ads. What the ads caused is a separate question, and the two answers can be very far apart.
Where the overcount is worst: your own organic shelf
Branded search is the clearest case. A shopper types your brand name into the retailer's search bar. Your product would land in the first organic slot anyway. Your sponsored listing sits right above it, they click the ad, and you pay for a sale you already had.
On the dashboard these campaigns tend to look like the best in the account. High conversion, cheap clicks, a ROAS that makes everything else look weak. That's what makes them dangerous. Budget drifts toward them because the number says it should, and in a lot of cases what the brand is buying is the right to appear twice on a page where it already ranked.
Sponsored products on your own hero SKUs work the same way. Some of that spend is genuinely defensive, especially where a competitor is bidding on your name and would otherwise take the click. Whether it's worth paying for depends on one thing: would that shopper have left without the ad? Attributed sales can't tell you, in either direction. So you end up unable to separate the branded campaigns doing real work from the ones quietly billing you for organic demand.
What incremental sales measures instead
Incremental sales are the sales that wouldn't have happened without the ad. You're comparing what you sold against what you would have sold with that spend removed and everything else held equal.
That baseline is the hard part, and it's why incrementality needs more data than attribution does. You need total sales, online and in-store, across retailers. You need to account for everything that moves sales with no ad involved: price changes, promotions, seasonality, new distribution, a competitor going out of stock. Only then can you ask how much of what's left belongs to each campaign.
When brands get there, the two numbers tend to split in a pattern. Branded search and retargeting shrink, sometimes by a lot. Upper-funnel display, video and offsite media grow, because they were driving sales attribution never connected back to them. Our research has found upper-funnel retail media ROI to be understated by 3 to 5x when judged on attributed results alone.
Divide incremental sales by spend and you get incremental ROI, or iROI. That's the number that actually answers finance's question. (Note that we've also see this as referred to as “iROAS” in some places.)
ROAS and iROI don't move together
If attributed ROAS were a rough stand-in for the real thing, you could live with the gap. Rank your campaigns by ROAS and the order would roughly match the order by incremental return.
It doesn't. Repeatedly in Incremental's research, there's no correlation between ROAS and iROI. A campaign at the top of the ROAS list tells you very little about whether it's growing the business.
That's the practical problem with the overcount. Every budget decision made on ROAS, which campaigns get more and which get cut, is close to a coin flip from an incremental standpoint. Brands that make those calls on iROI instead tend to see big swings. Batiste improved iROI by 292%. TheraBreath improved it by 161%.
What to do about it
Keep using attributed sales for what it's good at. It's fast, it's free, and it's fine for in-flight calls like pausing a broken ad or catching a runaway bid.
Stop using it to split budget between campaigns, ad types or retailers. That's the decision where the overcount does the most damage, because it systematically rewards the campaigns closest to checkout.
Pull branded search and branded sponsored products out and look at them on their own. If they make up a big share of your attributed sales, assume a good chunk of that figure is organic demand you'd have captured anyway. Make the campaigns prove otherwise.
Then look at what attribution can't see. If you're running offsite or upper-funnel media and grading it on attributed ROAS, there's a real chance you're about to cut the campaigns doing the most work.
And when someone asks what retail media is returning, give them iROI. If you don't have it yet, say that plainly. A ROAS figure offered as the answer to that question is the one that comes back to bite you in next year's planning.
See what your campaigns are actually driving
Incremental is the causal intelligence layer for commerce media. We show brands which campaigns are driving sales across retailers, online and in-store, and help them move budget toward the ones that are. If you want to see how your branded campaigns hold up, talk to us or run your numbers through the iROI calculator.
