Incrementality vs. Attribution: Why Last-Click Is Lying to You

You ran a retail media campaign. Sales went up 18%. Your agency sent a deck with a ROAS of 6.2. Everyone in the room felt good about it.

Now ask the harder question: would those sales have happened without the campaign?

If your measurement methodology can't answer that question, you don't know whether your retail media program is working. You know that sales happened while the campaign ran. That's not the same thing.

The gap between those two answers — what actually happened vs. what the campaign caused — is the gap between attribution and incrementality. And it's where most retail media budgets quietly leak.

‍

"Last-click ROAS can overstate retail media performance by 3x or more in high-frequency CPG categories." — Forrester / Skai analysis

‍

What attribution actually measures

Attribution is an accounting system. It answers: which touchpoints were present before a conversion happened, and how do we divide credit among them? Last-click attribution — the default for most retail media network reporting — gives 100% of the credit to the final ad impression before purchase.

The problem isn't the math. The math is correct. The problem is what it's measuring. A shopper who buys your cereal every two weeks, every single week of their adult life, who happened to see your sponsored product ad last Tuesday — that's not a conversion your ad drove. That's a conversion your ad witnessed. Last-click attribution can't tell the difference.

In low-frequency categories with long purchase cycles, this distortion is manageable. In high-frequency CPG categories — snacks, beverages, household consumables, personal care — where baseline purchase rates can exceed 30-40% among your target audience, last-click attribution turns every naturally occurring repeat purchase into apparent ad ROI. The numbers look great. The budget is buying credit for purchases it didn't cause.

What incrementality actually measures

Incrementality doesn't ask "what was near the conversion?" It asks "what caused the conversion?" The methodology is simple in principle: compare a group of shoppers who saw your ad against a matched group who didn't. The difference in purchase rate between those two groups is your incremental lift — the sales your ad actually drove.

That number is almost always smaller than your last-click attributed sales figure. Sometimes meaningfully smaller. The brands that have done honest incrementality testing and compared it to their attribution reports often discover that their actual retail media ROI is 40-60% of what the dashboard was showing.

This sounds like bad news. It's actually good news — because now you have a real number to optimize against. You can identify which campaigns have genuine incremental lift and double down on them. You can identify which campaigns are buying credit and cut them. That's how budgets get smarter.

‍

→ Retail Media Measurement  —  agencyfiveeighty.com/retail-media-incrementality

→ Data & Analytics  —  agencyfiveeighty.com/data-analytics

‍

Why most brands don't measure incrementality

It's harder. Incrementality testing requires planning before the campaign launches, not after. You need a holdout group — shoppers who match your target audience but don't receive the campaign — and you need that holdout to be large enough for statistical significance. That means asking the network to withhold inventory from a portion of your audience, which networks are understandably reluctant to do without a push.

It also requires patience. You need the campaign to run long enough for the purchase rate difference between exposed and control groups to become statistically significant. For low-frequency categories with monthly purchase cycles, that can mean running a test for eight to twelve weeks before you have a clean read.

Most brand teams are optimizing in-flight, making weekly decisions based on whatever the dashboard shows. That cadence isn't compatible with rigorous incrementality testing. The result: brands default to attribution because it's available in real time, even though they know it overstates performance.

The practical middle ground

You don't have to choose between perfect incrementality measurement and no measurement. There's a practical middle ground that most brands should be running.

• For your largest campaigns: request holdout methodology from the network before the campaign launches. Walmart Connect and Kroger Precision Marketing both support this; smaller networks are negotiable.
• For ongoing campaigns: compare your attributed sales to your organic baseline rate in the same category, same time period. If your baseline purchase rate is 35% and your attributed conversion rate is 37%, the lift is 2 percentage points — not 37%.
• At minimum: track your spend-per-incremental-sale over time, not just ROAS. As you add more spend to a campaign, incremental lift typically flattens while attributed sales keep growing. That divergence is the canary.

Five Eighty builds measurement frameworks before campaigns launch — not after. Because the question "did this actually work?" should have an answer before you agree to the budget.

‍