Attribution Gap Calculator: Meta vs. Amazon Attribution

When you run Meta ads for a brand that sells on Amazon, you get at least two answers to "what did my ads make?" Meta Ads Manager reports one number. Amazon Attribution reports another. They rarely agree, and neither is the number that matters: the sales your ads caused, which you'd lose if you stopped. That number is called incremental revenue.

The attribution gap is the distance between what a report credits your ads with and what they actually caused. A positive gap means the report over-credits your ads, so they look better than they are. A negative gap means it under-credits them, so you might cut ads that work.

Meta counts purchases its pixel or Conversions API can see, within its attribution window. For website-conversion campaigns, the default since March 2026 is a click in the last 7 days, or a view or engagement in the last day. It counts people who would have bought anyway. And it can't see purchases on Amazon, because Amazon doesn't send them back to Meta.

Amazon Attribution counts Amazon purchases made within 14 days of a click on a tagged link. It misses everyone who saw your ad, didn't click, and searched for you on Amazon later. For many brands that's a big share of what Meta ads do on Amazon.

Both are rules for handing out credit. Neither can tell you whether a shopper would have bought without the ad. For how to set up tags and where to send the traffic, see Facebook ads for Amazon.

ROAS for each source     = revenue from that source ÷ ad spend
gap vs. incremental      = (source revenue − incremental revenue) ÷ incremental revenue
Meta vs. Amazon (no test) = Meta-reported revenue ÷ Amazon Attribution revenue − 1

• Ad spend: what you spent on the campaigns over the period, from Meta Ads Manager.
• Meta-reported revenue: the purchase conversion value Meta credits to those campaigns.
• Amazon Attribution revenue: product sales Amazon credits to the same campaigns' tagged links.
• Incremental revenue: extra revenue the ads caused, measured with a holdout test over the same period. It's optional; without it, the calculator compares Meta and Amazon Attribution only.

Use the same dates for every number, or the gap measures the calendar instead of the reports.

A made-up brand spends $5,000 on Meta in a month. Some ads link to its own site, some to Amazon with Attribution tags. These are the example values in the calculator.

1. Meta reports $14,000 of purchases: a ROAS of $14,000 ÷ $5,000 = 2.80.
2. Amazon Attribution credits $3,500: a ROAS of 0.70.
3. A holdout test measured $8,000 of extra revenue across both channels: an incremental ROAS of 1.60.
4. Meta's gap: ($14,000 − $8,000) ÷ $8,000 = 75%. Meta over-credits the ads.
5. Amazon Attribution's gap: ($3,500 − $8,000) ÷ $8,000 = -56.3%. Amazon under-credits them.

Judged by Meta alone, the brand would think its ads return 2.80. Judged by Amazon Attribution alone, 0.70, and it might switch them off. The real answer, 1.60, is in between. Whether that's profitable depends on the brand's break-even ROAS.

There's no good gap, only a known one. Here's how I'd read the size of it:

• Within about 10–20%: the report is a usable stand-in for day-to-day decisions, as long as nothing big changes.
• 20–50%: use the report for direction only. Correct it with your measured ratio before deciding budgets.
• Over 50%: decisions based on the report alone are likely wrong, in whichever direction the gap points.

These are rules of thumb, not benchmarks. Gaps vary by category, price, how often shoppers search Amazon after seeing an ad, and how much of your traffic is tagged. They also move when your ads, prices or season change, so a gap measured once won't hold forever.

• Adding Meta and Amazon Attribution together. The sum still counts credit, not cause. It can be bigger or smaller than what the ads caused.
• Mismatched dates or windows. Meta's window is up to 7 days, Amazon Attribution's is 14. Pull every number for the same period, and leave a gap after the period ends for late purchases.
• Untagged links. Amazon Attribution only sees clicks on its own tags. Any Meta ad linking to Amazon without one is invisible to it.
• Treating a gap as fixed. If spend doubles, the gap can change. Re-measure after big changes.

Incremental revenue is the only one of the three numbers that answers "what would I lose if I stopped?" The cleanest way to measure it is a holdout test: pause Meta ads in some regions, keep them on in matched regions, and compare Amazon sales before and during. That catches shoppers no click ever tracked.

Size a test so it can detect the effect you care about with the holdout test calculator, then turn its results into incremental revenue with the incremental ROAS calculator. Put that number in the calculator above and keep the ratio. Each test makes the next estimate better. See how lift is calculated for the full method.