To separate incremental ad sales from sales you would have had anyway, stop trusting attributed sales and measure total sales with the ads on and off. Pause or cut ads on one SKU for a fixed window, compare its total sales to the period before, adjust for whatever moved in a control group of similar SKUs you left alone, and price the difference at contribution margin. What remains after subtracting ad spend is what the ads earned. On branded and best-selling products, that figure is often a fraction of what the ad console reports, and sometimes it is negative.
Amazon's own definitions explain the gap. Its ad campaign attribution page on advertising.amazon.com, updated October 20, 2026, states that for sellers, Sponsored Products uses a 7-day lookback window, counts clicks only, and assigns credit on a last-click basis. Eligible products include the advertised product and any other product sold by the same Seller Central merchant.
Read that carefully. A shopper who typed your brand name, saw your ad sitting above your own organic listing, and clicked the ad gets counted as an ad sale. So does a shopper who clicked an ad on Monday, left, and came back through your organic listing on Saturday to buy a different product of yours. Neither purchase needed the ad. Both are attributed to it.
The same page notes that attributed sales do not reflect discounts applied in the cart and are reported on the date of the ad click, not the date of the sale. Attributed revenue therefore sits on a different basis from what your settlement records, which is one more reason not to compare it directly to your books.
Attribution answers "which sales followed an ad click." Incrementality answers "which sales would not have happened without the ad." Profit depends on the second.
The cleanest test a marketplace seller can run without special tooling is a pause test on one SKU at a time.
The control matters because two consecutive weeks are never identical. Seasonality, a competitor's stockout, or a payday weekend move every product. If the control group fell 4 percent, part of the test SKU's drop was going to happen anyway.
Illustrative numbers. Take a seller with a kitchen scale on Amazon. Its contribution margin before ad spend is 34 percent of sales, after referral fees, fulfillment fees, refunds, and FIFO landed cost.
Weeks 1 and 2, ads running:
At those numbers the campaign looks profitable. $9,800 of attributed sales at a 34 percent margin is $3,332 of contribution, against $2,600 of spend, a gain of $732.
Weeks 3 and 4, ads paused:
The adjustment. If the control group fell 4.0 percent, the test SKU would have been expected to fall about the same with ads still running. Expected sales with ads in weeks 3 and 4: $18,400 times 0.96 equals $17,664.
Incremental sales: $17,664 minus $14,900 equals $2,764. That is 28 percent of the $9,800 the console attributed. The other 72 percent of attributed sales happened anyway, through organic search or a buyer who would have found the listing regardless.
Incremental contribution: $2,764 times 34 percent equals $939.76.
Net result of the ads: $939.76 minus $2,600 equals negative $1,660.24 over two weeks.
To break even, the ads would have needed to produce $2,600 divided by 0.34, or $7,647 of incremental sales. They produced about a third of that. The campaign that reported a 3.77 ROAS returned $1.06 of incremental sales per dollar spent.
Two weeks of pausing measures the short-run effect. It can understate what ads do if ad sales feed organic rank and rank decays over a longer window. It can also overstate it if a competitor happened to raise prices during your test. Run it twice, on different weeks, before you cut a campaign for good. Look at total units ordered in Business Reports along with revenue, because a price that moved mid-test contaminates the revenue line.
Branded search terms and best sellers tend to show the biggest gap between attributed and incremental. Launch products and non-branded discovery terms often show the smallest, because without the ad, the shopper never sees the listing.
TACoS, ad spend as a share of total sales, is the first place the gap shows. If ad spend rises and total sales do not, TACoS climbs while ACoS stays flat. The two ratios and what each hides are in TACoS vs ACoS.
The test itself only works if ad spend is booked against the SKU it promoted. A single monthly Amazon Ads invoice posted to one marketing account cannot support a pause test, because you cannot see that the kitchen scale's spend went to zero. ConnectBooks reconciles Amazon settlements and carries a PPC ROI view alongside SKU-level profit, so ad spend and sales can be read together for the same period.
The comparison above is the kind of question Crunch, the AI CFO inside ConnectBooks, takes in its ad spend deep-dive area. A seller can ask "what did pausing ads on the kitchen scale do to its profit" and Crunch works through its five steps on the seller's reconciled ConnectBooks data: what changed in sales, ad spend, and margin; which periods to compare; which products moved; why; and what to do next. The question patterns for broader campaign review are in PPC analysis with AI.
It cannot run the experiment. Crunch compares periods you already have. If ads never paused on a SKU, there is no "ads off" period to compare, and any incrementality figure becomes an estimate built on assumptions you should see stated.
It does not pick the control group with your knowledge. You know that one "similar" SKU had a stockout in week 3 or that another got a new main image. Choose the controls yourself, or check the ones it uses.
It does not see the long tail. Organic rank effects, repeat purchases, and review velocity play out over months. A two-week comparison misses them, whoever runs it.
It does not decide your strategy. A launch SKU might deserve unprofitable ads for a quarter while it builds reviews. That is a business call, and the test only tells you its price.
Sort your ad spend by SKU. Run the pause test first on the three products with the highest spend and the most organic sales already, because that is where attributed sales most overstate the effect. Keep the spend where incremental contribution exceeds cost. Cut or restructure where it does not. Then move the freed budget to the products where the test showed ads finding buyers who would not have found you otherwise.
Running an e-commerce business comes with plenty of challenges, but ConnectBooks is here to make your life easier. With real-time insights, seamless integrations, and detailed tracking of your profitability and inventory, you can stay ahead of the game. Whether you’re selling on Amazon, Shopify, Walmart, TikTok or eBay, ConnectBooks helps you manage your finances with 100% accuracy and confidence, so you can focus on growing your business.
Ready to level up? Start making smarter, data-driven decisions every step of the way. Try ConnectBooks Free Today or Schedule a Demo