Amazon Ads: about 20% of brand sales land after the attribution window

Amazon Ads estimates that roughly 20% of a brand's sales value on Amazon arrives after the standard attribution window closes, Digiday reported on 1 October 2026. Amazon reached the figure by studying more than 2 million campaigns that ran on its ad products in June, and it is using the result to argue that advertisers should judge campaigns on a full 12 months of sales rather than the usual 7 to 30 days.

What Amazon Ads found

Attribution windows on Amazon typically run from seven to 30 days, according to Digiday. Sales that happen after that cutoff are not credited to the ad in standard reports. Amazon's study of more than 2 million June campaigns put the size of that uncredited share at about a fifth of a brand's value.

Lily Tong, a director of measurement at Amazon Ads, told Digiday that "roughly 20% of the sales value for a brand shows up over time and isn't captured in that immediate attribution window because that's just a cutoff date for measurement." She added: "It doesn't mean that you didn't create value."

The study is Amazon's own analysis of its own ad products. Digiday did not report an independent audit of the figure, and the article did not break it down by category, region or ad format.

How the long-term sales metric works

The 20% figure rests on a long-term sales metric that Amazon unveiled in closed beta in October 2024. Digiday describes how it is built:

1. It looks at shoppers who are new to a brand, meaning they have not bought from it in the past year.
2. It records their first key actions after seeing an ad, such as searching for the brand or adding one of its products to their cart.
3. It applies the brand's historical 12-month return from shoppers who took the same actions, and uses that to estimate what the campaign will generate over the next year.

So the metric is a projection. It does not wait 12 months to count actual purchases. It assumes that new shoppers who search or add to cart today will behave the way similar shoppers behaved in the brand's past. Tong acknowledged the gap to Digiday: not every shopper who searches or fills a cart will buy, though some will, and shoppers who add a product to their cart and leave often come back and buy it within 12 months.

A check is coming, Digiday reports: marketers will "soon" be able to see whether projected long-term sales came true and compare them with actual sales. The article did not give a date for that feature.

The key figures at a glance

The table below collects the numbers reported by Digiday on 1 October 2026.

Why Amazon is pushing a longer view

Digiday frames the push as part of a wider fight over credit for sales. Better upper-funnel measurement and AI-driven shopping are pulling money out of search, and TikTok, Instagram, YouTube and creators claim credit for sales they helped drive, even when the purchase happens on Amazon. Proving value over a longer stretch is becoming table stakes, and Digiday notes Amazon, like Google, needs it more than most.

The metric fits a run of Amazon moves toward longer-term selling. Digiday lists full-funnel campaigns, announced in November 2025, which try to shift budgets across Sponsored Products, Sponsored Brands, display and streaming to drive long-term sales. This spring Amazon also said it would extend the lookback windows in its clean room.

Tong's argument is aimed most at brand advertisers. She told Digiday they "can't put a dollar value on" what their brand ads do, so they fall back on brand lift surveys or proxy signals such as search. That makes brand spend hard to weigh against performance spend, which she called "very apples and oranges." A dollar figure would let a marketer set brand spend on Amazon against the rest of the performance plan.

Amazon is also working on sales outside its store. Omnichannel metrics, a tool built with industry partners, takes attributed Amazon ad exposures and looks at the sales they drive elsewhere. It covers consumer packaged goods and now automotive. Tong wants those figures in standard end-of-campaign reports, and pitched them at smaller advertisers because marketing mix models tend to be "expensive and infrequent." Digiday also reports that one study with an offline sales partner found Amazon's streaming TV ads returned more than $2 in incremental sales off Amazon for every dollar spent. Benchmarks against direct peers and category leaders are at an earlier stage.

Where agencies push back

Digiday reports that agencies are split. Some say long-term sales helps brands look past last-touch attribution and see how ads move shoppers from discovery to purchase. Others treat it as a forecast that shifts whenever the campaign mix changes, and say the uplift Amazon credits to streaming TV does not show up when they check it against actual product sales week by week.

The skeptics tend to build their own marketing mix models to tie Amazon spend to direct-to-consumer and in-store revenue. That is where data access bites. Amazon gives agencies 90 days of historical sales data, while a mix model needs closer to 18 months, though Digiday says access improved this year through a beta program. Amazon also prefers analysis to run in its clean room, where advertisers take out aggregated insights; Tong put that down to a "really high bar on customer privacy."

Agency executives quoted by Digiday made the same point from the buy side. One said full visibility is possible when the right data exists, but that this is a big "if" depending on where the data comes from. Another said that so far they had seen off-Amazon spend appear in reports across their campaigns "with no sales behind it," and that they were keeping that functionality to a minimum. Tong's answer was that advertisers can bring in almost any measurement firm they like: "We don't point advertisers to our measurement."

What the article did not say

• No launch date for the feature that compares projected long-term sales with actual sales.
• Nothing on whether the long-term sales metric is available outside the US or in which marketplaces.
• No breakdown of the 20% by product category, price point or ad type.
• No independent validation of the 2 million-campaign study.

How to read long-term sales claims

This section is analysis. A projected 12-month figure and a measured 12-month figure are different things, and reports should keep them apart. A few practical points follow from what Digiday described:

• Treat the long-term sales number as a model output. It depends on the brand's history and on how many new shoppers searched or added to cart, so it will move when the campaign mix moves, as the skeptics noted.
• Keep last-touch ROAS and long-term projections in separate columns. Adding them together double-counts nothing on paper, but it mixes a count with a forecast.
• When the verification feature arrives, compare projections against actual repeat purchases before shifting budget toward upper-funnel formats on the strength of the projection.
• If you run a mix model, the 90-day data limit matters. Building your own sales history outside the platform, in your store analytics and order data, is what lets you test a platform's claims over 18 months.

The core idea, that a short attribution window undercounts slower purchases, applies to any platform. The specific 20% figure applies to Amazon's own study and should not be carried over to other channels.

What this means for Thai marketers

The Digiday article contains no Thai data and does not say whether Amazon's metric is offered in Southeast Asia, so this section is analysis. Thai brands mostly sell through marketplaces such as Shopee and Lazada and through their own stores, and many judge marketplace and social ads on short attribution windows. Amazon's study is a reminder that some sales land after the window closes. It is not evidence of how large that share is on other platforms.

The practical step is to own the longer data. Brands that keep order-level history and clean GA4 measurement on their own sites can check repeat purchases and new-customer behaviour over 12 months themselves, instead of relying on whatever window a platform reports. For teams planning ecommerce marketing across marketplaces and direct stores, separating first-purchase revenue from later repeat revenue gives a fairer view of upper-funnel spend.

Frequently asked questions

What did Amazon Ads find about sales after the attribution window?

Amazon Ads estimates that about 20% of a brand's sales value on Amazon arrives after the 7 to 30 day attribution window closes. Digiday reported the figure on 1 October 2026, based on Amazon's study of more than 2 million campaigns run in June.

Is the long-term sales metric a measurement or a forecast?

It is a forecast. The metric tracks new-to-brand shoppers' first actions after an ad, such as branded search or add-to-cart, and projects 12 months of value from the brand's historical return on similar shoppers.

Can advertisers check whether the projection came true?

Not yet, according to Digiday, which said marketers will "soon" be able to compare projected long-term sales with actual sales. No date was given.

Is this available in Thailand?

The source did not say. Digiday did not mention which markets have access to the long-term sales metric.

Why are some agencies skeptical?

Some agencies say the forecast shifts whenever the campaign mix changes and that credited streaming TV uplift does not appear in weekly product sales checks. Digiday notes Amazon shares 90 days of historical sales data, while a marketing mix model needs about 18 months.

Relevant Audience helps brands in Thailand set up measurement that looks past short attribution windows, across marketplaces and owned stores. If you want to know how much of your sales arrive late, talk to our ecommerce team.