- Why Amazon wants to change how you measure its campaigns
- The difference between a projection and real data
- The problem with data: Amazon provides little, and the right data arrives late
- What changes if you also sell outside of Amazon
- What to check before moving budget to Amazon
- Frequently Asked Questions
- Amazon states that 20% of sales generated by its campaigns occur after the standard measurement window (7-30 days) and are not counted.
- The platform is developing 12-month metrics to demonstrate longer-term value, but many agencies still treat them as estimates, not certain data.
- The historical data Amazon provides to verify these numbers is often insufficient: 90 days instead of the 18 months needed for a reliable analysis.
- Before increasing your budget on Amazon, ask for independent data and compare the platform's projections with actual week-over-week sales.
Amazon claims that about a fifth of the value generated by your ad campaigns doesn't show up in the reports you read every week. Before shifting budget based on this claim, it's worth understanding what's behind it — and what to ask.
Why Amazon wants to change how you measure its campaigns
Each advertising platform measures campaign results within a time window: usually 7 or 30 days from the last click or ad view. Anything that happens after that date is not counted. Amazon claims this system penalizes it, because a significant portion of purchases happen later.
According to the platform's internal data, derived from the analysis of over 2 million campaigns, about 20% of the value generated by a campaign comes after the measurement window closes. That's why Amazon is building a 12-month metric that estimates future sales based on new customer behavior: if a user saw an ad and then searched for the brand or added a product to their cart, the platform uses historical data to project how much they will spend over the year. You can read the original analysis on Digiday .
The practical takeaway for you: if Amazon shows you better numbers thanks to these new metrics, it doesn't necessarily mean your actual sales have increased. It could simply mean the counting window has widened.
The difference between a projection and real data
Amazon's 12-month metric is an estimate. Here's how it works: the platform looks at new customers who interacted with your ads, then uses the historical behavior of similar customers to calculate how much they'll spend in the next year. It's not a recorded sale, it's a forecast.
Many agencies working with Amazon treat these projections with caution. The concrete problem is that an abandoned cart or a brand search does not guarantee a purchase. Some of those customers will return, others will not. When the more skeptical agencies check the numbers week by week, they often find no correlation with actual sales.
This doesn't mean the metric is absolutely wrong. It means it needs to be verified. If you run campaigns on Amazon, the article about advertising attribution and the models that deceive the budget helps you understand exactly where these systems can lead you astray.
The problem with data: Amazon provides little, and the right data arrives late
To independently verify if Amazon campaigns are truly generating value, agencies build their own models — tools that cross-reference ad spend with actual sales, both online and in-store. These models need about 18 months of historical data to work well.
Amazon provides 90 days. That's less than a fifth of what's needed. Access to longer data exists, but only through a beta program and not available to everyone. This creates an imbalance: Amazon asks you to trust its projections, but doesn't give you enough data to verify them yourself.
For those selling on Amazon, this is the central point. If you're considering moving significant budgets to the platform — perhaps after reading about new features like those described in the article about Amazon Ads AI multi-format — the first question to ask isn't "how much can I spend?" but "what data can I use to verify the results?"
For us, technology is useful when it gives you control, not when it takes it away. Relying on a platform's projections without being able to compare them with independent data means optimizing blindly.
What changes if you also sell outside of Amazon
Amazon is developing a tool called "omnichannel metrics" — multichannel metrics — that tries to measure the sales generated by its ads even outside the platform: in physical stores or on other websites. In one study cited by the source, video campaigns streamed on Amazon reportedly generated more than 2 euros in additional sales off-platform for every euro spent.
This data is interesting, but needs context. It's a single study, conducted with a partner selected by Amazon. For a company that sells both online and in-store — think of a consumer goods manufacturer with national distribution — the risk is attributing sales to Amazon that would have happened anyway, or that depend on other factors.
The theme of multi-channel attribution is the same one that emerges when talking about How YouTube counts views or how platforms redefine metrics to their advantage. The logic is always the same: the more value the platform can demonstrate, the more budget it attracts. Keep this in mind when reading reports.
You can learn more about other developments related to the Amazon ecosystem in the hub dedicated to Marketing and ADV performance .
What to check before moving budget to Amazon
- Ask Amazon what historical data you can export and for how many months: if it's less than 12, long-term projections are not verifiable.
- Compare actual week-over-week sales with the platform's projections: if they don't match, don't increase the budget before understanding why.
- Check if the new metrics (like the 12-month estimate) are active on your campaigns and how they affect the reports you receive: a higher number isn't always good news.
- If you also sell outside of Amazon, build an independent measurement system that cross-references advertising spend with actual sales across all channels.
- Before accepting
- Consider relying on an external measurement partner, not suggested by Amazon, to get a second opinion on campaign results.
Frequently Asked Questions
Not necessarily. Amazon is expanding the measurement window to include estimated sales in the 12 months following the campaign. These are projections based on historical behavior, not sales that have already occurred. Before increasing your budget, check if these estimates align with your actual sales.
The attribution window is the period within which a sale is linked to an ad. If a customer sees your ad today but buys 40 days later, with a 30-day window, that sale won't be counted. Amazon wants to extend this window to 12 months, which would make campaigns appear more effective — but only if the estimated sales actually happen.
You can build a measurement model that crosses ad spend with actual sales, week by week, across all channels. To do this, you need at least 12-18 months of historical data. Amazon only provides 90 days by default, so you may need to request access to additional data or rely on an external measurement partner.
Amazon is also promoting them to smaller advertisers, precisely because independent measurement models are expensive and complex for SMEs to manage. However, this doesn't mean the projections are more reliable: it means SMEs have fewer tools to verify them, and therefore need to be even more cautious.
A metric you can't verify isn't data: it's a promise.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
