Key Takeaways
From 1 January 2027, Amazon plans to count a display impression on its store only when the ad appears on the shopper’s screen, not when it starts downloading, according to its notice to advertisers.
Amazon expects about 63% fewer reported impressions on the affected placements on average, but a smaller count is not a smaller audience, because the impressions that disappear are ads that loaded and never reached the screen.
If spend, clicks and purchases stay the same, CPM, click-through rate and Amazon DSP’s purchase rate (purchases per impression) all rise about 2.7 times, while spend, CPC, ROAS and cost per purchase do not move from the counting change alone.
Better-looking ratios are not better performance either, and Amazon’s claim that advertisers in testing saw higher ROAS and lower cost per action comes with no published figures or method.
Billing, fees charged per thousand impressions and frequency caps are the open questions that could change what you pay or deliver, so get written answers from Amazon before January.
Mark 1 January 2027 in every report, reset impression-based benchmarks afterwards and judge the change on spend, sales and cost per purchase.
It means a break in your impression reporting, not necessarily bad news: from 1 January 2027, Amazon plans to count a display impression on its store only when the ad appears on the shopper’s screen, and it expects reported impressions on those placements to fall by about 63% on average. In this article, impressions means the default impression count Amazon reports for DVA (Display, Video and Audio) display ads, the programmatic display ads advertisers know from Amazon DSP, on the Amazon store’s home page, product pages, search results and order pages. The details come from an Amazon notice to advertisers that the trade press reported in early November 2026, not from a public Amazon announcement, so the answers that matter most for cost, such as billing, are still missing. Below is what changes, what the 63% does and does not mean, how each metric moves, what could change your costs and how to reset your reporting.
The change at a glance
| Fact | Detail |
|---|---|
| What changes | A display impression on the Amazon store counts when the ad appears on the shopper’s screen, not when it starts downloading |
| Start date | 1 January 2027 |
| Ads covered | DVA display ads (Display, Video and Audio campaigns, the successor to Amazon DSP display) |
| Placements named | Home page, product detail pages, search results pages and order pages on the Amazon store |
| Expected effect | About 63% fewer reported impressions on average across those placements, more for placements lower on the page |
| What it is not | A cut in audience, spend or sales: the impressions that stop counting are ads that loaded but never reached the screen |
| Not named | Video, audio, ads on third-party sites and apps, Sponsored Products and Sponsored Brands |
| Reported unchanged | Viewable impressions, attribution and reports for dates before 1 January 2027, according to one trade report; the notice as reproduced does not mention them |
| Still open | Billing, fees charged per thousand impressions, frequency caps, reach, the exact on-screen threshold, countries and Amazon Marketing Cloud data |
| Source | An undated Amazon notice to advertisers, reported by the trade press; no public Amazon page as of 10 November 2026 |
What is Amazon changing about display impressions in 2027?
From 1 January 2027, Amazon will count a display impression on its store when the ad appears on the shopper’s screen, instead of when the ad starts downloading to the shopper’s device, according to its notice to advertisers. The notice calls the current method count-on-download: an impression registers as soon as the ad begins to load, even if the shopper never scrolls far enough to see it.
The notice names four placement types on the Amazon store: the home page, product detail pages, search results pages and order pages. It covers DVA display ads and does not mention video, audio, ads on third-party sites and apps, Sponsored Products or Sponsored Brands.
One overlap needs checking. Amazon now says Sponsored Display is known as display ads and lists existing Sponsored Display campaigns under Display, Video and Audio campaigns, but the notice does not say whether those campaigns are included. If you run them, ask your Amazon account manager before January.
The notice gives no list of countries and no time zone for the switch. Advertisers in the UK and 20 other countries still see the Amazon DSP console rather than DVA, so confirm which of your marketplaces and campaign types the change covers.
Why does Amazon expect 63% fewer impressions?
Amazon expects about 63% fewer reported impressions because, on average, roughly six in ten display ads that start downloading on these placements never appear on the shopper’s screen. The notice says placements lower on the page will see larger drops, which fits the mechanism: the further down an ad sits, the fewer shoppers scroll to it.
The 63% is an average across the affected placements, not a forecast for your account. According to the notice, the impact varies by campaign and placement, and it gives no sample, period, markets or weighting behind the figure.
A smaller count is not a smaller audience. An ad that loads below the fold and never reaches the screen had no opportunity to be seen, so dropping it from the count removes nothing a shopper saw. Reported reach can still fall if Amazon recalculates it on the new count, because shoppers whose only impressions never reached the screen would drop out of it.
The figure does not mean:
63% less spend: the notice describes reports, not invoices;
63% fewer shoppers reached: the people who saw your ads on screen still saw them;
63% fake or invalid impressions: the ads were delivered to the device but were not on screen when counted, and Amazon filters invalid traffic separately;
63% fewer conversions: attribution and impression counting are separate rules, and Amazon has based view-through conversions on viewable impressions, not every served one, since 2019.
What happens to CPM, click-through rate and other metrics?
Metrics divided by impressions rise and metrics that do not use impressions stay where they were: if spend, clicks and purchases are unchanged, 63% fewer impressions makes reported CPM, click-through rate and Amazon DSP’s purchase rate about 2.7 times higher, while spend, CPC, ROAS and cost per purchase are untouched by the counting change itself.
| Metric | Effect if spend, clicks and purchases are unchanged | Watch out for |
|---|---|---|
| Impressions | About 63% lower, if the campaign matches the average | The drop varies by placement |
| CPM and eCPM | About 2.7 times higher | Spend has not risen |
| Click-through rate | About 2.7 times higher | Creative has not improved |
| Purchase rate (Amazon DSP) | About 2.7 times higher, because Amazon DSP divides purchases by impressions | Use purchases per click when you mean conversion rate |
| CPC | No change | Spend and clicks are its only inputs |
| ROAS and cost per purchase | No change from the count itself | Any movement has another cause |
| vCPM | No change, if viewable impressions stay the same | Viewable and total impressions are separate counts |
| Viewable rate | Depends on measurable impressions, its denominator | Any viewability share of total impressions rises by arithmetic |
| Average frequency | Lower, if reach stays the same | Frequency is impressions divided by reach |
| Reach | Cannot be worked out from the 63% | Depends on who loses their only impression |
In an illustrative US example, a campaign spends $10,000 on 2,000,000 impressions and gets 1,000 clicks: a $5.00 CPM and a 0.05% click-through rate. Count 37% of those impressions and the same delivery shows 740,000 impressions, a $13.51 CPM and a 0.135% click-through rate, with identical spend, clicks and sales.
Campaigns that mix store display with other inventory will see a smaller drop. If 40% of a campaign’s impressions come from the affected store placements and those fall by 63%, total impressions fall by about 25% (63% of 40%), and CPM rises by about a third.
Google made a similar move in 2018, when AdSense switched from counting served impressions to downloaded ones. It told publishers to expect fewer impressions and higher impression-based ratios such as RPM and click-through rate, and said earnings should not be affected. A counting change moves the ratios; money moves only if something else changes.
How does on-screen counting compare with industry standards?
On-screen counting is stricter than the industry’s minimum for counting a display impression, which is the moment the ad begins to render, but it is not the same as a viewable impression. Three trigger points matter, from earliest to latest:
Count-on-download: the ad starts loading on the device. This is Amazon’s current default count for store display.
Begin to render: the ad starts drawing on the page, whether or not it is on screen. Guidelines from the Media Rating Council, IAB Tech Lab and the Mobile Marketing Association made this the minimum for accredited impression counting in 2017.
Viewable: at least 50% of a display ad’s pixels on screen for at least one continuous second, which the standard treats as an opportunity to see the ad, not proof that anyone looked at it.
Amazon’s new trigger, the ad appearing on screen, sits between begin to render and viewable. The notice does not say how many pixels must show or for how long, so an on-screen impression is not automatically a viewable one.
Amazon already reports begin-to-render impressions accredited by the Media Rating Council, but only in Amazon DSP’s industry standard report, a custom report template; Amazon says reporting outside that template is not accredited. The 2027 change concerns the default impression count in everyday reports.
That template gives you a way to size the drop in your own account before January. Your new on-screen count should land roughly between your viewable impressions and your begin-to-render impressions, so comparing both with today’s default count gives a working range. This is our estimating method, not an Amazon figure.
What could change your costs or results?
Four things could change what you pay or deliver from January, and the notice answers none of them: billing, fees charged per thousand impressions, frequency caps and whatever produced Amazon’s test results.
Billing: the notice talks about reports, not invoices. Amazon’s DSP terms say the amount charged for each placement is based on Amazon’s ad server measurement, so ask which impression count your invoices will use from January, and whether CPM bids, floors and forecasts will be recalibrated for the new unit.
Fees: some fees are charged per thousand impressions, such as Amazon’s DSP audience fee, while third-party ad-serving and verification fees follow those vendors’ own counts. Ask which count each fee will use, and do not assume any of them falls by 63%. Fees charged as a percentage of spend do not change with the count.
Frequency caps: if ads that never reached the screen used to count towards your cap and no longer do, the same cap could allow more on-screen exposures per shopper. If caps keep the old counter, nothing changes. Ask which applies before you touch a cap, and do not multiply or divide caps by 2.7.
The test results: the notice says advertisers in testing saw ROAS rise and cost per action fall, with no figures, sample or method. A counting change cannot do that on its own, because neither metric uses impressions; it would take lower costs, better placement choices or better bidding signals, none of which the notice describes. Treat it as an encouraging claim, not a forecast for your account.
Three smaller details are open too: how much of the ad must be on screen and for how long, which marketplaces switch and in which time zone, and whether Amazon Marketing Cloud impression tables, exposure-based audiences or custom attribution change.
Amazon has changed measurement for the same inventory before. On 1 January 2026 it moved view-through attribution for Amazon DSP ads in its store, and for Sponsored Brands and Sponsored Display campaigns billed per viewable impression, to a shopping-signal enhanced last-touch model with a shorter window, a change covered in our post on the brand that paused $48K a month of DSP. A January 2027 report compared with January 2025 will sit across both changes.
What should advertisers do before and after 1 January 2027?
Before January, confirm the details in writing, save baselines and fix anything tied to impression counts; after January, measure the real drop by placement before you reset any benchmark, and explain the change before anyone panics about it.
What should you do before 1 January 2027?
Get it in writing. Ask Amazon to confirm which accounts, marketplaces, placements and campaign types are covered, and how billing, fees and frequency caps will work.
Size the drop. Compare default, begin-to-render and viewable impressions in the industry standard report to estimate your own range.
Save baselines. Export December 2026 data with impressions, viewable impressions, clicks, spend and conversions, with the metric definitions and attribution settings beside them. Amazon plans to retire the legacy Sponsored ads and Amazon DSP report pages by 31 December 2026, so rebuild scheduled reports in unified reporting before then.
Mark the date. Annotate dashboards and client reports at 1 January 2027, switch off year-on-year impression alerts and check automated rules that react to swings in impressions, CPM or click-through rate.
Review impression-based terms. Check impression targets, CPM ceilings, click-through goals, impression guarantees and every fee charged per thousand impressions.
Budget on spend. Plan 2027 on spend and outcomes, not on a 63% saving or a 2.7 times price rise.
What should you do after 1 January 2027?
Measure the real drop by placement, device and campaign against your December baseline.
Read impressions next to spend, clicks, viewable impressions and conversions, so a counting change is not mistaken for a delivery problem.
Check that frequency caps and automated rules behave as they did before, and that invoices match the count Amazon said it would bill on.
Recalculate CPM and click-through rate from totals rather than averaging campaign-level rates, and label any restated history as an estimate.
Rebuild CPM and click-through benchmarks from at least four weeks of post-change data, and allow for post-holiday seasonality when you compare.
Year-on-year charts need the most care. January 2027 against January 2026 crosses the new count, and against January 2025 it crosses the view-through attribution change of 1 January 2026 as well, which our post on the brand that paused $48K a month of DSP covers. For those charts, swap impression-based lines for viewable impressions and outcomes, or carry a note on the chart itself.
Why should you report viewable impressions after January?
Because viewable impressions should be the one impression count the change leaves alone. An ad that met the viewability standard was on screen, so it counted under the old method and still counts under the new one, which makes viewable impressions, and the cost per thousand viewable impressions (vCPM) you can work out from them, the steadiest line to carry across 1 January.
This is our reading of the two definitions, and it holds only if Amazon keeps measuring viewability the same way. One trade report says viewable impressions stay as they are, but the notice as reproduced does not mention them, so confirm it with Amazon before you rebuild a dashboard around viewable counts.
The viewability standard for display ads asks for at least half of the ad’s pixels in view for one continuous second, which the industry treats as an opportunity to see. Amazon’s notice does not say how much of an ad must appear on screen, or for how long, to count under the new method, so expect your new impression count to sit above your viewable count and below the old one. Our guide to performance measurement and analytics in Amazon DSP covers the wider set of DSP metrics and reports.
From January, we would report Amazon DSP delivery in four layers:
Viewable impressions and vCPM: for delivery and cost comparisons across the switch, with vCPM worked out as spend divided by viewable impressions, times 1,000;
Impressions: for pacing and frequency only, labelled with the counting method in force;
Viewable rate: with care, because Amazon’s viewability guide defines it as viewable impressions divided by measured impressions, and the measured count may change with the new method;
Outcomes: spend, sales, ROAS and cost per purchase for every judgement about performance.
How should you explain the change to clients or your board?
Separate what the dashboard counts from what shoppers saw and what you paid, and answer the six questions people ask first:
| Question | Short answer |
|---|---|
| Have we lost 63% of our audience? | No. Ads that loaded but never reached the screen stop counting; we will track reach and delivery separately. |
| Why has CPM jumped? | The same spend divided by fewer counted impressions. We will check whether invoices or cost per purchase changed too. |
| Why has click-through rate improved? | Fewer counted impressions with the same clicks. It is not evidence that the creative got better. |
| Will ROAS improve automatically? | No. ROAS uses sales and spend, not impressions, so it moves only if those move. |
| Were the old reports wrong? | No. Download, on-screen and viewable are different counting points, and the old figure measured downloads. |
| Should we cut the budget? | Not because of the count. We judge budget on spend, sales, cost per purchase and incremental results. |
Send the answers before the first January report goes out: a drop explained in advance reads as a reporting change, while the same drop explained afterwards can sound like an excuse.
Add one line to every January report so the change is explained where people see it. For example: “From 1 January 2027, Amazon counts store display impressions only when the ad appears on screen, so lower impressions and higher CPM and click-through rates reflect the new count rather than a change in the campaign.”
Frequently asked questions
Why is Amazon changing how it counts display impressions?
Amazon’s notice gives no reason beyond describing the new count and saying advertisers saw higher ROAS and lower cost per action in testing. The change moves Amazon’s default store display count past the industry’s begin-to-render minimum, but Amazon has not said whether accreditation, advertiser pressure or something else prompted it.
Should you scale down 2026 impressions by 63% to compare years?
No. The 63% is an average across placements, and your mix of home page, product page, search and order page impressions will differ from Amazon’s average, so a flat cut distorts placement-level trends. Keep 2026 figures as reported and build 2027 benchmarks from post-change data instead.
What happens to campaigns that run across 1 January 2027?
Their reports will mix both methods: days before the switch on the download count and days after it on the on-screen count. Split any flight that crosses the date into before and after periods, and ask Amazon how it will treat delivery commitments and guarantees that span the change.
What should Amazon Marketing Cloud users do before January?
Save aggregated baselines now: spend, impressions, reach, frequency distributions and the conversion-path queries you rely on, with each query’s version and date range. AMC returns aggregated results rather than shopper-level records, so a saved baseline is the simplest way to repeat the same comparison after the switch. Our guide to Amazon Marketing Cloud explains how queries and audiences work.
Do other ad platforms count display impressions on screen?
No, not as a rule, and methods differ even within one platform. In February 2027, Google moves display impressions in its ad tools from download to begin to render, which does not require the ad to be on screen, so a Google display CPM and an Amazon store display CPM will rest on different definitions.
What does this article not cover?
This article covers the 2027 change to how Amazon counts store display impressions. Related topics live elsewhere:
-
Performance measurement and analytics in Amazon DSP for the wider set of DSP metrics and reports;
-
Why DSP campaigns can look like they are failing for how attribution and Amazon Marketing Cloud change the picture.
What is the next step?
This month, ask Amazon or your agency to confirm coverage, billing, fees and frequency caps in writing, and export your December baseline before the legacy report pages close.
Clear Ads is an Amazon Ads Advanced Partner that manages Amazon PPC, Amazon DSP and full-service accounts for brands worldwide. If you want help preparing your DSP reporting for the change, request a call back from the Clear Ads team.
