
Your campaigns are not about to get worse. The ruler you read them with is about to change. As of November 2026, Amazon has told advertisers that brands will see 63% fewer display impressions across its retail properties once it counts an impression only when the ad renders on screen, from 1 January. If you buy media for a Singapore business, the decision in front of you is whether to cut spend or to rebuild the baseline your SEM agency reports against. This piece works through the second option and says where it does not apply.
Read the mechanics before the headline. In a notice distributed among advertisers and seen by Digiday, Amazon set out the shift: display ads on its homepage, product pages and search results will be counted when they render on the user's screen rather than when they begin downloading to a device. The new rule applies from 1 January. Google announced a similar transition at the beginning of September, effective 17 February. Spokespeople for both companies declined to comment when Digiday asked.
Digiday's briefing of November 2026 reports that six buyers it spoke to are amending reporting methodologies and calming clients. None of them described changing what they buy. The work is in how the buy is described and compared, which is a different discipline from media planning and often sits with whoever owns the client report.
A shopper who never scrolls to the fourth row of a Store page was not really reached by the ad sitting there, even though the old count logged it as downloaded. Under the new rule that impression falls out of the total. The number of people using the page has not changed; the threshold for being counted has.
The arithmetic then does something predictable. Cost per thousand impressions divides spend by counted impressions. Spend stays flat, the count drops, and CPM rises. A client looking at that line will ask whether the buying got worse. Your answer has to show that delivery held and the denominator moved.

SEM, search engine marketing, is the practice of buying placement in search results and in the surfaces that behave like them: text ads, Shopping listings, Performance Max, and the app and video inventory a search platform bundles alongside. Media buying is the wider discipline that also covers display, video and retail-media space. Paid search is one line inside it, and in many agency structures the same team runs both lines and adds their numbers into one client report.
That habit is what the new rules expose. When you sum Amazon display impressions, Google display impressions and search impressions, then divide a single budget by the total, you are averaging figures measured with different rulers. The definition matters because a metric only travels as far as the counting rule behind it.
Search is bought against intent: a query, a keyword, a match type. Display and retail media are bought against audience and context. The two halves respond to different levers, and the counting change lands mostly on the display and retail-media half. Search reporting keeps its own definitions, which is why an Amazon date in January and a Google date in February can both be true and still produce two different percentage drops in the same dashboard. A SEM agency that reports the two halves as one blended figure will struggle to explain either drop.
On paper, yes, and buyers have already said so out loud. Hannah Schatz, vice-president of programmatic operations at Coegi, told Digiday that fewer impressions will mean higher CPMs. Nola Ladd, awareness media supervisor at Collective Measures, expects more competition on the platform as buyers chase the placements that still count.
Neither expects the market to reprice overnight. Dru Sil, global product managing director at Stagwell Media Platform, said Amazon has not issued guidance on the effect on pricing or costs, nor on whether billing will follow the new definition. In his words, the implications are for buyers to discover. Any budget conversation held before that guidance arrives is a conversation held on assumptions.
A stricter definition sorts placements by whether they can plausibly be seen. Display slots low on a Store page rarely enter the viewport. They lose counted volume and become less attractive to buy. Slots near the top of a page stay in view, so they keep their volume and draw more demand. If you buy a mix of both, your blended CPM will move even though no single placement was repriced.
If display is a large share of your buy, the Video and Display Buying: An SEM Agency View piece explains how that inventory is planned and priced.
For most advertisers the answer is no, and that is the part to tell a worried client first. Jared Smith, senior director of strategy and innovation at January Digital, told Digiday that the shift will make reporting metrics look different but should not move ROAS or CPA, the outcomes clients care about. Ernie DeColte, senior vice-president of media at Level, put it harder: impressions and CPMs are tool metrics used to optimise towards business impact, and the change may be the wake-up call that pushes some advertisers to buy differently.
The practical risk sits in automation. Paul DeJarnatt, who heads investment at Novus, warned that automated processes and reconciliations built on impression data will need recalibration. A bidding rule that chases impression volume, or a forecast trained on the old count, will misread the new one and keep acting on it long after the label on the chart has changed.
A SEM agency that ties bidding to revenue per dollar of spend will feel this shift later and less painfully than one whose targets are set on impression volume.

Five steps, in the order that keeps the report usable, for any team that buys display or retail media alongside search.
If your team runs its own Google Ads account, Google Ads API v26.2: What to Check covers the version change worth checking before any reporting script is rebuilt.
A budget cut is the wrong reflex for most advertisers. Dru Sil of Stagwell Media Platform told Digiday that he expects clients to keep running what they have been running, because the only way to build a baseline in January and February is to keep ads running. Cutting spend in the same month the counting rule changes leaves you with two unknowns and no comparison.
The reconciliation is where the counting change stops being a reporting annoyance. Pull the platform's counted impressions and spend into the same sheet as your CRM's signed deals, then divide. That single ratio tells you whether a falling impression count was accompanied by a falling contribution to revenue, or not.
In the SEM agency Singapore market, clients have started asking for the reconciled number before the impressions chart, and the four questions below are the ones a report now has to answer.
| Question the client asks | Report that fails | Report that survives |
|---|---|---|
| How many people saw the ad? | A raw impression count | Counted impressions, labelled with the counting rule and the date it changed |
| Is our CPM rising? | CPM compared with last year's CPM | CPM shown beside delivery, with the denominator stated |
| Did the ads work? | Conversions recorded inside the platform | Signed deals reconciled from the CRM, with the attribution window named |
| What should we change? | A dashboard of charts | A ranked action list, each item with an owner and a review date |
A report built on those four lines answers the question a manager will ask in the first week of January: did performance get worse, or did the ruler change? The answer decides whether the next conversation is about budget or about benchmarks.
Topkee Media runs Google Ads as a partner on a revenue objective rather than a click objective, across Search, Display, Performance Max, YouTube, App and Shopping. Its Campaign Intelligence Report pulls Google Ads, Yahoo Ads, Facebook Ads and GA4 into one report, with attribution starting from the advertiser's own data layer instead of any platform backend, and scores every channel on one fixed yardstick: signed-deal value divided by ad spend. The same report tracks a six-layer funnel from impressions to signed deals, so a smaller impression count appears as one line in a chain that ends in revenue, and it closes with a ranked action list.
That structure answers the question this news leaves open. A published case on the service page shows the yardstick in use: an enterprise broadband client reached ROAS peaking at 654.34% over five months, and a repair business grew from a single outlet to 13 locations within a year on a system tracked through TTO CDP and Leadbox CRM. A buyer comparing a SEM agency Singapore shortlist can put one question to each candidate: which of these numbers would still mean the same thing if the platform changed its definition tomorrow?
Topkee lists each case on that page with its spend, income and return, so a reader can check the figures against the same calculation rather than take a headline percentage on trust.

Advertisers with a single channel and a short purchase cycle can sit part of this out. A clinic buying search ads for bookings has one conversion path and one countable outcome, so the display counting change barely touches its report. A brand that sells through distributors and never receives deal-level data cannot use signed-deal value as a yardstick at all; it needs a different check, such as verified lead quality from its channel partners. And a business with too little volume to build a monthly baseline will spend more on reconciliation than the insight is worth.
Neither does the advice suit a team that cannot get read access to its own ad accounts and CRM. Rebuilding a baseline requires the data, and permission to join it, on your side.
Delivery should stay close to what you bought. The count of impressions falls because fewer of them meet the stricter definition, and Amazon expects the drop to reach 63% on display placements across its retail properties. Reach, clicks and sales are separate figures and need their own comparison before you draw a conclusion.
For most advertisers, no. Buyers told Digiday they intend to keep spending at familiar levels through January and February so that a new baseline exists to compare against. A cut in the same month as a measurement change leaves you unable to tell which effect caused the result.
Both fall inside the same planning cycle. Amazon's rule starts on 1 January and Google's on 17 February, so a dashboard that blends the two platforms will show a first movement in January and a second in February. Label the dates per platform rather than per report.
Ask the vendor or the internal owner which definition each column uses. A dashboard built before the change may keep summing download-based impressions beside render-based ones, which makes month-on-month comparison meaningless. Rebuilding the calculation matters more than restyling the chart.
Long enough to need a deliberate baseline. If you keep spend steady through January and February, you have two clean months under the new rules, and the comparison after that is between like and like. Advertisers who change budget, creative and counting rule in the same month will not know which change moved the result.