Amazon Will Count DVA Display Impressions Only When Viewed - Expect a 63% Drop Starting Jan 1
What happened: On November 3, Amazon announced it is changing how it counts DVA display impressions on the Amazon Store. Today, an impression registers the moment an ad begins downloading to a shopper's device - whether or not it ever actually appears on screen. Starting January 1, 2027, Amazon will only count an impression once the ad renders in the viewport. Amazon's own notice says counts will drop roughly 63% on average, with bigger drops for placements lower on the page.
What is actually changing
The shift is from count-on-download to count-on-render. Per Amazon Advertising's own announcement and reporting from PPC Land, the change applies only to DVA (display, video, audio) display ads on Amazon Store placements - homepage, product detail pages, search results, and order pages. Sponsored ads and third-party inventory are unaffected. Amazon's viewability bar follows the Media Rating Council standard: 50% of an ad's pixels in view for at least one second for display, two seconds for video.
Why Amazon is doing this now
Count-on-download has been below industry standard since the IAB and MRC set begin-to-render as the viewability floor back in 2017. Amazon's move brings its own Store placements in line with how the rest of programmatic display has measured impressions for years, and it comes alongside the broader DVA+ rollout that merges DSP and Ads Console reporting into one view. Cleaner impression data makes that unified reporting layer more defensible to advertisers who already compare it against third-party verification.
What this means for brands
- Reported impressions drop, not actual delivery. The ~63% figure is a counting-methodology change, not a cut to how many ads Amazon serves. Budgets and bids don't need to move on this alone.
- CPM math gets messier. Amazon has not confirmed whether CPM pricing adjusts to match the new, smaller denominator. Until it does, a brand's effective cost-per-viewable-impression could look like it jumped even though nothing about delivery changed.
- Historical comparisons break at the boundary. Any dashboard or quarterly report that trends impressions across January 1, 2027 will show a cliff that has nothing to do with campaign performance. That line needs an annotation before someone on the brand side flags it as a crisis.
- Homegrown reporting pipelines need a definition update. Brands pulling DVA data via the Amazon Ads API into their own BI stack or a custom attribution model are pulling a field whose definition is quietly changing underneath them on a fixed date.
That last point is where most brands get caught off guard: a custom reporting pipeline that blends Amazon Ads API pulls with Google Analytics or an internal data warehouse has to be updated to handle the new impression definition on both sides of the January 1 cutoff, or every downstream metric built on top of it - CPM, CTR, even blended ROAS - quietly skews. Rebuilding that logic correctly, backfilling the old definition for historical comparisons, and testing it before the holiday reporting crunch is exactly the kind of focused engineering work that's hard to staff internally on short notice, which is why brands often bring in YuSMP Group's staff augmentation team to get a dedicated engineer on the fix without a multi-month hiring cycle.
What to check before January 1
- Find every place DVA impressions feed a downstream metric - native Amazon reports, any API pull into a warehouse, and any dashboard that calculates CPM or CTR from raw impression counts.
- Confirm whether CPM billing adjusts. Watch for Amazon guidance closer to the date; if pricing doesn't rescale with the new count, effective cost-per-impression will look higher even with flat spend.
- Annotate the transition in reporting tools. Add a note at the January 1 boundary in any dashboard or spreadsheet that trends impressions over time, so a 63% drop doesn't read as a campaign failure internally.
- Re-test custom pipelines against both definitions. If a pipeline pulls raw impression data via API, verify it handles the old and new counting methods correctly on either side of the cutoff before relying on year-over-year comparisons.
For the broader question of when a connector is enough versus when a dedicated build earns its cost, see our breakdown of syncing Amazon and Walmart orders to Salesforce, and for the API layer underneath reporting work like this, see our guide to integrating your ERP with the Seller Central API.
