How should we compare Meta results across 2026 reporting changes (7-day and 28-day view windows dropped from the Ads Insights API on January 12, 2026; click-through narrowed to link clicks in March 2026) when reading year-over-year or before-and-after performance?
Our answer
Treat January 12, 2026, March 2026 and November 12, 2026 as reporting breaks. Compare years and before-and-after periods on backend revenue, orders or new customers, not Meta-reported conversions, and judge Black Friday on store orders.
- Confidence
- Medium
- Strongest proof
- 8 Meta rules
- 2 datasets, 3 practitioners, 4 opinions or claims
- Sources
- 13 sources
- 17 statements
- M
New here? Take the 1-minute tour
Why this is the answer
4 reasons and our judgmentTwo of the 2026 reporting breaks change what Meta counts. Since March 2026, Meta counts only link clicks as click-through for website and in-store conversions, and on January 12, 2026 its Ads Insights API stopped returning 7-day and 28-day view windows (Meta).
From November 12, 2026, outside the EU and UK, Meta's setting to measure conversions on customer devices is on by default (Meta's help page, date reported by Stape). Stape and PPC Land expect more credited purchases with no change in sales.
The March 2026 change to link clicks cut Meta's click numbers. Using Haus holdout tests as the benchmark, Olivia Kory found it cut 7-day click conversions by about a sixth relative to Incremental Attribution.
Measurement sellers point to backend results instead. Wicked Reports, FunnelFox and AdAdvisor advise comparing backend revenue, CRM orders or blended ROAS.
Our judgment
- No Meta-reported series is like for like across March 2026, so compare backend results. Click-through counts only link clicks, and AdNova notes engage-through is fixed at one day, so some conversions fell out of reporting.
- Treat January 12, 2026 and the March 2026 rollout as two separate breaks. One removed view windows from the API and the other changed what counts as a click, and FunnelFox notes click-through windows were not affected in January. For API reports, rebuild 2025 comparisons on the windows Meta still returns, knowing that even 7-day click is not like for like across March.
- Use Meta's own numbers across a break only to size the shift, not to judge performance. Compare counts from both sides and add engage-through and other columns with Compare Attribution Settings, as AdNova and Jon Loomer advise. Re-check any ROAS target or incrementality factor built on pre-March 7-day click, as Olivia Kory advises.
- Treat November 12, 2026 as a third break for any dataset that keeps measuring conversions on customer devices. Set the daily ratio of Meta-reported purchases to store orders before it, so a Black Friday rise in Meta numbers reads against store orders, not as better ads. Stape says reported cost per purchase could fall and ROAS rise with no change in sales, and Yiqi Wu of Aimerce, as PPC Land quotes him, advises comparing Meta purchases with Shopify orders by day.
Our judgment, built from Meta, AdNova, Wicked Reports, FunnelFox, AdAdvisor, Meta for Developers (Marketing API blog), Jon Loomer, Olivia Kory, Stape and PPC Land: our reading of these sources, not a finding from any one of them.
What the sources say
17 statements from 13 sourcesMeta's help page says on-device conversion measurement is on by default, is not active in the EU and UK, works alongside the Pixel and Conversions API, and does not change attribution windows or models.
“Measuring conversions on customer devices affects which conversions can be measured, not how they're credited to your ads.”
Meta's on-device conversion setting is on by default for eligible pixels, which must be installed directly on the page on a site that requires login by email or phone. It is not active in the EU and UK.
“Your website must require people to log in to an account using an email or mobile phone number.”
Meta turns on measuring conversions on customer devices by default and says keeping it on shows more conversions and gives a stronger signal. It gives no figure and does not say if bidding uses it.
“Measure conversions on customer devices is turned on by default.”
Andrew Foxwell says Meta measures differently than it did on Black Friday 2025, so comparing 2026 results with 2025 is flawed.
“Meta is measuring differently than it was last black Friday.”
PPC Land says that from November 12, Meta's on-device conversion counting will make campaigns look like they sell a little more, while attribution windows and models stay the same.
“Nothing about the campaigns will have changed. What changes is the counting.”
PPC Land reports that from November 12, 2026 Meta will measure some purchases on shoppers' devices, on by default and switchable per dataset. Whether it feeds bidding or only reporting is unknown.
“I wouldn't opt out by default”
Stape, which sells server-side tracking, relays that Meta's on-device conversion matching starts November 12, 2026, on by default outside the EU and UK, and may lift reported conversions without real sales changing.
“advertisers may see an increase in attributed conversions and reported conversion value.”
Captain Compliance, a privacy tool vendor, relays Meta's notice that from November 12, 2026 the Pixel will measure some missed conversions on the customer's device. It says the notice is silent on consent.
“no updates to existing integrations are required”
AdAdvisor, which sells an AI buying agent, says to treat 2026 as a reporting break: Meta stopped returning 7-day and 28-day view windows via its Insights API on January 12 and narrowed click attribution in March.
“Mark 2026 as a reporting break.”
Haus's holdout tests found Meta's Incremental Attribution tracked measured lift more closely than standard windows, and optimizing to it drove about 1.2x the incremental return. Sample size not given.
A creative analytics vendor says after March 2026 conversions after likes, comments and shares moved to a fixed one-day window, so some vanished from reports. A drop that starts one day and holds is a counting change.
“a counting change”
FunnelFox, an app funnel vendor, says Meta's attribution changes caused reported conversion drops even where actual sales hadn't moved. Compare Meta numbers with CRM and revenue data, not pre-March reports.
“Stop comparing this month's reported conversions directly to pre-March numbers”
Code3 says clients' Meta-reported results rose after Meta's March 2026 update: ROAS 30% higher and cost per purchase 26% lower than a year earlier. The author thinks the update mainly credits more conversions.
“Meta isn't necessarily driving 67% more efficient traffic overnight. It's doing a better job connecting conversions back to Meta touchpoints.”
Wicked Reports, an attribution vendor, says click-through now counts only link clicks, so a brand that reported 4.2x ROAS may now see 3.1x, not because sales dropped but because passive engagement is no longer counted.
“not because sales dropped, but because passive engagement is no longer counted as a conversion event.”
Jon Loomer says Meta's default attribution is 7-day click, 1-day engage-through and 1-day view, and the Compare Attribution Settings columns show other windows, including 28-day click, beside it.
“Meta did bring 28-day click attribution back for reporting purposes”
Meta changed click-through attribution in March 2026 to count only link clicks, to match tools like Google Analytics. Other clicks move to engage-through, and a video engaged view drops from 10 seconds to 5.
“allow your Meta reporting to align better with third-party tools like Google Analytics”
Meta stopped returning 7-day view and 28-day view attribution windows in the Ads Insights API from January 12, 2026; 1-day click, 7-day click, 28-day click, 1-day engaged view and 1-day view remain.
“On January 12, 2026, we will deprecate certain attribution windows available today with the Ads Insights API’s”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer