The ROAS figure in Meta Ads Manager is not your true return on ad spend. It counts conversions that would have happened anyway and misses some that it drove. Accurate Facebook and Instagram ROAS measurement requires geo holdout tests and marketing mix modelling alongside the platform's own reporting.
Meta Ads Manager reports impressive ROAS figures for most advertisers. The problem is that those numbers are based on attribution windows that give the platform credit for conversions it did not cause. A customer who was going to buy anyway, and who happened to see a Facebook ad in their feed the day before, gets counted as a Facebook-driven sale.
This is not a conspiracy. It is simply how last-click and view-through attribution work. The fix is not to distrust Meta entirely, but to add measurement methods that separate genuine incremental sales from sales that would have occurred regardless.
Why Meta reporting inflates ROAS
Meta uses a default 7-day click, 1-day view attribution window. This means any purchase made within seven days of a click or one day of an ad view gets credited to that ad. Given how often your existing customers and organic website visitors are in your retargeting audiences, a significant portion of these attributed conversions would have happened without the ad.
iOS 14 and subsequent Apple privacy changes made the problem worse. With less signal available from iPhones, Meta's models started inferring more conversions rather than observing them directly. Reported conversions sometimes exceed actual conversions measured server-side, a discrepancy that can reach 30 to 50 percent for some advertisers.
The Meta Conversion API: fixing the signal problem
The Meta Conversion API (CAPI) sends conversion data directly from your server to Meta, bypassing browser-based tracking limitations. It improves the accuracy of the signal that Meta receives and reduces discrepancy between what you measure server-side and what Meta reports.
CAPI does not fix the attribution window problem or the over-counting of conversions that would have happened anyway. It fixes the under-counting of conversions caused by blocked browser pixels. Both problems are real. CAPI addresses one of them.
- Implement CAPI alongside your pixel, not as a replacement for it. Running both in deduplication mode gives Meta the strongest possible signal.
- Use server-side event matching with hashed email addresses and phone numbers to improve match rates.
- Set your attribution window in Ads Manager to match your actual purchase cycle, not the default. Shorter windows produce more accurate comparisons.
- Check your event match quality score in Events Manager. Scores below 6 out of 10 indicate a weak signal that is affecting your campaign optimisation and reporting.
Geo holdout tests: measuring true incrementality
A geo holdout test (also called a ghost ad test or conversion lift test) withholds Facebook and Instagram advertising from a matched control group while running normally for the test group. By comparing conversion rates between the two groups, you measure what percentage of conversions are genuinely incremental to the ads.
Most brands that run their first Meta incrementality test find that true incremental ROAS is 30 to 60 percent lower than the number in Ads Manager. That is not a reason to stop advertising on Meta. It is a reason to reallocate budget toward your highest-incrementality audiences and campaigns.
Meta offers a built-in Conversion Lift tool that runs holdout tests within the platform. This is useful and free but has a limitation: Meta controls the methodology. For an independent validation, work with a third-party measurement provider who runs the holdout externally using your CRM and sales data.
Marketing mix modelling as a cross-check
An MMM treats Meta as one channel among many. It estimates Meta's contribution to sales using variation in spend over time, rather than tracking individual users. Because MMM does not rely on cookies, attribution windows, or pixel data, its estimate of Meta ROI is completely independent of the platform's own reporting.
When your MMM estimate of Meta ROI is materially lower than what Ads Manager reports, the true figure almost certainly lies somewhere between the two. The MMM may be understating Meta's effect (it often struggles to attribute fast-acting social response correctly), and Ads Manager is overstating it. Understanding both numbers lets you negotiate between them intelligently.
Audience segmentation and incrementality
Incrementality is not uniform across your Meta campaigns. Prospecting campaigns targeting new audiences tend to have higher incrementality than retargeting campaigns that show ads to people who already visited your website. Your most loyal existing customers produce near-zero incremental conversions when retargeted because they would have bought anyway.
Running separate incrementality tests for prospecting and retargeting, or analysing them separately in an MMM, reveals where your budget is genuinely working versus where it is spending money on conversions that were already secured.
Should I change my bidding strategy based on incrementality results?
Yes. If your holdout test shows that retargeting audiences produce low incrementality, the correct response is to reduce the retargeting budget or tighten the audience definition to exclude recent purchasers and high-intent segments. Reinvest that budget in prospecting or in the audiences where incrementality is high. Meta's own Advantage+ campaigns and broad audience bidding can improve incrementality by pushing spend toward genuinely unexposed users.
How do I reconcile the difference between Meta-reported conversions and my own sales data?
Start by pulling both datasets for the same period and calculating the ratio of Meta-attributed conversions to total sales. A ratio above roughly 30 to 40 percent for a mid-sized brand usually indicates over-attribution. Check for obvious causes: a long attribution window, retargeting campaigns targeting existing customers, or a recently changed pixel. Then commission an incrementality test to establish the ground truth. Use that test result to apply a correction factor to your ongoing Ads Manager reporting.
Does Instagram measurement work the same way as Facebook measurement?
Within Meta Ads Manager, Instagram and Facebook share the same attribution and measurement infrastructure. If you break out performance by placement, you can compare Instagram versus Facebook ROAS within the platform. But both placements will have the same over-attribution issues, and both benefit from the same corrections: CAPI, holdout testing, and MMM. The incrementality differences between placements are best analysed once you have established your overall Meta incrementality baseline.