Meta Artificially Inflated Ecommerce Ad Metrics, Former Employee Claims

Meta, the parent company of Facebook and Instagram, has been accused of artificially inflating the performance metrics of its ecommerce advertising product, Shops ads, according to a whistleblower complaint filed Wednesday in a U.K. employment tribunal.

The complaint, brought by Samujjal Purkayastha, a former product manager on Meta’s Shops ads team, alleges the company misled advertisers by overstating the return on ad spend (ROAS), making its newer ad offering appear more effective than competing products, ADWEEK reports.

Allegations of Inflated Metrics

According to the filing with the London Central Employment Tribunal, Meta allegedly boosted Shops ads’ performance numbers by:

• Counting shipping fees and taxes as part of total revenue

• Subsidizing bids in ad auctions to secure more prominent placement

• Applying undisclosed discounts to give the impression of stronger results

Internal reviews conducted in early 2024 revealed Shops ads’ ROAS had been inflated between 17% and 19%, according to the complaint. Meta’s other ad products – as well as competitors like Google – calculate ROAS using net figures, excluding shipping and taxes. Without the added fees, the filing claims, Shops ads performed no better than Meta’s traditional ad products.

“This was significant,” the complaint states. “In addition to the ROAS performance metric being overstated by nearly a fifth, it meant that, rather than having exceeded our primary target, the Shops Ads team had in fact missed it once the figure was reduced to take account of the artificial inflation.”

Meta’s Push After Apple’s Privacy Changes

The filing links these alleged practices to a broader effort inside Meta to recover from the effects of Apple’s App Tracking Transparency (ATT) feature, rolled out in 2021.

Apple’s policy limited access to iOS user data, a cornerstone of Meta’s ad business. Former Meta CFO David Wehner warned during a 2021 earnings call that the change could cost the company “on the order of $10 billion.”

By encouraging advertisers to use Shops ads, which kept transactions inside Meta’s apps, the company could collect more first-party purchase data and reduce its reliance on Apple’s tracking permissions.

According to Purkayastha, Meta began subsidizing Shops ads in auctions, sometimes by as much as 100%, ensuring they appeared more often than other ad formats. This increased visibility, artificially boosted conversions and made Shops ads seem like a stronger investment.

Purkayastha joined Meta in 2020 as part of the Facebook Artificial Intelligence Applied Research team before being reassigned to the Shops Ads team in March 2022. He remained at the company until February 19, 2025.

The complaint says Purkayastha repeatedly raised concerns in meetings with senior leadership between 2022 and 2024, questioning the accuracy of Shops ads’ reported results. He claims the company continued using the disputed methodology despite internal objections.

The complaint also points to Meta’s tracking tools as part of its strategy to maintain advertising performance after Apple’s privacy changes.

• Aggregated Event Measurement (AEM1), introduced in April 2021, used machine learning to estimate conversions while respecting users who opted out of tracking.

• AEM2, rolled out shortly thereafter, allegedly linked in-app activity to browsing and purchases on third-party sites using personal identifiers like names, emails, phone numbers, and IP addresses.

According to ADWEEK;

In the complaint, Purkayastha said he believed AEM2 bypassed restrictions imposed by Apple’s privacy framework, though it mitigated much of the loss of data from the privacy changes.

Purkayastha was terminated from Meta in February 2025, according to the complaint. His filing with the employment tribunal is part of an application for interim relief, requesting that his former position be reinstated. 

Meta has not yet commented on the complaint. A request for comment from ADWEEK, which first reported the filing, was not returned by the time of publication.