Meta Is Making a Major Change to Its Algorithm Around Conversions

Meta is making a change to its advertising algorithm that will make the platform temporarily look like it’s performing worse for some advertisers. The change could raise ad prices but also increase advertisers’ trust in the fidelity of Meta’s ad systems, ADWEEK has learned.

The change affects conversion campaigns where a marketer wants someone to make a purchase. Meta will now give more weight to the conversions occurring one day from when a user first clicks, specifically looking at the one-day conversion window. A conversion window is the time between when a user sees an ad and completes an action.

Shorter conversion windows are harder for platforms like Meta because they have less time to prove their ads worked. With longer conversion windows, like a seven-day window, there is more likelihood that a user purchased for reasons outside of the ad.

Previously, the algorithm behind conversion campaigns was equally weighted toward all lengths of conversion windows.

“This ads ranking change will more heavily favor 1 day click conversions, whereas previously, all attribution windows were weighted evenly,” according to a note digital agency Wpromote sent its clients about the change, which was seen by ADWEEK. “From early testing, Meta has seen an average 30% improvement to Meta-attributed conversions in third-party analytics tools. However, this also means that Meta’s in-platform data will be impacted, with the likelihood of decreased conversions reported, along with potentially higher CPAs and CPMs.”

With more conversions only counted by Meta if they happened within one day of seeing the ad, advertisers might see fewer conversions reported for a given campaign than they did previously. Ad prices also might be higher because the change will cause Meta to target a smaller pool of users, looking for those who are more likely to convert in one day.

Advertisers who hadn’t already been optimizing for one-day conversion windows or who sell luxury products with longer purchase windows are more likely to see ad performance decline, said Darren D’Altorio, vice president of social media at Wpromote.

The change is significant because Meta typically doesn’t tell advertisers much about how its algorithm works. Brands have also recently been increasingly skeptical of how Meta measures its ads, ADWEEK has reported.

“We are super excited about these updates,” D’Altorio said. “This is an exciting update for performance advertisers who validate results with third-party measurement tools (which most do) and illustrates that Meta consistently leads the pack as it relates to adtech innovation.”

Separately, Meta announced a slew of changes on Aug.14 to its ad system around how it optimizes and measures conversions. The greater weighting of the one-day attribution window is not directly mentioned in Meta’s blog post, though it was part of how these changes were explained to advertisers, according to D’Altorio.

“With today’s updates we expect advertisers, especially those who use third-party analytics tools, to see improvements in Meta-attributed conversions,” a Meta spokesperson said in response to questions about the role of the one-day attribution window in its ads algorithm. “As the holiday season approaches, our goal is to continue to help improve advertiser performance and deliver more of the results they value.”

Meta did not respond to comment on the change in the weight of a one-day conversion window.

More incrementality and comparison with third parties

Besides the change to conversion windows, Meta is also making changes that show advertisers how many users took an action on an ad as a result of specifically seeing it on one of Meta’s platforms.

When Apple deprecated mobile identifiers in 2021, Meta had less data about who to target and how to measure whether those ads worked, causing ad performance to plummet.

In the past three years, Meta has solved this problem with AI which has improved performance, but advertisers worry the campaign reports Meta gives are more modeling than reality. They worry that the modeling makes Meta ads look like they perform better than they do.

Meta’s updates to its ad systems could alleviate those concerns.

Meta is now allowing advertisers to optimize for incremental conversions. A user who scrolled past a Meta ad for shoes might have bought the shoes anyway, but an incremental conversion would represent a customer brought in uniquely by the ad.

Meta calculates incrementality by comparing conversions from a group of people exposed to an ad to conversions from people who didn’t see any advertisement. The difference between the two helps figure out how many conversions were incremental.

Incrementality testing is a big reason why ad buyers have been turning to third-party measurement solutions like Rockerbox, TripleWhale, and Measured.

Another reason advertisers have turned to these vendors is to figure out which platform is most responsible for a sale. A user could have seen an ad for a product on Meta, TikTok, and Google before making a purchase, and each platform could be taking credit for driving the sale.

Now, advertisers can connect these third parties with Meta, starting with Google Analytics and Northbeam. Meta will soon expand similar tools to Adobe and Triple Whale. Meta will use this data in campaign optimization to improve its relative campaign performance.

“One of the biggest criticisms of Meta, specifically from performance advertisers, is that its in-platform attribution is wildly inaccurate compared to a third party ‘source of truth,’ ” D’Altorio said. “These changes will better calibrate their ad system to close that gap.”

Meta has already started to use some preliminary third-party data from early tests with advertisers to make changes to its overall advertising algorithm.

“We would like our system to be customized to deliver [advertisers’] specific way they think their media performs. We’re not there yet,” Fred Leach, vp of product management, told ADWEEK. “That’s ultimately what the connection with third-party analytics tools should allow us to do. But right now, we’ve gained enough aggregate learning where we can tune the system to where, on average, we believe it’s going to deliver more value for advertisers”

Correction: The headline on an earlier version of this article misstated that Meta was changing how conversions are measured. The headline has been updated to reflect that Meta has changed how its algorithm is optimized. This article has been updated to clarify that the change affects campaigns only where a marketer wants someone to make a purchase, not for app download conversions. And to reflect that Meta will now give more weight to the conversions occurring one day from when a user first clicks.