Facebook has been selling itself short.
A bug in Facebook’s measurement system caused Facebook to undercount app install campaigns on iOS 14 via SKAdNetwork (SKAN) for nearly seven months.
The bug, which Facebook identified on August 20, only impacted iPhone 12 users that were using the Facebook app. Instagram users and Audience Network partner apps were unaffected, a Facebook spokesperson told AdExchanger.
But still, that could be tens of millions of the most desirable consumers out there.
Global sales of iPhone 12 handsets surpassed 100 million units as of April, according to market research and consulting firm Counterpoint Research.
Eric Seufert, editor of Mobile Dev Memo, first called attention to the bug on Monday, which is also when Facebook sent an email to advertisers and tech partners alerting them of the issue.
Umm, isn’t that a big deal?
— Ari Paparo (@aripap) August 30, 2021
In the email, Facebook said the undercounting started when it started running iOS 14 app ad campaigns in February.
“During the affected period,” Facebook wrote, “SKAN app ads were being delivered and people were potentially converting, but we were unable to measure those conversions since we were not receiving the postbacks from SKAN for iPhone 12 users.”
Specifically, the issue affected the measurement of SKAdNetwork app ad conversions for three campaign types on Facebook: app event optimization (targeting people who are likely to take a specific action), value optimization (targeting people based on how much they are likely to spend in a given amount of time) and mobile app installs (targeting people with the goal of getting them to install an app).
Facebook said the bug has since been fixed, but it led to undercounting for around 10% of total SKAdNetwork conversions.
“The impact on specific campaign metrics will vary,” the Facebook spokesperson told AdExchanger. “We are communicating with advertisers whose metrics may have been impacted to make sure they are aware.”
In the email they received, advertisers were told they may notice an increase in measured app ad conversions for iOS campaigns now that the bug has been fixed.
But what’s lost is lost. The fact is, for at least half a year, Facebook was undercounting paid attributions for the wealthiest, most lucrative owners of iOS hardware.
Talk about shooting yourself in the foot. Facebook underreported the impact of its own paid media on “the very cream of the cream – the highest net worth owners on the most profitable mobile platform,” said mobile industry analyst John Koetsier.
“This certainly had an impact on where advertisers invested growth dollars,” Koetsier said.
But there was no way for anyone outside Facebook to validate what was happening as it was happening, said Shamanth Rao, CEO of mobile user acquisition startup RocketShip HQ. Rao noted that “Facebook doesn’t share postbacks with advertisers or MMPs [mobile measurement partners].”
To be fair, though, SKAdNetwork implementation has been a slog for many players across the mobile industry. For one, SKAdNetwork has experienced bugs of its own, which made early testing a challenge, and the way the technology itself functions is quite confusing, including the privacy threshold and the conversion value timer.
Advertisers only have access to a limited amount of conversion data through SKAdNetwork, and that doesn’t include any user- or device-level data. The sparse amount of shared data, which Apple restricts on purpose in the name of protecting user privacy, admittedly makes it rather tough to fix or even find a bug like the one Facebook just uncovered, Koetsier said.
But it also begs a rather big question, he said: “What other bugs in mobile advertising reporting on iOS are we missing simply because the data to vet the results simply doesn’t exist?”