Apple will soon require consumers to opt in if they want to allow businesses to track their data and use it for personalized advertising. Facebook is fighting this decision with an aggressive ad campaign, citing evidence that the decision will hurt small businesses. But that evidence turns out to be false, as Facebook surely knows.
This article has been updated.
On January 28, at the Computers, Privacy, and Data Protection conference, Apple’s CEO, Tim Cook, spoke about the importance of giving users more control over how their data is used for online advertising. To that end, Apple’s new operating system will soon require consumers to opt in if they want to allow businesses to track their data and use it for personalized advertising.
Facebook, which relies heavily on personalized advertising, responded fiercely to this decision, placing full-page ads in the New York Times, the Wall Street Journal, and the Washington Post that accuse Apple of hurting small businesses. Facebook has also created a website where small businesses can voice their concerns about Apple’s decision.
Facebook’s central claim is that small businesses will lose revenues if they can’t use personalized ads. “Without personalized ads,” the company says in its ads and on its website, “Facebook data shows that the average small business advertiser stands to see a cut of over 60% in their sales for every dollar they spend.” It’s an eye-popping figure, and one that suggests that Apple’s pro-privacy policy is poised to deal a devastating blow to small businesses. But where does the data for this apocalyptic claim come from? And does it hold up under scrutiny?
To properly evaluate this claim, you first need to understand the popular metric that Facebook used here to quantify advertising success: return on ad spend, or ROAS. The metric indicates the amount of revenues associated with advertising — but it does not indicate the amount of revenues caused by advertising.
To understand why this difference matters, imagine a company that knows its customers very well. It can predict with a high degree of accuracy how much a customer will spend in the coming month. If the company targets its advertisements to those customers who are expected to spend a lot, each dollar spent on advertising will be associated with high revenues. That’s great — the company has achieved a high return on ad spend. But here’s the thing: These customers would have generated high revenues anyway. That’s why they were targeted in the first place. So it would be a mistake to conclude that these customers spent more because of the personalized ads.
In its campaign against Apple’s new policy, Facebook is claiming that when it compared the ROAS for campaigns that leveraged personalized information with campaigns that didn’t, it found that small businesses would suffer a 60% cut in revenues if they were deprived of personalized advertising.
That scary-sounding number, however, does not say much about how Apple’s privacy policy will impact the revenues of small businesses.
After publication, Facebook shared more details about their methodology.* They split more than 25,000 campaigns across many industries and regions into two groups. Campaigns in the first group were optimized for link clicks, which does not require data sharing. Campaigns in the second group were optimized for purchases. This requires data sharing, so Facebook calls these personalized campaigns. Of course, the second type of campaign will be associated with higher revenues per advertising spend. Facebook’s algorithms were instructed to find people likely to buy. That doesn’t mean they were able to increase people’s likelihood to buy.
Without randomized controlled trials, Facebook should refrain from making claims that imply large effects of personalized ads. Randomized controlled trials that compare personalized advertising with no advertising suggest only a small impact on sales.
That isn’t the only problem with Facebook’s argument.
According to Facebook, Apple’s decision is especially damaging during this pandemic, because, as Facebook’s ads and website state, “Forty-four percent of small to medium businesses started or increased their usage of personalized ads on social media during the pandemic, according to a new Deloitte study.”
We took a close look at the Deloitte study. While the statistic used by Facebook might be vivid, it is hardly informative. In its study, Deloitte surveyed companies from nine industries across seventeen countries. It sure sounds impressive to say that 44%** of small businesses in the U.S. started using, or increased their usage of, targeted advertising. But let’s put this number into context.
First, it implies that 56% of businesses did not increase their spending on targeted advertising. The majority of businesses said they had never used targeted advertising, stopped using it during the pandemic, decreased their usage, or that their usage stayed the same.
Second, it says very little about how much businesses actually spent on digital advertising. Even a relatively small increase in advertising expenditures (of say 5%) is likely to be spread over a large number of firms that increased their spending by some amount.
Third, in the same table, just below the 44%, Deloitte reports the share of small businesses that started using or increased their usage of non-personalized ads. That number (35%) is not much lower than the one underlined by Facebook, suggesting there’s nothing special about personalized advertising in pandemic times.
Facebook, in our view, cherry-picked the data they thought best supported its case, and the cherry wasn’t even good.
In pointing this out, we don’t mean to dismiss the concerns that many small businesses have about the changes Apple is making to its privacy policy. They’re real: Under Apple’s new plan, companies will have to explain their data-collection practices when submitting new apps or making updates, and many users won’t give permission to have their behavior tracked online. Facebook says it wants to stand up for small businesses in the face of these changes, which it is perfectly entitled to do. But disinformation about advertising effectiveness isn’t the way to do that.
Editor’s note: After the original version of the piece was published, Facebook reached out to us and the authors to clarify its methodology, add context, and correct an error. After consulting with us, the authors have made changes to analyze that methodology* and correct that error**.