Edmon de Haro
Online data privacy is a hot-button political issue, and a handful of data-restrictive policies have been introduced to address it. In the European Union, the General Data Protection Regulation—in effect since May 2018 and commonly known as GDPR—requires users to give explicit consent for their data to be used for targeted ads. In 2021, Apple rolled out its “Ask App Not to Track” feature, which allows iPhone users to set their default to not sharing data with mobile apps, requiring them to opt in for tracking to occur.
Walling off consumer data from online advertisers may cost businesses billions of dollars, according to Facebook’s Nils Wernerfelt, Northwestern’s Anna Tuchman, Chicago Booth’s Bradley Shapiro, and Facebook’s Robert Moakler. Without access to browsing history, past purchases, shopping-cart contents, and other detailed data, the median digital advertiser would have to spend 37 percent more to gain new customers, the researchers calculate.
The findings come from a model they built using data from about 107,000 global advertisers on Facebook to measure the costs of attracting customers with or without the use of data. Online advertising would suffer without detailed consumer data, and tougher barriers would disproportionately hurt smaller companies selling niche products, according to the findings.
“Targeted advertising works by finding someone it would be appropriate to send an ad to—people who have not yet bought the product but who would be interested in it,” Shapiro says. “The more niche the product, the harder it is to find those people, so companies need better data to find them. If you take some of that away, it’s going to be harder to find those appropriate customers.”
The researchers set out to determine the effectiveness of targeted ads using detailed consumer data from Facebook, the bulk of whose revenue comes from a process known as offsite conversion optimization. Retailers and other platforms share purchasing data with Facebook, which combines all the information it receives and uses a prediction model to identify similarities between consumers. It’s able, for example, to receive data from a shoe store about one shopper’s purchases, then find other people on Facebook who appear similar to that shopper and might also be interested in buying a pair of shoes. This is how Facebook can send out targeted ads for a product even to potential customers who didn’t indicate that they were in the market for it.
The researchers find that offsite conversion optimization has a median cost per incremental customer of $43.88, which is relatively low, Shapiro says. (Each new customer acquired solely because of advertising is considered incremental.) Over a week’s time, advertisers spend a median of $1,259 on Facebook ads and net about 29 new customers.
The researchers then modeled an alternative scenario without the offsite data required to do conversion optimization, instead assuming that Facebook only has access to onsite information (collected entirely on web addresses that it owns) to train its prediction model, in particular whether someone clicked on an ad on Facebook. They find a pricier median cost per incremental customer of $60.19. If advertisers kept their spending constant, they’d generate only 21 new customers a week without the data, the researchers find.
The loss of access to these data would most likely have other effects. Because Facebook ads would become less effective at generating sales, some advertisers would opt to leave the platform altogether—probably smaller companies or those selling niche products—and the value of Facebook ads would drop. Facebook users would likely see more ads from big companies selling products with wide appeal (Coca-Cola, for example), and fewer ads from smaller businesses selling less-popular items, Shapiro says.
This understanding of data’s value to advertisers should add more nuance to conversations around potential future regulation, the researchers argue. One possible path involves giving individual consumers ownership and control over their own data. Considering how valuable the information is to advertisers, maybe they’d be willing to pay consumers for it, the researchers suggest.
“When you hear Congress talk about this question, they completely leave out how much value these data bring,” Shapiro says. “Usually, the option they discuss is to shut it all down and not allow any data sharing, but they don’t come up with other alternatives. Adding our research allows us to think about a broader range of policy levers they might pull.”
Nils Wernerfelt, Anna Tuchman, Bradley Shapiro, and Robert Moakler, “Estimating the Value of Offsite Data to Advertisers on Meta,” Working paper, August 2022.
Chicago Booth’s John Paul Rollert looks for a balance between honoring customers and indulging them.
Is the Customer Always Right?
An experiment demonstrates that officers can learn to apply critical thinking in stressful situations, reducing the use of force and discretionary arrests.
How to Redesign Police Training to Reduce the Use of Force
Lars Peter Hansen and Kevin M. Murphy discuss how data can inform policymaking.
A Nobel Laureate on the Limits of Evidence-Based PolicyYour Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.