A Blog by Jonathan Low


May 5, 2019

Is Ad Targeting Worth the Price (To Marketers And Society)?

Humans' ability to choose is increasingly being subsumed by marketers' desire to push consumers toward the choices that the companies doing the advertising want them to make.

Is that really good for civilizations - and even for marketers who will eventually bear the brunt of societies' inevitably disappointed expectations? JL

Stuart Thompson reports in the New York Times:

Nearly every ad you see online is tailored just for you. These digital ads are powered by vast, hidden datasets that allow advertisers to make eerily accurate guesses about who you are, where you’ve been, how you feel and what you might do next. Data providers can receive information from almost every imaginable part of your life. “Surveillance companies are learning to nudge, coax, tune and herd human behavior in the direction that they want people to go for the sake of the commercial outcomes that their business customers seek.”
Nearly every ad you see online is tailored just for you. These digital ads are powered by vast, hidden datasets that allow advertisers to make eerily accurate guesses about who you are, where you’ve been, how you feel and what you might do next.
While targeted ads may be familiar by now, how they work — and the power they have — often seems invisible.
We decided to lift the veil on this part of the internet economy, so we bought some ad space. We picked 16 categories (like registered Democrats or people trying to lose weight) and targeted ads at people in them. But instead of trying to sell cars or prescription drugs, we used the ads to reveal the invisible information itself.
Here’s what one ad looked like. The underlined text below shows what the ad knows about the person who’s seeing it.
This was an experiment in how far digital advertisers go to collect and use information about every part of our lives for profit. But it’s also the story of how our information is used not just to target us but to manipulate others for economic and political ends — invisibly, and in ways that are difficult to scrutinize or even question. And it’s a warning sign about the real-world political risks that come from this sophisticated guessing game, which is played in billions of transactions each day.
“The way ads are targeted today is radically different from the way it was done 10 or 15 years ago,” said Frederike Kaltheuner, who heads the corporate exploitation program at Privacy International. “It’s become exponentially more invasive, and most people are completely unaware of what kinds of data feeds into the targeting.”
With that in mind, we want to share how we targeted these ads, what we learned, and why it might disturb you.
Targeted advertising was once limited to simple contextual cues: visiting ESPN probably meant you’d see an ad for Nike. But advertising services today use narrow categories drawn from a mind-boggling number of sources to single out consumers. (Like many publishers, The Times uses targeted advertising to find potential subscribers and readers.)
To build the ads for our experiment, we imagined some extremely specific targets and built profiles of those people. Then we chose 16 attributes that matched those profiles from a list of about 30,000 — a list that’s rarely seen by people outside the industry.
We could do this because many companies, like retailers and credit card providers, sell customer information to data companies. Most data providers declined to tell us where their data comes from or how they built their models, so the sources in the ads below come from the ad experts who helped us create the campaign. Our experiment would have been blocked on Facebook because the company bans most ads showing how you’ve been targeted.
Here are more sample ads that we bought as part of our experiment. (These ads, like all the ads shown in this story, aren’t live and don’t show data about you specifically.)But that’s just the start. Launching a political campaign — or trying to interfere with one? You could target potential political converts.
“In the next election, I think it is inevitable that every single voter will have been profiled based on what they have been reading, watching and listening to for years online,” said Johnny Ryan, the chief policy officer at Brave, a private web browser that allows users to block ads and trackers.
Companies can target by location, too. It’s possible to single out bankers on Wall Street or political aides on Capitol Hill. If you wanted to target any specific location, you could find an ad platform that could do it (but the accuracy would vary depending on the technology used).
It’s not limited to what advertisers know about your past behavior, either. Data providers claim they can predict things you didn’t directly disclose, including what you’ll do in the future. They don’t have your voting history, but they can still guess whether you’re a likely liberal or a likely conservative. They also say they know whether you’re planning a foreign vacation in the next year.
“Surveillance companies are learning to nudge, coax, tune and herd human behavior in the direction that they want people to go for the sake of the commercial outcomes that their business customers seek,” said Shoshana Zuboff, Harvard Business School professor emerita and author of “The Age of Surveillance Capitalism.”
Predictions about who people are and what they will buy offer the most promise to advertisers because they enable targeting nearly everyone, even those with little data available.
Acxiom, a major marketing data company, said it makes predictions by comparing specific people to a much larger group. For example, it might predict whether you’ll buy an S.U.V. by looking at people who did buy S.U.V.s, then checking whether you live in a similar region, or whether you’re around the same age and share similar interests. If you’re similar, they’ll put you in a bucket that says you’re highly likely to buy an S.U.V. And if you’re the one who’s buying the S.U.V., your characteristics might be used to refine the pool of potential future S.U.V. buyers.
“Those predictive models and those audiences are defined by the simple principle of ‘birds of a feather flock together,’” said Chad Engelgau, chief strategy and marketing officer at Acxiom. He added that prediction categories are used when data companies don’t have more specific data, so they might guess the likelihood that you’re a parent who needs to buy diapers using a predictive category rather than learning directly by collecting new birth records.
“Nobody in the industry is saying this is a fact,” he said of the predictions. “It’s used to create efficiencies.”
[As technology advances, will it continue to blur the lines between public and private? Sign up for Charlie Warzel’s limited-run newsletter to explore what’s at stake and what you can do about it.]

The accuracy of predictions made by data providers is difficult to verify. The companies release little evidence that those included in these groups actually belong there. A study from 2018 found that the gender assigned by data brokers was accurate, on average, only 42 percent of the time — that’s worse than just flipping a coin. So in the ads we bought, we expected many men would see ads aimed at women and vice versa.
Acxiom said it’s working on adding more transparency to its models, including better labeling and third-party audits. It has also joined many other data companies calling for federal regulation that would place limits on how data is collected and shared while leaving room for marketing purposes. But such a law could also weaken or replace much more aggressive state regulations, like California’s data protection law.
Data providers also say the information stored and shared is anonymous, but that doesn’t mean it stays that way. In 2017, for example, German researchers tied data that was supposedly anonymous back to specific individuals. They found revealing things like the medication used by a German politician and a German judge’s use of pornography.

Where the data comes from

Today’s data providers can receive information from almost every imaginable part of your life: your activity on the internet, the places you visit, the stores you walk through, the things you buy, the things you like, who your friends are, the places your friends go, the things your friends do, and on and on. One provider boasts about having “precise second-by-second viewing of tens of millions of televisions in all 210 local markets across the country.” If you responded to an online event invitation, the details could be used to target you. If you’re listening to energetic music, you could be targeted almost instantly in an energetic mood group. Much of this data could be shared among brands.
You might find targeted ads preferable to general ones since there’s a better chance that you might actually want to buy the products being marketed. But privacy advocates say targeting has gotten out of control.
“Do you really need to track people on all the websites they visit, every single app they use, on their entire platform, and keep the data for a shockingly long time?” asked Ms. Kaltheuner, from Privacy International. “Is all of this really necessary to show relevant ads?”
Ads in our experiment combined data from many of these sources.
Just by browsing the web, you’re sending valuable data to trackers and ad platforms. Websites can also provide marketers specific things they know about you, like your date of birth or email address.
Ad companies often identify you when you load a website using trackers and cookies — small files containing information about you. Then your data is shared with multiple advertisers who bid to fill the ad space. The winning bid gets to fill the ad slot.
All of this happens in milliseconds.
Dr. Zuboff, the author and professor, argues that companies today are mining private lives the same way they exploit natural resources, turning them into profitable goods. She said that makes every “smart” device and “personalized” service just another way to collect data for the surveillance economy.
That change happened so quickly and so secretly, Dr. Zuboff argues, that the public is not equipped to fight back.
Ms. Kaltheuner agreed. “It’s about ads, but it has all these other implications that have nothing to do with ads,” she said. “And that’s the price we’re paying to see ‘relevant’ ads.”


Post a Comment