Advertising campaigns used to be planned and managed by media buyers—usually 22-year-old, newly graduated communications majors. If that media buyer needed to help a car manufacturer reach men looking to buy a car, she might place an ad in Car and Driver, or in the automotive section of the newspaper. Advertising used to be something you could place, count, then see in the front cover spread of a magazine.
But this is not digital advertising today. Digital advertising is automated, data-driven, and opaque in its mechanics. That 22-year-old communications major has had to make way for data scientists, mathematicians, and computer programmers who, behind the scenes, use statistics, calculus, and linear algebra to optimize advertising campaigns, by micro-targeting users and constantly tweaking algorithms.
Does that car manufacturer still want to reach men looking to buy a car? A data scientist may tell them the optimal target is a 39-year-old man, carrying on an extramarital affair, who’s on the brink of divorce. They can model this hypothesis (and prove it works), and advertising companies like Google and Facebook can put that into execution, finding ways to home in and target those types of people online.
When you go to a website and load a page, in the milliseconds that it takes for that page to load, there are real-time auctions running in the background that determine which ads to load on your page. Almost all online ads are delivered in this way, where highly complex auction markets make their money by competing on who can better track users and invade their privacy more thoroughly.
The targeting begins the moment you as a reader visit any website. Typically, your IP address, your location, and the URL of the page you are on are swiped from your browser without your explicit knowledge, and shared with advertising companies that run these ad auctions. The goal, of course, is to build as specific a portrait about you as possible—by linking your device with your identity—and cookies are a common tool for doing so.
A “cookie” is a small text file that a site can install on your computer when you visit. The text file fingerprints your device with a unique identifier, or “cookie ID” (such as 12345qwert). If the website knows your real identity (for example, if you log on to the site with your real name), the company can link it to your cookie (here, 12345qwert) and begin to gain an advantage in determining which ads to load onto your page.
For example, if you’re on the hypothetical URL newspaper.com/how-to-fight-melanoma, this probably means you’re reading an article about melanoma. Companies might use that information to make a prediction about whether you or someone you love may have cancer. And they most certainly use that info to determine which ads to load onto your page.
The prices that any company is able to fetch for its ads depend on two crucial factors: the ability to identify who is loading the page, and the ability to then connect the user’s identity with more information about the user.
Imagine a person visits espn.com to read an article about the upcoming Super Bowl. Assume first he doesn’t log on to the site, and blocks his browser cookies, so maybe the website he is visiting can’t know who he really is. An advertiser can nonetheless bid on the opportunity to display an ad to this anonymous reader. Maybe the slot goes to a beer brand that wants to generally reach people who like football. Perhaps the going price is a $2 CPM (cost per thousand) and the ad gets sold at this price (meaning, this is the clearing auction price).
But you’re not usually anonymous when you’re online, even when you think you are. Again, advertising companies might know your identity because you log in, or because you are using a browser that allows tracking. Now it’s not simply an anonymous person loading a page about the Super Bowl, it’s “Michael Greenberg,” of Wichita, Kansas.
Now, companies can combine Michael’s identity with other commercially available datasets in real time. For example, they might stitch Michael’s identity with the fact that he makes $1 million-plus per year, which means that they can match Michael with an ad for a private jet service instead of a Bud Lite. The private jet ad might sell at a $200 CPM as opposed to the $2 CPM beer ad targeted to an anonymous user.
“The exact same ad, on the same website, at the same time, could be worth vastly different amounts to two different buyers depending on how much they know about the consumer being targeted,” explains Ari Paparo, now founder and CEO of advertising company Beeswax and a former Google exec. “User data is everything.”
Advertisers gain an even better advantage when they’re able to track what users do as they move from site to site, app to app, site to app, and vice versa, which is exactly how Facebook and Google operate (and exactly the type of information traditional publishers don’t have).
If a company that sells online ads can know what their readers are reading on other sites, then they can target the users based on that information when the user returns to their own site. For example, say Michael visits CNBC’s website in the mornings and reads about the markets, but visits The New York Times in the evenings and only reads the book review section. CNBC knows Michael is someone who follows the markets, and might monetize his view at a $30 CPM. The Times knows that Michael is someone who likes to read books so might only monetize Michael at a $10 CPM. If the Times can somehow find out that Michael is reading CNBC in the mornings, then when Michael visits the Times book section in the evening, the Times can target him as someone who follows the markets and monetize him at $30, too.
Would CNBC want to share with the Times what Michael reads on cnbc.com? Of course not. The two are competitors on the advertising side of the market. If CNBC is selling its audience of financial readers at a cost of $30, and the Times can copy CNBC’s readers and their reading patterns, then the Times could theoretically undercut CNBC and sell ads targeted to CNBC financial readers for, say, $20 instead of $30.
But publishers like the Times and CNBC have no choice but to share this information with Facebook and Google. How, might you ask, does Facebook currently get this data from news publishers that are also advertising competitors? Well, Facebook has a number of derivative products that flow from the social network, including “Like” buttons and log-in tools. Facebook licenses Like buttons to publishers so that their readers can “like” and then “share” news stories across the Facebook social network. But Facebook now conditions these licenses on the ability to track publishers’ readers, whether the readers click the Like buttons or not, and Facebook can now use publishers’ reader data to sell its own ads.
Google, which now tracks users on over 70 percent of the top one million sites, also uses its ability to track users across the internet to extract an advantage in advertising markets. Google tracks users via its analytics and ad-serving products, which Google consolidated and rebranded last summer as the Google Marketing Platform. Google was actually the first of the two companies to consolidate products under a rubric of privacy.
The implication of all this is that the money that Google and Facebook can make selling advertising goes well beyond what other ad sellers can demand in the market. The Big Tech duopoly can track billions of users across millions of sites and mobile apps, creating longitudinal profiles on users. News publishers simply cannot compete with that kind of an informational advantage.
But there is another thing going on in these markets that explains the duopoly in the advertising market. When most people think about Google and Facebook, they think the companies make so much money by selling ads on their own properties—Google search, Gmail, the Facebook social network, Instagram, and so on. This is partly true. Google and Facebook also run auctions through which publishers now sell their own advertising.
Unlike in finance, there are several auction markets where digital ads trade. Anyone can create one. But Google and Facebook make sure their own advertising inventory (YouTube, Facebook) can only be bought through their own, proprietary auctions. Google made almost $20 billion last year from selling other companies’ ads. This is why Google today is the largest seller of advertising, globally, period.
by Dina Srinivasan, American Prospect | Read more:
But this is not digital advertising today. Digital advertising is automated, data-driven, and opaque in its mechanics. That 22-year-old communications major has had to make way for data scientists, mathematicians, and computer programmers who, behind the scenes, use statistics, calculus, and linear algebra to optimize advertising campaigns, by micro-targeting users and constantly tweaking algorithms.
Does that car manufacturer still want to reach men looking to buy a car? A data scientist may tell them the optimal target is a 39-year-old man, carrying on an extramarital affair, who’s on the brink of divorce. They can model this hypothesis (and prove it works), and advertising companies like Google and Facebook can put that into execution, finding ways to home in and target those types of people online.When you go to a website and load a page, in the milliseconds that it takes for that page to load, there are real-time auctions running in the background that determine which ads to load on your page. Almost all online ads are delivered in this way, where highly complex auction markets make their money by competing on who can better track users and invade their privacy more thoroughly.
The targeting begins the moment you as a reader visit any website. Typically, your IP address, your location, and the URL of the page you are on are swiped from your browser without your explicit knowledge, and shared with advertising companies that run these ad auctions. The goal, of course, is to build as specific a portrait about you as possible—by linking your device with your identity—and cookies are a common tool for doing so.
A “cookie” is a small text file that a site can install on your computer when you visit. The text file fingerprints your device with a unique identifier, or “cookie ID” (such as 12345qwert). If the website knows your real identity (for example, if you log on to the site with your real name), the company can link it to your cookie (here, 12345qwert) and begin to gain an advantage in determining which ads to load onto your page.
For example, if you’re on the hypothetical URL newspaper.com/how-to-fight-melanoma, this probably means you’re reading an article about melanoma. Companies might use that information to make a prediction about whether you or someone you love may have cancer. And they most certainly use that info to determine which ads to load onto your page.
The prices that any company is able to fetch for its ads depend on two crucial factors: the ability to identify who is loading the page, and the ability to then connect the user’s identity with more information about the user.
Imagine a person visits espn.com to read an article about the upcoming Super Bowl. Assume first he doesn’t log on to the site, and blocks his browser cookies, so maybe the website he is visiting can’t know who he really is. An advertiser can nonetheless bid on the opportunity to display an ad to this anonymous reader. Maybe the slot goes to a beer brand that wants to generally reach people who like football. Perhaps the going price is a $2 CPM (cost per thousand) and the ad gets sold at this price (meaning, this is the clearing auction price).
But you’re not usually anonymous when you’re online, even when you think you are. Again, advertising companies might know your identity because you log in, or because you are using a browser that allows tracking. Now it’s not simply an anonymous person loading a page about the Super Bowl, it’s “Michael Greenberg,” of Wichita, Kansas.
Now, companies can combine Michael’s identity with other commercially available datasets in real time. For example, they might stitch Michael’s identity with the fact that he makes $1 million-plus per year, which means that they can match Michael with an ad for a private jet service instead of a Bud Lite. The private jet ad might sell at a $200 CPM as opposed to the $2 CPM beer ad targeted to an anonymous user.
“The exact same ad, on the same website, at the same time, could be worth vastly different amounts to two different buyers depending on how much they know about the consumer being targeted,” explains Ari Paparo, now founder and CEO of advertising company Beeswax and a former Google exec. “User data is everything.”
Advertisers gain an even better advantage when they’re able to track what users do as they move from site to site, app to app, site to app, and vice versa, which is exactly how Facebook and Google operate (and exactly the type of information traditional publishers don’t have).
If a company that sells online ads can know what their readers are reading on other sites, then they can target the users based on that information when the user returns to their own site. For example, say Michael visits CNBC’s website in the mornings and reads about the markets, but visits The New York Times in the evenings and only reads the book review section. CNBC knows Michael is someone who follows the markets, and might monetize his view at a $30 CPM. The Times knows that Michael is someone who likes to read books so might only monetize Michael at a $10 CPM. If the Times can somehow find out that Michael is reading CNBC in the mornings, then when Michael visits the Times book section in the evening, the Times can target him as someone who follows the markets and monetize him at $30, too.
Would CNBC want to share with the Times what Michael reads on cnbc.com? Of course not. The two are competitors on the advertising side of the market. If CNBC is selling its audience of financial readers at a cost of $30, and the Times can copy CNBC’s readers and their reading patterns, then the Times could theoretically undercut CNBC and sell ads targeted to CNBC financial readers for, say, $20 instead of $30.
But publishers like the Times and CNBC have no choice but to share this information with Facebook and Google. How, might you ask, does Facebook currently get this data from news publishers that are also advertising competitors? Well, Facebook has a number of derivative products that flow from the social network, including “Like” buttons and log-in tools. Facebook licenses Like buttons to publishers so that their readers can “like” and then “share” news stories across the Facebook social network. But Facebook now conditions these licenses on the ability to track publishers’ readers, whether the readers click the Like buttons or not, and Facebook can now use publishers’ reader data to sell its own ads.
Google, which now tracks users on over 70 percent of the top one million sites, also uses its ability to track users across the internet to extract an advantage in advertising markets. Google tracks users via its analytics and ad-serving products, which Google consolidated and rebranded last summer as the Google Marketing Platform. Google was actually the first of the two companies to consolidate products under a rubric of privacy.
The implication of all this is that the money that Google and Facebook can make selling advertising goes well beyond what other ad sellers can demand in the market. The Big Tech duopoly can track billions of users across millions of sites and mobile apps, creating longitudinal profiles on users. News publishers simply cannot compete with that kind of an informational advantage.
But there is another thing going on in these markets that explains the duopoly in the advertising market. When most people think about Google and Facebook, they think the companies make so much money by selling ads on their own properties—Google search, Gmail, the Facebook social network, Instagram, and so on. This is partly true. Google and Facebook also run auctions through which publishers now sell their own advertising.
Unlike in finance, there are several auction markets where digital ads trade. Anyone can create one. But Google and Facebook make sure their own advertising inventory (YouTube, Facebook) can only be bought through their own, proprietary auctions. Google made almost $20 billion last year from selling other companies’ ads. This is why Google today is the largest seller of advertising, globally, period.
Image: Mark Lennihan/AP Photo
