Case: Capital One’s Online Profiles

In 2010, Capital One Financial Corporation began using special software to create instantaneous profiles of visitors to its website. Constructed from information such as recent purchases, web browsing history, and geographic location, these profiles were used mainly to determine which credit card offers to display on a visitor’s computer screen.⁷⁰⁸

Customer Profiles

In the case of one customer, Carrie Isaac, Capital One’s website used “cookies” left by other websites, her Internet Protocol (IP) address, and other technical information transmitted by her computer to conclude that she was a member of the “White Picket Fences” group, a profile for customers who are thought to be middle-class parents who live in a metropolitan suburb and have reliable creditworthiness. Capital One used sophisticated algorithms to determine correctly that she was female and a young parent and that she earned approximately $50,000 annually, had attended, and shopped at discount department stores. On the basis of this information, Capital One’s software displayed a credit card designed for people of average creditworthiness with no annual fee and an initial monthly interest rate of zero percent, increasing to 19.8 percent thereafter. Overall, Capital One’s inferences about Ms. Isaac’s identity were accurate.

The same appeared to be true of another potential customer, Paul Boulifard. Capital One’s website focused on Mr. Boulifard’s residence in Nashville, Tennessee, and his interest in travel. It displayed the “VentureOne Rewards” credit card to him, which allows the accumulation of points that can be used to purchase airline tickets. The images surrounding this card included a beach scene and the slogan “Still Searching? Get double miles with Venture.”

In the case of Karyn Morton, however, Capital One’s software was less accurate. Ms. Morton was profiled as a member of the “City Roots” segment. Capital One accurately determined that she was a homeowner living in Detroit, a member of the National Association for the Advancement of Colored People (NAACP), and a regular reader of major newspapers. It inaccurately inferred that Ms. Morton was retired without children, had little education, and was living on a modest income of $28,000. She actually earned three times that amount, was 33 years old, and held a law degree. Capital One offered Ms. Morton two credit card options, one for individuals with average credit scores and an interest rate of 24.9 percent and one for customers with excellent credit scores and an interest rate of 13.9 percent.

Use of Profiles

Capital One emphasized at the time that it did not use the information gathered in a visitor’s online profile to determine who actually received certain lines of credit. It used only the concrete information voluntarily offered by a customer on a credit application for such purposes. Capital One, therefore, did not violate the Equal Opportunity Credit Act, a federal law that prohibits banks and other lenders from targeting or restricting financial services based on race, ethnicity, national origin, or residency.⁷⁰⁹ Capital One claimed that it simply made an “educated guess” about what it thought customers would want and featured products based on those inferred preferences.⁷¹⁰

Capital One’s efforts at product placement were not unique. Other online retailers have used similar methods in setting online prices.

• In 2012, Orbitz, the online travel site that provides low-priced deals on car rentals, hotel rooms, and airfares, offered the same products to different customers at different prices.
• Customers who used desktop computers with an Apple operating system paid 30 percent more for hotel rooms compared with customers who booked the same rooms using computers with a Microsoft operating system.⁷¹¹
• The office supply giant Staples has sold products at different prices depending upon a customer’s proximity to competitors’ stores. A recent investigation found that the Staples.com website displayed different prices to different people by “estimating” their location based on their computer’s IP address. Staples considered the distance from a competitor’s store, such as OfficeMax or Office Depot, and if a store was located within 20 miles, then a discounted price was shown.⁷¹²

Profiling Technology

Capital One arguably refined a common practice. Marketing decisions involving product placement and pricing have long been guided by the concept of “segmentation.” The marketplace is composed of groups of customers—or segments—with different experiences, demographic traits, and preferences. The rise of information technology and e-commerce has enabled marketers to modify the manner in which they sell products based on their knowledge of the segment to which a potential customer belongs. Segments provide a useful, if imperfect, guide to quickly predict a customer’s likely purchases.

Capital One’s software was engineered by a little-known supplier, [x+1], Inc. Neither this fact nor the exact methods employed by the profiling software were disclosed to visitors on the website. Capital One did disclose that it collected and used visitors’ IP addresses, browser and operating system information, “cookies” placed by other websites, navigation preferences, social media activity, and geographic data. These disclosures, however, were placed within the “privacy” section of Capital One’s website, located at the bottom of the user’s screen in small font. This is typical in the online commercial environment. Internet users are rarely cognizant of how they are being profiled, and privacy disclosures are not easy to find without some effort.⁷¹³ Users also expect their online activity to take place in a market that provides impersonal, even anonymous, interaction. This expectation is apparently important to Internet users. Marketing studies⁷¹⁴ indicate that consumers typically find product and price customization problematic when there is a lack of transparency regarding the customization efforts. When consumers expect standardized sales experiences, customized experiences are considered unfair, but if there is an expectation that product offers or prices will differ between consumers, then variations are perceived as less problematic.⁷¹⁵

Capital One’s algorithms were focused exclusively on the information that could be gleaned from visitors’ computers at the moment that they started using Capital One’s website. More advanced technology exists, however, which can combine the up-front data provided by a visitor’s computer, web browser and IP address with larger sources of data that contain historical records of Internet transactions, in-person retail purchases, and e-mail addresses.⁷¹⁶ This technology could conceivably enable customer profiling that combines online with offline behavior. It also holds the prospect of eliminating anonymity in Internet transactions. As more data, such as ZIP codes, telephone numbers, birth dates, e-mail addresses, and online social activities, are accessed and used by online advertisers, the accuracy with which companies can place a customer within a segment, or even construct a concrete identity profile, is increased. This capability would expand and refine the ability of companies like Capital One to customize experiences for each consumer.