Stanford CIS

Does E-commerce Personalization Make All Consumers Vulnerable Consumers?

By Darren Stevenson on

It is the first week of March, which in the U.S. means it is also National Consumer Protection Week. This coordinated annual effort encourages people to assert their rights in the marketplace and make better informed consumer decisions. While you may have never heard of National Consumer Protection Week, given that you may also have recently paid more when shopping online for a hammer or hotel room based on your computer or smartphone's operating system, browser history, or other information marketers know about you, perhaps you are in need of some consumer protection yourself. Yes, even you savvy web user, you too might be a vulnerable consumer in light of emerging 'personalization' strategies used by firms.

Achieving Personalization

It comes as no surprise that the marketing industry is one of the largest constituents of digital personal data. One 2013 study by professors from Harvard Business School and Columbia University estimated in the U.S. alone marketing activities linked to individual-level consumer data contributed over $150 billion to the national economy and supported 600,000 jobs. By most measures, the business of monetizing personal information is booming.

Advances in online user tracking are largely made possible by the industry's ability to attract some of the most talented scientists, engineers, and statisticians of our time. As mathematician, early Facebook employee, and now venture capitalist Jeff Hammerbacher lamented in 2011, "The best minds of my generation are thinking about how to make people click ads. That sucks." The combined efforts of many genuinely brilliant people is the seamless tracking of behavioral data that occurs on most online platforms today. These data are collected by proprietors of websites and apps, marketers, and information middlemen unaffectionately referred to as data brokers.

Additionally, this information describing individuals is continually bought, sold, rented, and traded between these firms and put to a variety of uses. Most of this activity occurs under the hood, unobservable to the person whose information is in transit. As a result, rather than most people seeing the same advertisements and product prices as in traditional television commercials or newspapers, dynamic content personalization means individuals are routinely exposed to online marketing that has been customized, to some degree, using information about the viewer.

To achieve web content personalization, marketers observe and track internet users in ways some people may find surprising. This ranges from studying and quantifying overt text communications to collecting less explicit behavioral data. For example, sending messages in private channels (e.g. email), on mixed public/private platforms (e.g. Facebook post), and in public outlets (e.g. online Q&A forum), each of these 'transactions' generate quantifiable marketing information. This information feeds into the larger pool of consumer data ultimately used for audience segmentation and content personalization.

Similarly, additional forms of online communications contribute to this band of consumer profile data, often derived from single mouse clicks and touchscreen taps rather than user-supplied text. Examples include upvotes/downvotes, 'likes', and selecting star ratings for particular digital or physical products. The most common form of tracking these less overt communications occurs when firms monitor the series of websites an individual visits and in which order, data commonly referred to as one's 'clickstream'.

Additionally, activities occurring offline are also incorporated into the mix and used to tailor online content. This includes integrating behaviors like brick-and-mortar credit card purchases or those made with cash but accompanied by use of a customer loyalty card, for instance. These offline data are then integrated with online content personalization systems, used to further enhance personalization across internet-connected devices.

Of course, the concept of individual-level consumer profiling is nothing new, growing out of database marketing and direct mail techniques pioneered in the 1980s. Further, firms have flirted with dynamic online pricing since the early days of the web. However, under recent increases in the availability of digital personal information, which has become not only more granular but also faster, cheaper, and more widely available to marketers, presenting consumers with customized experiences is becoming more feasible and sophisticated. Consequently, marketing efforts are increasingly automated, driven by algorithms that rely on these varied forms of personal information as inputs. The result is a personalized online media space where no two individuals experience the same ads, product discounts, recommendations, and prices.

Consenting to Personalization

Oftentimes, internet users consent to this type of data collection and corresponding personalization processes in a single click of the mouse (or tap of the touchscreen). Legally, by extending one's index finger a few millimeters an individual expresses he or she has read, understood, and agreed to a multitude of privacy policies and terms of service. Typically, these agreements specify that firms are granted rather wide access to collect, sell, rent, trade, and otherwise use information collected about users. For many web users, issuance of these legally binding online agreements occurs so often they fail to gain much attention or consideration.

Of course, few people read much less fully grasp privacy policies and terms of service agreements. These documents tend to be lengthy and written in legalese, alienating readers while obfuscating third-party data sharing agreements. A recent study estimated it would take the average internet user around 200 hours each year to read the privacy policies of all the websites they visit in that same year. This process would require repeating annually given new sites visited and updated policies. It is safe to say most individuals do not take a month off of work each year to read the thousands of arcane privacy policies specifying the formation and whereabouts of their data doppelgangers.

Further, these legal agreements often arrive under a disclosure paradigm that assumes an informed and rational participant or consumer. Criticism of this notice and consent regime is common. Critiques include challenging the way this paradigm favors service providers over users by placing unrealistic expectations on the recipient to evaluate technical matters for issues of fairness.

Additionally, notice and consent becomes particularly ineffectual under third-party data sharing arrangements, where consumers are simply notified that their information may be shared with "partner organizations" or "affiliates" but not precisely how these third-parties will use their information. Those who consent to terms of service agreements are rarely cognizant of how third-party data sharing works nor the ways marketers connect disparate sources of information to triangulate consumers.

In addition to more active consenting via checkboxes, as in the case of creating various online accounts (email, social media, coupon sites, etc.), consent by use is even more common. Just as it sounds, consent by use occurs in most cases when simply visiting a website or using a pre-installed app on smartphone or tablet. This type of always-on consent occurs on most commercial web properties, for instance search engines, news, and entertainment sites, where the act of landing on a proprietor's webpage grants the owner necessary permission to conduct a range of online monitoring and information gathering practices.

Personalized Harm

Admittedly, this 'harm' incurred by digital personalization may appear less dire when compared to something like protecting consumers from telephone scams and damaging forms of identity theft. Yet the subtle effects of sophisticated data-driven content personalization may hold widespread consequences for the ability of consumers to make informed purchasing decisions. Discerning whether a website is displaying a higher price due to one's inferred household income or whether or not they are a parent, for instance, is largely impossible for consumers. Further, given the opacity under which these marketing practices function, implications may far exceed current levels of attention and related perceived needs for consumer protection.

To better appreciate this danger, it helps to consider both current practices and the likely trajectory of industry developments. Given recent advances, including the application of machine learning and predicative analytics to deliver targeted content, we may be approaching a critical moment in the emergence of subtle yet sophisticated forms of data-driven exploitation.

In practice, these harms might range from personalized political ads that are a bit too persuasive to objectionable forms of first- and third-degree price discrimination. The problem is captured by University of Washington Law Professor (and CIS Affiliate) Ryan Calo in his analysis of the potentials for so-called 'digital market manipulation.' A term referring to the process where firms use known or estimated information about individuals to manipulate purchasing decisions, primarily by leveraging information asymmetries to exploit consumers' cognitive limitations. Under advances in data-driven personalization to support dynamic marketing content, potentials for this type of undue influence loom large.

Consumer Protections to the Rescue?

In calling attention to these practices and the potential harms for consumers, the hope is not to raise a red flag unaccompanied by solutions. While current conditions make it virtually impossible for consumers to determine when e-commerce leans in their favor, a number of recent activities in the U.S. signal that help may be on the way.

In the past few years, many state and federal lawmaking and regulatory bodies have dedicated attention to how digital data is used to influence consumer decision making. In September, the FTC convened a workshop to assess the effects of big data on consumers while specifically examining implications for low income and underserved populations. The workshop came in a larger series of events and reports by the FTC on consumer privacy. Similarly, a recent U.S. Senate Commerce Committee's investigation spanning multiple years paid particular attention to the questionable practices of data brokers, garnering attention from the marketing industry and privacy groups alike. Similar recent reports and updates from the White House attest to growing concerns at the federal level regarding social impacts of big data, personalization, and potentials for consumer harms.

Just last month, the White House published a report on Big Data and Differential Pricing in which concerns were raised over fairness and possibilities for disadvantaging consumers through dynamic online pricing. The report also highlighted how, in cases of differential pricing, consumers typically cannot know nor control which data about them is being used to determine the price they are given.

This recent government attention reflects palpable concern, confusion, and perhaps a tinge of anxiety on the part of consumers over uses of personal data by firms. As knowledge of these practices has spread, public opinion surveys regularly indicate that individuals are uneasy about how their personal information is used by companies. Nevertheless, in the U.S. most big data consumer protections are yet to be realized. However, increasing federal attention along with recent movement on a new Consumer Privacy Bill of Rights, point towards better safeguards on the horizon.

Firms to the Rescue?

Updating laws and policies designed to protect consumers is one way to hold firms in check. At the same time, nobody should expect red tape to completely protect individuals when it comes to having their data used in unacceptable ways. As is often the case, regulation is not the panacea some would believe it to be. For example, specifying a comprehensive set of unallowable practices remains a fool's errand. Even developing more general guidelines has proved challenging given the pace at which new varieties and applications of personal data continue to emerge.

From third-party web tracking to fitness tracker data to the next trail of digital breadcrumbs to come along, preemptive consumer protections are impossible to imagine for yet-to-be-imagined applications. This means firms will continue to bear much of the burden, at least initially, in preventing undue persuasion in e-commerce situations.

The rate of development of these innovations and the opacity under which they function have made it impossible to fully comprehend the myriad of ways personal data is being used in attempts to persuade. While this presents a problem for consumers, it is also a problem for firms needing to generate sales while still maintaining the interests of their users, clients, and buyers on whom they depend.

It remains to be seen whether this kind of self-restraint on the part of firms will come at the behest of consumer backlash or as the result of more preemptive reasoning and action. In support of the latter, recognizing the value many individuals place on fairness in retail transactions may help curb differential pricing and discounts in e-commerce, particularly as maintaining consumer trust is often of paramount concern to most business strategies.

The Vulnerable Everyone

Historically, consumer protection efforts have paid special attention to groups especially vulnerable to commercial influence. This includes those suffering from mental illnesses, addictions, or medical conditions, children, the elderly, low-income individuals, and others who might be unduly influenced in the marketplace due to their unique circumstances.

Alternatively, focusing on the experience itself rather than persistent attributes offers another conceptualization of the vulnerable consumer. This type of vulnerability may be temporary, as in the case of moments of intense grief, extreme tiredness, or when under the influence of drugs or alcohol, situations where individuals are prevented from fully knowing and controlling their own consumer decisions. As one definition puts it:

Consumer vulnerability is a state of powerlessness that arises from an imbalance in marketplace interactions or from the consumption of marketing messages and products. It occurs when control is not in an individual's hands, creating a dependence on external factors (e.g., marketers) to create fairness in the marketplace. (Baker, Gentry, & Rittenburg, 2005)

Undetectable personalization experiences like receiving a higher price based on your location or type of computer signal a lack of control on the part of consumers. In most cases, e-commerce users are powerless to affect change and rectify these information asymmetries. Thus, when evaluating products and offers online, we might qualify all internet users as vulnerable consumers as no user is completely free from personalization nor able to adequately determine when personalization leads to disadvantaged content.

If invoking notions of powerlessness or lack of control seem overblown, notice that aside from resorting to rather drastic measures consumers have no comprehensive way to evaluate if personalized e-commerce transactions are affecting them in adverse ways. For instance, consider the financial burden and impracticality of purchasing two separate computers (or smartphones), used in parallel at all times, just to see which one displays the better personalized outcomes such as prices or discounts. "Now should I clear the browsing history on the Mac before or after buying the plane ticket? Or for this purchase would it be better to use the one running the Windows operating system, but only after visiting a few payday loan sites? Or does going to those high-priced designer product sites first produce a better discount on the travel site? Or should I..."

Though this strategy could allow someone to compare the prices or discounts aroused by these different operating systems and/or browsing histories, along with similar techniques, it is of course impractical. Additionally, while a tiny fraction of people might have the time or resources to employ these eccentric measures, the same cannot be said when it comes to testing for other consumer profile metrics that cannot be so easily manipulated as one's operating system or browsing history. This includes consumer profile items like income, race, relationship status, whether someone is a parent or not, and an endless list of known and unknown factors that steer forms of e-commerce personalization at various points.

Of course placing the burden on individuals to pursue these types of counter-personalization strategies in order to make informed consumer decisions is untenable. Envisioning internet users trying to conduct their own ad hoc A/B testing under personalized e-commerce illustrates the difficulty of something as basic as price comparison, an act typically considered routine in evaluating purchases.

The Vulnerable You

National Consumer Protection Week aims to highlight resources to help people avoid scams, identity theft, and many other explicit forms of consumer exploitation. However, a less perceptible harm has been inching onto the radar of consumer advocates in recent years. This concerns the potential for firms to strategically use detailed consumer information to dynamically generate and personalize online ads, offers, recommendations, and product prices, doing so in ways that leverage known vulnerabilities on the part of consumers. Most often this is achieved with little to no transparency and consumers remain in the dark.

As the personal data ecosystem has become more connected and sophisticated in tracking users, these decade old concerns over personalization in e-commerce have steadily gained attention. Further, there is some indication that individuals, when made aware, strongly object to firms using many of the types of information (e.g. financial, racial, location) currently used to achieve online content personalization.1

While in theory dynamic pricing is often simply considered to be more efficient and effective pricing, when consumers are unaware of this practice and when it may leave them disadvantaged, questions of fairness emerge.

In assessing whether you might be a vulnerable digital consumer, consider the following questions: Is the information marketers glean from your data exhaust steering the offers and prices you see in your favor? Or, could you be experiencing online personalization in ways that limit your ability to make informed purchase decisions or obtain the best deal? In the case of the latter, this could simply mean having your personal information (e.g. inferred income) used in attempts to nudge you to pay a little more, pushing you up towards your unique reservation price for a particular product or service.

While considering these questions may be worrisome for some, perhaps more problematic is that there is really no way to know the answers.

In light of information asymmetries, lack of control on the part of consumers, never fully knowing the extent to which online content is uniquely generated, and the resulting murky state of personalized e-commerce, all consumers become vulnerable consumers. Yes, even you.

[1] Stevenson, D., & Sandvig, C. (2015). "Evidence of Contextual Integrity for Personal Information Use in Targeted Online Advertising." American Marketing Association (AMA) Marketing and Public Policy Conference, Washington, DC (June 4-6).

Edits: For clarity, omitted a reference to Netflix and Amazon.