Dynamic pricing is the practice of offering different prices to consumers based on various factors designed to maximize sales and profits, which may include the retailer’s perception of the willingness of a particular consumer to pay at a given price point, often in connection with other factors such as a given point in time. Airlines use dynamic pricing based on complex data analyses involving a myriad of factors including time of day and week, fare class and availability. The ride share service Uber’s surge pricing dynamically bases fares on supply and demand at a given moment. Making projections of whether consumers will pay more or less under different circumstances is an evolving art that can be aided by data analytics, including, now that the data is increasingly available, consumer profiling based on historical consumer behavior and even A/B testing – the practice of testing for different reactions by the same subject based on variables. This can be the basis for personalized pricing, also known as first-degree or primary price differentiation, the “holy grail” of which is to develop a methodology for “perfect price discrimination” that maximizes the amount each individual consumer is willing to pay. Perfect price discrimination is only theoretically possible since the seller must know each buyer’s reservation price (the maximum they will pay) and individualize an offer to them at that price, thus not leaving any potential revenues uncaptured. Beyond the difficulty in ascertain that information, the market prevents perfect price discrimination through competition and data enables real time competitive offering. However, since less price conscious consumers may be less inclined to shop the competition and more cost conscious consumers are more inclined to look for lower equivalent offers, retailers have an incentive to try to identify which consumers are which when they are in front of them and offer them personalized pricing based on their price sensitivity.
Some e-commerce companies have experimented with personalized pricing, sometimes with negative consumer and press reaction, and there have been suggestions by some commentators that the practice is or should be illegal. This issue has been on the radar of the Federal Trade Commission (FTC or Commission) in regard to its potential discriminatory and negative impact on low-income Americans, leaving many wondering about the legal ramifications of its use by retailers. Some of the legal issues retailers should consider in developing dynamic, data-driven, personalized pricing methodology are considered here, though it should be noted that the law is not very developed in this area.
There have been a couple of lawsuits over the years challenging dynamic pricing based on common law and fraud theories, but they have not been successful. However, in compiling consumer profiles upon which to base dynamic pricing models, e-commerce operators should be cognizant of anti-discrimination laws, such as California’s Unruh Civil Rights Act (the Unruh Act). The Unruh Act, violations of which can be the basis for a consumer class action lawsuit, prohibits discrimination in public accommodations (which would include e-commerce) when such discrimination is based upon sex, color, race, religion, ancestry, national origin, disability, medical condition, marital status, sexual orientation, or other arbitrary personal characteristics, with narrow statutory exceptions for senior housing and discounts to consumers “who have suffered the loss or reduction of employment or reduction of wages.” It is important to note that California courts have applied the Unruh Act’s protection to not just immutable protected classes such as gender, sexual orientation, heritage, race, etc. but also to personal attributes such as occupation (in some but not all circumstances), political affiliation and belief systems, physical appearance, and even manner of dress. However, the California Supreme Court has determined that the Unruh Act is not a bar to discrimination for all types of classifications of consumers and does not prohibit discrimination based on mere financial or economic status.
For instance, economic ability to meet the consumer’s obligations is a permissible distinction, but exclusion of a particular vocation unrelated to economic ability or any other legitimate business reason is not. When it comes to selective discounts, the California Supreme Court has stated that businesses may offer “reduced rates to all customers on one day each week,” “might offer a discount to any customer who meets a condition which any patron could satisfy,” or might offer “discounts for purchasing commodities in quantity, or for making advance reservations.” The Court has explained that the “key is that the discounts must be ‘applicable alike to persons of every sex, color, race, etc.,’ instead of being contingent on some arbitrary, class-based generalization.” The California Supreme Court has suggested that discounts based on a consumer’s geographic location of residency might violate the Unruh Act, but has yet to have had that issue properly before it such that it could rule on the issue, and an earlier case has suggested the contrary. Notwithstanding the Unruh Act’s potential ban on discrimination because of personal attributes that go beyond neutral economic factors (e.g., ability to pay), it requires intentional discrimination (disparate impact is not enough, though it might be evidence of pretext). Further, California courts have found that discrimination may be justified by legitimate business interests, which, although the question has not been addressed in a written decision, in the context of dynamic pricing might be established by showing that the pricing distinctions are reasonably calculated to maximize sales and profits without intentionally targeting any particular classification unrelated to purchase willingness or behavior. In other words, pricing designed to maintain as nearly perfect a balance between consumer surplus and producer surplus is an effort to achieve profit maximization, and that ought to be a legitimate business purpose.
Furthermore, using consumer behavior data to set pricing is distinguishable from a single-trait pricing model, which the California Supreme Court has held illegal at least regarding gender (e.g., ladies’ night discounts). Such a single-trait criteria may yield increased traffic or sales, but the triggering trait (women, student, seniors, etc.), the triggering trait is itself the protected category and is attenuated from the economic trait of willingness to purchase. Take senior or student discounts for matinee movies for instance. It may be a legitimate business interest, sufficient to permit the practice under the Unruh Act, to try to match the sale of early afternoon movie tickets with populations that may not be working and thus have availability to attend. However, these categories will be both under and over inclusive. Big data helps solve for this problem by adding consumer behavior and allowing a more accurate and individualized offering. Analytics and machine learning can better isolate the legitimate economic characteristics. Accordingly, data-driven, dynamic pricing based on consumer tracking and targeting actually should be seen as a device that helps avoid the types of personal trait consumer discrimination that the courts have explained the Unruh Act is intended to prohibit, instead relying on economic criteria that the courts have said it permits.
However logical that may seem, it remains untested by the courts. Given the potential for claims under laws such as the Unruh Act, e-commerce operators should make efforts to avoid knowingly creating price distinctions based on personal attributes of their consumers other than those that have a legitimate economic purpose and apply across consumers aside from the economic factors being isolated. This is particularly the case for sensitive (sexual orientation, health status) and historically protected (race, gender) classes. How this is done in practice requires knowledge of how the particular dynamic pricing is being determined and of the case law, and then application of risk tolerance analysis. In addition, the retailer may help insulate its dynamic pricing operations by establishing, through testing and other research, that its criteria selection practices are based on multiple factors and that the methods are proven to foster maximization of profitability and/or sales volume. This said, the FTC’s expressed concerns regarding the negative impact on purely economic classes of consumers (e.g., the poor and, by disparate impact, potentially minorities) leaves some types of dynamic pricing potentially open to claims that it is an unfair practice in commerce. However, at least under federal unfairness authority (as opposed to some state laws), a practice is only illegal if the unfair impact of the commercial practice could not have been reasonably avoided by the consumer and is not outweighed by a benefit to consumers or to competition. A strong argument can be made that a soundly designed and implemented dynamic pricing program benefits both consumers (more consumers are getting opportunities to buy at a price they find appropriate) and the marketplace (retailers are offering consumers more choices and are innovating differing offers in competition with each other). Furthermore, there are almost certainly going to be, for the foreseeable future, many retail opportunities to consumers where dynamic pricing does not occur, such as going to a brick-and-mortar store.
As big data matures, so will dynamic pricing based on e-commerce consumer behavior likely continue to evolve and personalized pricing become more prevalent. Employing privacy-by-design principles, to respect consumers and minimize the negative impacts of data collection and use on them, for the development and operation of dynamic pricing will help retailers avoid potential consumer backlash and possibly even legal liability, as will efforts to avoid mere profiling of consumers based on particular personal traits (e.g., gender, race, age, etc.) as opposed to price personalization based on projections with a basis in consumer behavior attributes derived from data.