If you’ve been following the news recently, you may have heard the term ”surveillance pricing.” In Canada, the NDP has made it a plank of their campaign, creating the Ban Predatory Pricing website and the federal government has promised to enact protections against “harmful practices such as…surveillance pricing” as part of its AI for All strategy. In the U.S., Maryland and California have banned the practice and numerous other states have required companies to disclose when they use a person’s data to set a price. That is the crux of what “surveillance pricing” refers to: when companies create different prices for different people, and set that price based on the wide range of surveillance data they have access to. The practice is also called “personalized pricing.”
Our research shows that surveillance pricing is already here. When news stories recently broke with allegations of Instacart and Delta using surveillance pricing, both firms quickly disavowed the technology, denied using the practice, and stated they would never use personal information to set a price. But other firms are already using the practice and without regulation requiring its disclosure, you might never know how your personal information is driving the price you receive.
Perhaps because of deep public distrust, companies hide the details of how surveillance pricing operates. In one forum of financial firms discussing the potential of using surveillance pricing, participants described the limit not as a lack of information about their customers, nor as a limit in their technological capacity, but around not wanting to seem “too creepy” and acting unfairly or risking backlash. Yet, the temptation is too strong for these firms to resist as industry insiders often refer to surveillance pricing as the “holy grail” of prices—an ability to charge each consumer at their personal maximum willingness to pay.
In attempting to understand how firms are using this practice, we conducted an analysis of corporate earnings calls and pricing firm webinars to find out what has enabled surveillance pricing and the practices of the firms who use it. What we found is that surveillance pricing should not be understood as a stand-alone technology, but as the acceleration of the existing centralization of market power and of practices of data surveillance.
What is surveillance pricing?
Surveillance pricing (also known as personalized pricing and predatory pricing) refers to the automated creation of individual prices based on data about consumers. Companies are now able to personalize prices due to two developments: (1) companies’ gathering and selling an immense amount of data about individuals, and (2) the emergence of generative AI technologies which can sort and use that data to make instantaneous “decisions” to set a price. It is the latter which separates surveillance pricing from the long-standing practice of dynamic pricing, where companies like airlines adjust prices in real-time according to generalized information such as supply/demand. With surveillance pricing, each consumer is presented with a price based on the decisions of algorithms which are constantly learning about what drives a person’s willingness to pay.
In order for surveillance pricing to work, a firm must have a platform through which they can deliver a separate price to each consumer. As companies try to maximize their leverage, these platforms also allow them to isolate the buyer and restrict their ability to compare prices. In this way, surveillance pricing has to be understood as enabled by the growth of intermediary platforms such as Amazon or Instacart and the prevalence of loyalty apps offered by companies like Starbucks and Target.
It is through these platforms that companies can send each consumer their own price without providing an ability to compare it to what others are receiving. In intermediary platforms like Amazon, this is achieved directly through offering different prices for different consumers. For loyalty programs, companies implement differential pricing through discounts, where users are prompted with a lower price based on what the firm knows about the consumer’s buying patterns through data it has gathered through the app and other sources.
As TASK software describes in the case of Starbucks: “Customers who are already hooked on a product are unlikely to decline an offer so tempting… Starbucks can then track their customer data, and use it to drive more personalization efforts.” In these cases, consumers may receive lower prices, but with inequities in who receives what level of discount and with the firm driving purchasing which might otherwise not have occurred, allowing the company to capture more profits overall.
Where does the data for surveillance pricing come from?
Loyalty apps are particularly instructive about the types of surveillance data that companies collect on their consumers. Home Depot, for instance, describes how they receive 3.5 billion visits to their website a year, with 1.7 billion transactions. As a result they told investors “we know what area you live in, what the weather is in your region. If you’re a new homeowner, if you’re moving, if you’re looking to renovate a bathroom or a kitchen… we capture all of this data so that we can turn it into the right message on the right platform at the right time for the right customer.”
However, not all the data gathered about individuals is based on a consumer’s direct transaction with a company. Instead, there is a thriving data broker industry which gathers and resells information about both individual consumers and general patterns about types of consumers. On the more established end of this spectrum, companies like Mastercard use their position as a financial services firm to provide insights about people who live in different zip codes to companies for a fee, highlighting how likely a consumer is to spend on different product types like “children’s clothing and apparel.” Companies that buy this data can then use it in their pricing algorithms.
A more grey area are data brokers like Kochava who gather individualized data about people who may not be aware that they are the target of surveillance. As alleged in a U.S. Federal Trade Commission lawsuit that recently settled and resulted in a ban on some of its practices, Kochava is said to have sold location data on over 300 million people that it gathered through providing geolocation services to other apps, which then passed that data on their users movements to Kochava as part of the agreement. According to the FTC, Kochava then proceeded to aggregate the data, assign those users Mobile Advertising IDs (MAID), and sell that data to other companies for use in profiling. Kochava’s staggering level of detail (in the words of the FTC) included “names, MAIDs, addresses, phone numbers, email addresses, gender, age, ethnicity, yearly income, ‘economic stability,’ marital status, education level, political affiliation, ‘app affinity’ (i.e. what apps consumers have installed on their phones), app usage, and ‘interests and behaviors’.”
How is this data used to generate a price and who benefits?
The practices of companies isolating consumers and gathering data on them is what enables surveillance pricing. Different companies then have different strategies on how to actually implement that pricing. While some seek to maximize how much a consumer is willing to pay for each transaction, others try to maximize sales over a longer period by offering discounts that build loyalty and make it less likely that users ‘churn’ (go elsewhere) so that they can maximize profits over the long term.
For the first, the approach is relatively straight-forward. A company will profile a consumer buying through a website or platform via the available data, and the algorithm predicts how much they are willing to pay before going elsewhere. The company’s algorithm then sets the price at that level so that the firm can maximize their profit on the sale.
In cases of market concentration and where multiple firms are using algorithms to set prices, this can result in a rise in prices across a market. In a study of German gas stations, for instance, the authors found that when all gas stations used algorithms, the technology was trained over time that a drop in prices would lower profits as all other stations matched, whereas increases in prices drove profits. This resulted in a 38 per cent increase in profit margins for gas companies.
Maximizing sales over time is more complicated. Consider UK retailer Sainsbury’s description of their approach to investors: “for each customer, the first step is to calculate their potential value to Sainsbury’s. So, customers with a higher potential value through either their immediate or future predicted lifetime sales, receive a larger Nectar prices discount budget.” What Sainsbury’s describes here is their pricing program, where they offer personalized discounts to those who use its loyalty program, but where these prices are not based on their willingness to pay price but on their predicted “customer potential sales value score”. Here, companies offer higher discounts to customers that are more likely to spend more to drive loyalty—presumably prioritizing higher income consumers that are less price-sensitive. In these cases, company decisions can result in customers who are more “valued” receiving lower prices, driving inequality.
Companies may also factor in geographic and time flexibility in the prices that consumers receive, although the lack of transparency around algorithms make it difficult to be sure. Companies have already been found to engage in geography-based (but not personalized) pricing in order to charge higher prices in underserviced, low-income neighbourhoods. Staples, for instance, was charging higher prices for consumers located in locations with less competitors (which just so happened to generally be low-income communities). With companies able to gather data about consumer travel patterns and then prompt them with lower prices via their apps, companies may target consumers with less competition in their neighbourhood (or with less time to shop around) for lower discounts than consumers with more flexibility.
The window to regulate is closing
On its own, surveillance pricing has rightly received attention from people and their political representatives for concerns around privacy invasions and inequity. Earnings call data and webinars cement this viewpoint, highlighting the sheer amount of data that companies have on individuals and the ways those companies can use it to drive higher prices or to prompt consumers to spend more.
All this reflects not just technological changes enabled by AI, but how these technologies interact with concentrated market power. Surveillance pricing is made possible because of the ability of a small number of powerful firms to isolate consumers, gather information about them and profile them—all while keeping them in the dark about what prices they are charging to other consumers. This accelerates corporate power as consumers lose out, exacerbating an already unequal system. If regulators want to have any chance to actually slow this process down, they need to move quickly—before surveillance pricing becomes normalized and widespread.
