AI and the art of pricing

Whether individuals know the term “dynamic pricing” or not, they have almost certainly been impacted by it. From airline tickets and hotel rooms to cab rides, pricing increases at times of peak demand and this has become a commonplace experience in those sectors.

Most of these systems optimize revenues using basic algorithms designed to juggle a few variables at a time—but the systems are getting smarter.

Now, applying AI to dynamic pricing opens the door to far more powerful pricing models that can be applied to more products and the use of what some call “surveillance pricing.”

To vendors this approach offers new levels of business efficiency. At the same time, it has been criticized as pushing up against the ethical - and, in some cases, regulatory - limits of AI use.

We look at the opportunities and risks.

Traditional pricing models vs. AI-driven pricing

Typically, pricing models take a number of factors into account. At the most basic level, production costs, market perception, competitor pricing, and other market factors play into deciding what to charge for a product. Then, companies have to decide whether they will use straightforward value-based pricing or get into tiered models, per-user, flat rates, or any number of other models.

In the past, a lot of work went into figuring out how to maximize revenue, and sometimes opportunities were missed. The models were rigid and reactive, requiring more time to adjust. Dynamic pricing has helped change that, and now, AI is taking it to the next level.

AI pricing engines use machine learning algorithms to determine optimal pricing in real time. By constantly collecting and processing data, the engine can choose which algorithms to apply in a given situation. AI pricing engines can constantly change their strategies from predictive analytics to optimization algorithms, all with the aim of maximizing profit. And, while it may be easy to assume that means increasing prices, it does not have to. In fact, one way dynamic pricing can work is to lower the cost of items for a reluctant purchaser to encourage them to pull the trigger.

Maximizing profit may be the ultimate goal, but AI does that by adapting to consumer behavior. Creating personalized pricing is integral to success, and sometimes, that means lowering pricing to meet customer expectations.

The risks and rewards of AI pricing

By now, many of the pros of AI pricing should be obvious. The ability to proactively change pricing based on real-time user behavior to maximize revenue can also offer benefits for the customer. That does not mean it is without its possible pitfalls for brands and their consumers.

AI learns over time. The more data it collects, the more effective it can be. At times that may mean better deals for consumers, but companies can risk alienating customers if they are not careful about how they employ this polarizing strategy.

Elisabeth Buchwald of CNN wrote about hearing from a friend who saw a buy-one-get-one-free deal in their Starbucks app, only to sign in and see that the same deal was not being offered to her. Buchwald wrote, “It’s likely that Starbucks used artificial intelligence to determine that my friend, if offered a promotion, would make a purchase they wouldn’t otherwise have, whereas I would make a purchase regardless. The AI turned out to be correct."

Exploitative AI?

But when does personalization step over the line and become exploitative? The Federal Trade Commission (FTC) is trying to determine the answer to that question. In a recent statement on its website, the FTC ordered eight companies that offer “surveillance pricing” products to provide information in order to understand the potential impact this technology can have on consumer privacy and competition, as well as consumer protection.

FTC Chair Lina Khan said in the statement, “Firms that harvest Americans’ personal data can put people’s privacy at risk. Now firms could be exploiting this vast trove of personal information to charge people higher prices.” While the FTC looks into the legalities, consumer trust could be on the line.

The Uber company provided an early example of this back in 2017 when an emergency struck the streets of London. Consumers opened their Uber apps in an attempt to get home to safety—causing the surge pricing algorithm to increase prices by 200%, according to Harvard Business Review (HBR). It was not the first time something like this happened, but as consumers become more aware of what’s happening, brands need to be more careful. HBR says, “These systems can create an uncomfortable tension between earning customer loyalty and earning money.” With this in mind, companies must walk the fine line between profit and loyalty.

The FTC investigation into surveillance pricing is digging into the products and services being offered, data collection and inputs, customer information, and the impact on consumers and prices. They have pulled information from Mastercard, Revionics, Bloomreach, JPMorgan Chase, Task Software, PROS, Accenture, and McKinsey & Co.

The antitrust risks of AI algorithms

A related question is to what extent vendors using a shared pricing algorithm could be engaged in anti-competitive 'price fixing' in violation of antitrust legislation such as the Sherman Act in the USA.

As Bloomberg Law recently reported: "Federal Trade Commission Chair Lina Khan last month encouraged enforcers and regulators to be vigilant as AI-pricing algorithms “can facilitate collusive behavior that unfairly inflates prices.”"

The article concluded with a warning to users of pricing algorithms: "Companies eager to implement AI into their business should be aware of recent activity and take steps to make sure they comply with antitrust laws."

Effective - but ethical

In the meantime, companies may want to hedge their bets and use AI to aid in smarter pricing without crossing the line into surveillance or collusion. For instance, they may use AI to facilitate A/B testing to refine price elasticity.

It can also be used to bundle complementary products and analyze the success of past promotions to develop future offers. Seasonal or inventory-based pricing are other options for brands looking to employ AI-driven pricing without delving too far into customer data.

With a bit of careful consideration, companies can find ways to employ AI without alienating customers or crossing regulatory lines.