Algorithmic pricing is both efficient and absurd

Deterred by the £260 cost of a return train ticket to London, one Sunderland football fan has found a cheaper way to cheer on his team at Wembley this weekend: travel via Menorca. A Ryanair flight to the Spanish island, an overnight stay in an airport hotel, a return flight to London and a cadged car ride back home work out cheaper than the extortionate cost of a rail ticket. 

The experience of this Black Cats’ supporter encapsulates the costs, benefits and occasional absurdity of algorithmic pricing, now used by most train and airline companies to maximise yield. Companies can rapidly increase prices to exploit surging demand from football fans, for example. But they can also automatically lower prices to entice customers in weaker market conditions, as is the case with budget airlines. 

At a time when consumers are fixated on rapidly rising inflation, price-setting mechanisms should come under greater scrutiny. But digital marketplaces present a particular challenge given so many prices are set by automated algorithms rather than slow-moving humans. Plus, speed and flexibility of service can often prove more important determinants of online purchases than price. 

The promise of dynamic pricing is that it better matches supply and demand, producing greater economic efficiencies. But it should certainly be attracting more attention from regulators than it currently does. There is mounting evidence that algorithmic pricing can also lead to unfair discrimination between consumers and encourage implicit collusion between firms, raising prices overall.

The positive case for dynamic pricing is made persuasively by Yossi Cohen, who sells such services. The Israeli co-founder of Quicklizard explains how his company combines historic data, product inventories and cost structures with external variables, such as competitors’ prices, economic indicators and the weather, to automate pricing decisions. “We help companies find the sweet spot on the demand curve,” says Cohen.

The industry yardstick is that dynamic pricing can increase a retailer’s bottom line by 7 to 10 per cent, which makes a critical difference in a sector notorious for thin margins. But he argues that the ubiquity of price comparison sites also empower consumers to get the best deals. “Online commerce has changed the equilibrium in favour of the consumer,” Cohen says.

Intuition would suggest that an online retailer’s ability to track and automatically respond to rivals’ price cuts should increase competition. But two papers, co-authored by Alexander MacKay, an assistant professor at Harvard Business School, concluded the opposite.

In studying how dynamic pricing works in the real world, MacKay found that when a company with a sophisticated pricing algorithm instantly matched its rivals’ price cuts, product prices rose over time. A competitor with inferior technology would have no incentive to lower prices. “That takes price competition off the table,” says MacKay.

Even if there is no overt price collusion to excite the interest of the antitrust authorities, dynamic pricing can weaken competition. Savvy, time-rich consumers might be able to shop around and take advantage of time-limited offers and promotions, but others cannot. “The beautiful thing about competition is that when there is a lot of it consumers do not have to be particularly sophisticated to benefit from it,” MacKay says. 

In the past, regulators have stepped in to correct abuses of pricing innovations and market power. So, for example, retailers are banned in many countries from selling goods at below cost to drive competitors out of business. 

But the challenge of regulating online marketplaces is mind-bendingly complex considering their velocity and opacity. Companies can also personalise offers, meaning no two customers necessarily see the same one. The ecommerce giant Amazon changes prices 2.5m times a day across all its product lines. 

History suggests blanket price controls are a bad idea. MacKay and his co-authors offer two more innovative remedies. One would be to limit the frequency with which online companies can change prices. The second would be to bar companies from including competitors’ prices in their own algorithmic models. Both measures would encourage more blind competition and discourage implicit collusion.

Short of such regulatory intervention, consumers will have to rely on their own wits to navigate digital markets. The trouble is not all of us have the ingenuity of Sunderland football fans. 

john.thornhill@ft.com