When automated pricing becomes collusion

29 May 2018


Machine learning systems are not unlike raising a particularly brilliant child – they could move ahead and gain knowledge at speed, but you need to give them guidance and an ethical framework or they could break through lines that adults wouldn’t cross.

The possibility of automated collusion is one of those lines. As I’ve written previously, pricing is one of the most important issues for any business and Data and Analytics is bringing huge advantages to companies in making pricing decisions. Competition is tough in every sector, and companies are understandably keen to get the best prices for their goods and services.

That’s why many companies have turned to automated systems and algorithms to generate prices. Consider the famous taxi app where algorithms set prices that fluctuate according to circumstances (primarily demand). For most companies using these systems, the price that an online customer sees – the price of a particular holiday, for example - will vary according to the time of day, the number of holidays being sold, how many other people are on the site at that time, or even the browser the customer is using.

Companies are starting to look not only at what their customers are doing, but at what prices their competitors are setting at different times during the day. And this is where the potential for automated collusion comes in.

According to academic research, the automated pricing systems of some companies are so good that they can automatically align to the price set by competitors (or set it fractionally over or under in order to build a market). So even though no human has spoken to another and fixed the price, the computers have, in effect, colluded to identify what the best price should be to maximise profit amongst themselves. The computers have done an excellent job of setting the price and generating profit but in the process, by current definitions, they could also have done something illegal.

This issue is already on regulators’ radar. The European Union’s antitrust agency is looking at whether algorithms make it easier for companies to collude over prices online. The EU’s commissioner for competition, Margrethe Vestager, said that they would be ‘keeping an eye out’ for cartels that use software to work more effectively, and would consider stricter fines for companies considered to be misusing these systems. ‘I don’t think competition enforcers need to be suspicious of everyone who uses an automated system for pricing,’ she said. ‘But we do need to be alert.’

Not all automated pricing systems are suspicious, of course, but with regulators looking closely at the issue it’s important to be confident that your own system isn’t unwittingly taking you over a red line. It’s essential to have clear sight of what automated algorithms are doing, and to have control over the ability of the algorithm to analyse and align with competitors, so any issues can be quickly identified and investigated.

In summary, pricing algorithms and automated pricing systems play a very important role in helping companies set appropriate prices. But get it wrong – even innocently – and the regulators could come knocking, with all the financial pain and potential brand damage that entails