Dynamic pricing – don’t leave the thinking to artificial intelligence (AI)

Consumer advocates perceive dynamic pricing as deceptive. Retailers hope to achieve higher margins through the use of special tools. In this article, you will read about how AI is making prices more dynamic and why it is often better to allow common sense to prevail.

Compared to the possibilities for dynamic price displays offered by current software solutions such as prudsys or Blue Yonder, the mechanisms in place at petrol stations are little short of touching. However, changing as they do seasonally and in the course of the day, prices for fuels are instructive in two respects.

First, they show that consumers are aware of dynamic prices and are also learning how to deal with them. But they also demonstrate how narrow the corridor for price acceptance is. Because the only time there is talk in the media about a “rip-off” or “con” is if the oil companies overdo it with their price hikes.

Prices in the dark thanks to AI

When software giant IBM finally gets involved in a segment, you know that an issue has become mainstream. “Big Blue” is now also offering its customers an AI-powered tool for dynamic pricing. But what does artificial intelligence actually mean in this context?

Simple re-pricing systems rely in the first instance on observation of the competition and are controlled by a set of rules. If a competitor lowers its price, the software responds within a framework defined by the retailer. Understandably, dynamic pricing software providers like to keep their algorithms close to their chest. Their systems are increasingly complex to control and are adapted to retailers’ strategies.

Some parameters used for pricing are fairly obvious. These include:

• Time: Both the time of day and seasonal factors are taken into account. Prices rise in the run-up to Christmas, then they fall. Products are offered at slightly lower prices during the day, and then rise again in the evening. It may even be possible to react to special events such as sports broadcasts.
• Peaks: The systems respond to increasing demand for specific products. If an underdog gets into the final round of a football championship, interest in merchandising products grows. This can also be played in two directions, initially by undercutting competitors’ prices. If the systems recognize that competitors all of a sudden have to specify delivery times for a product, higher prices will be introduced if stocks are available.

This sounds very simple at first, but taken as a whole, it is a complex matter. Because if the AI is to be able to make the right decisions independently, the system has to experiment and learn. And to do that, it needs a great deal of historical data that can be enriched with additional information.

Furnished with this information, the system continuously analyses the interaction between prices and turnover. The hypothetical optimum price is confirmed or refuted using A/B testing during operations. Things get complex when the system is also designed to make product recommendations at the same time. Sophisticated solutions take dependencies between complementary products or substitutes into account, allowing the system to favour the sale of alternatives that offer higher margins, again making allowance for available stocks. There is really no such thing as off-the-shelf dynamic pricing.

Don’t overstep the mark

Customer appraisal is simply part of the retail trade. Using external characteristics to find out whether a customer is even solvent in the first place and what price they are likely to pay for a product has always been and remains part of bricks-and-mortar retail. In the self-service era with electronic labels and the Internet as a research source, most customers have simply forgotten about it.

Dynamic pricing opens up new dimensions in appraisal. Buyers of Apple products are generally considered to be better earners. By identifying these consumers, through information which their device transmits automatically anyway, it would therefore easily be possible to show Apple users a higher price than other customers for a trip or a product in an online shop. A hypothetical example, you say? By no means. Google will have no trouble leading you to reports that document strategies of this kind. It’s in the nature of things for mistakes to be made in this area, too. However, a jeweller who appraises a customer based on their car or clothing when they arrive could also make an error.

Consumer advice centers and advice programmes have in the meantime raised awareness of such strategies among consumers. They are starting to defend themselves, for example by using “privacy boxes” that do not identify their computer or smartphone.

In any case, retailers should ask themselves whether it’s a good idea to make full use of the possibilities of dynamic pricing. Not everything that AI can do in this area has been thought through in the longer term. While the margin generated from an individual customer may be higher, what good is that if the customer gets annoyed and never comes back?

Fluctuating prices are part and parcel of a market economy. And dynamic pricing supports retailers in their sales strategy. Taken to extremes, however, the technology can also be detrimental.