Is real time AI pricing all it’s cracked up to be? This Wharton Professor makes the case against immediate adoption

We may have become accustomed to pricing changes when booking holidays, but the Oasis ticket backlash brought to light the potential drawbacks of introducing real time price changes in novel settings. Professor Santiago Gallino, associate professor at the University of Pennsylvania’s Wharton School, tells us his concerns about implementing these tools, and how to overcome them. 

“These tools can be great, but at the moment, they only go so far.”

Why is there so much hype around AI real time pricing?

It sounds like the Holy Grail for brands – finding the sweet spot where the brand makes the best price that the consumer is happy to pay. But do not assume that this is the right tool for you, right now. For an Amazon of the world, prices change multiple times a day–sometimes within the hour–so it would be almost impossible for a company to navigate without some of these tools. But fashion brands, for instance, will usually decide on an initial price to go on the market with, and only change it in the sales: there’s really not much scaling involved.

Buy into the hype and the risk is that companies end up spending money on shiny pieces of technology that they don’t even need, unless their products are being priced minute-to-minute and at scale. Added to that, many companies’ systems may not be ready to implement the sophisticated models required. In the vast majority of situations, I would argue there is a simpler approach to achieve what’s required.

“[T]he risk is that companies end up spending money on shiny pieces of technology that they don’t even need.”

How about in the future? Will companies mature to the point where everything can be AI real time priced? 

Over time, companies will become more and more sophisticated with their pricing, trying to delegate some decisions to an algorithmic tool–especially when dealing with scale. If you have thousands of products in a new collection, even if you group items, the problem of how to price becomes incredibly complex.

Another area where real time pricing becomes interesting is if you’re part of the circular economy. I’m working with a clothing rental company. Their products are all slightly different to each other as they have been worn multiple times before being sold. How can that be priced objectively? Scalability is a big factor because now you don’t need to price 20,000 different SKUs; you need to price 1,000,000 individual items.

“Another area where real time pricing becomes interesting is if you’re part of the circular economy.”

What should companies do to make the most of AI real time pricing options? 

In our recent HBR article, my colleagues and I outlined the steps to making the most of the products, which in brief are:

• Decide your focus If most of your sales come from one area, focus there. If your sales are more dispersed, focus on mid-range or long-tail products.
• Build a consumer model Create a formula considering how your buyers make purchase decisions. Look beyond price, considering issues such as quality, return options and competitors.
• Experiment See how customers react to changes to determine price elasticity.
• Measure, test and optimise

Ultimately though, using these tools effectively is about making pricing decisions in a smart and robust way. Real time pricing often isn’t the whole answer. You need to incorporate non-rational components, such as pricing based on what is appealing to customers. That includes traditional tactics like using 99p or newcomer discounts but also more complex behavioural understanding. From what I’ve seen so far, this is where algorithms tend to fall behind. Furthermore, AI today cannot replicate you and your team’s customer understanding and experience.

This is why the technology can be useful in shaping decision-making, rather than full implementation. It’s like writing an article: ChatGPT may be useful for research or even a draft but it’s not going to stand up to a piece written by a journalist or an academic. These tools can be great, but at the moment, they only go so far.