Dynamic pricing is a hot topic in pricing circles these days. Back in October, it came up in a number of keynotes at the Professional Pricing Society’s Fall Conference. Some pundits are predicting that dynamic pricing will take over large swaths of the pricing industry over the next decade.

Dynamic pricing is not new. Back in the days where people bought most things through barter and then haggling, all pricing was dynamic pricing. Auctions are another form of dynamic pricing. One that is likely to dominate the future of dynamic pricing.

When pricing software companies talk about dynamic pricing they are generally talking about a system that uses a variety of inputs and calculations to create a near real time price for each buyer based on an estimate of their willingness to pay (WTP). These systems have their roots in revenue management and were originally developed for use by airlines. For airlines, load management is a critical issue. Every empty seat is lost revenue, and given the high fixed costs of a flight and the very low variable costs getting those last few people on board can determine if a flight is profitable or not.

You may have seen the add where there is a chicken pecking at a keyboard with the voice over telling you that this is how some airlines set prices. Well, the revenue management software is the chicken.

It is a smart chicken, applying a lot of machine learning to a lot of data. The same approach can be extended to any situation where there is data that indicates variability in demand. What these systems do is map variations in the data to build virtual elasticity of demand and cross elasticity of demand curves. Elasticity of demand is the degree to which demand changes when the price changes. Cross price elasticity is the tendency of buyers to change vendors in response to price changes (If Coke raises its prices to what extent do people switch to Pepsi?). These elasticity models are combined with external data of various types to estimate willingness to pay for any particular buyer and use it to predict that customer’s willingness to pay.

Price elasticity of supply and cross price elasticity interact in interesting ways. In many cases they operate on different timelines with price elasticity of demand having long term effects (what to by) and cross price elasticity operating on shorter timelines (which solution to buy). Dynamic pricing works best where either price elasticity of demand or cross price elasticity of demand is high. They serve different purposes in these two different scenarios.

In the upper left (low price elasticity of demand with high cross price elasticity, one is trying to edge out the competition without triggering a price war. One does not want two competing automated dynamic pricing systems inadvertently triggering a price war.

In the lower right quadrant, where price elasticity of demand is high and cross price elasticity is low, one is trying to optimize price and volume to maximize profits or maximize revenues (these are not the same price in most scenarios). Dynamic pricing can be incredibly effective in this scenario and the risks are relatively low.

It is the top right quadrant that is most interesting. Here price elasticity of demand and cross price elasticity are both high. These are the conditions for a chaotic system. It is not clear to me how well dynamic pricing will work in this scenario, especially when multiple market participants are all trying to apply these systems. This will be something to watch closely.

Even here, there are some serious problems. The first is around pricing fairness. People want prices to be fair and have a strong reaction when they think they are being manipulated or taken advantage of. Pricing fairness comes from transparency, consistency and shared value. If you are going to adopt dynamic pricing you will need to be able to explain how it works and why it is fair.

How to explain systems based on machine learning (as the best dynamic pricing systems are) is a general challenge. It is possible. See the ongoing work on xAI (explainable AI) being led by DARPA. These concepts should be applied to dynamic pricing.

In his excellent book The New Invisible Hand, Kyle Westra covers pricing transparency in depth and shows that you can have price transparency or price setting transparency but with dynamic pricing you cannot have both. In certain markets it may be important to provide transparency on both (healthcare comes to mind) and in these markets dynamic pricing is a poor solution.

The conflation of willingness to pay and differentiated value has to come to an end. Willingness to pay is an outcome of the creation, communication and delivery of differentiated value. It is an outcome and not a driver. Pretending that you understand value because you can estimate willingness to pay is wrong headed.

The second challenge to dynamic pricing is dynamic buying. If sellers are going to adopt dynamic pricing en mass then buyers would be well advised to do so to. If sellers want to optimize for willingness to pay - WTP (as a poor substitute for understanding value), then buyers are interested in willingness to sell - WTS and are eager to find the vendor that will cut them the best deal. This is also referred to as M2M (machine to machine) pricing. Assuming that dynamic pricing is widely adopted we should also assume that dynamic buying will also be adopted. We will then have buyers and sellers interacting in a dynamic environment. This sounds like an auction to me. A complex auction in which there are multiple players on both sides. Mechanism design (the design of auction systems) will become a key skill of pricing experts as M2M pricing becomes more common.

No doubt M2M pricing will become a significant part of B2B pricing in the future, just as it already is the dominant mode of pricing financial instruments (with mixed outcomes and the occasional market crash). But is this the only path forward? Dynamic pricing is transactional. it does not help build long-term commitments on either side. In the airline industry, the widespread adoption of dynamic pricing has contributed to weakening brands and reduced customer loyalty. Is this what you want for your company?

The counter to dynamic pricing is likely to be outcome or performance-based pricing. In this approach, buyer and seller share the economic gain. This has often been discussed but it has been hard to execute on, much harder than dynamic pricing in fact. Why has this been hard?

1. Causality is complex

2. Data is hard to come by

3. Risk-reward curves have different shapes for buyer and seller

4. Timelines are long

5. Trust is rare

At Ibbaka, we believe that all of these issues can be addressed. There have been technology advances that address the causality and data issues. New pricing designs can build alignment around differing distributions of risk and reward. With transparency around data, risk, reward and pricing mechanisms one can build trust.

Subscribe to this blog follow our work on performance-based pricing or reach out to us at info@ibbaka.com to discuss how we can make this work for you.