Ibbaka

View Original

Pricing is an open game

Steven Forth is a Managing Partner at Ibbaka. See his Skill Profile on Ibbaka Talio.

TL:DR: AI is changing the pricing game

Pricing is a special kind of game. One that plays out over multiple moves by many players where the rules can be changed. Pricing is an open game.

AI is changing the nature of the pricing game by

  • Revealing hitherto hidden information, making games of hidden information into games of (semi) perfect information

  • Reducing uncertainty, and thereby reducing risk discounts and increasing the rewards of collaboration and enabling outcomes based pricing

  • Making positive sum games the norm, the goal of pricing will be to make pricing and configuration decisions into games where all players can win

But, pricing has unique characteristics that make designing pricing AIs a challenge.

Pricing is an open game of hidden information played by teams over time. The teams have multiple, often conflicting, goals that need to be resolved.

Pricing AIs will need to be able to

  • Deal with high dimensional and sparse data

  • Combine supervised and reinforcement learning with Generative Adversarial Networks (GANs)

  • Be able to optimize for multiple, conflicting goals

  • Adapt rapidly to changing market conditions (external events)

  • Adhere to pricing AI ethics of

    • Transparency (explainable AI)

    • Non bias

    • Shared value

Pricing is often thought of as a kind of game. There is even work being down to apply computational game theory to pricing. Pricing is often taught using role playing games to help people understand basic principles, like the interactions between cross price elasticity (the tendency of a buyer to switch vendors in response to price differences) and price elasticity of demand (the impact of price on overall demand volume) or to understand how pricing decisions cascade through a supply chain.

But if pricing is a game, what kind of game is it?

Games can be characterized in various ways. One of the most fundamental questions is whether more than one player can win: is the game negative sum, zero sum or positive sum? Another question turns on how much information is available and who it is available to. And then there is the question of time and the number of players.

At Ibbaka, we think of pricing as an open game, one where the rules can be changed as part of the game play.

Let’s look at some of these characteristics of games and see how they apply to pricing.

Is pricing negative sum, zero sum or positive sum?

Zero Sum Game

In a zero sum game whatever one player gains the other loses. Wins and losses add up to zero.

When pricing is purely transactional it can become a zero sum game, leading to tough negotiations between buyers (often represented by procurement) and sellers. Value based pricing is meant to avoid this and to shift the conversation from price to value and make sure that pricing negotiations are part of, but do not dominate, the value conversation.

Markets with a high cross price elasticity (the tendency of a buyer to switch vendors because of price) have a tendency to be dominated by zero sum games.

Positive Sum Game

In a positive sum game all players can win. This should be the goal of value-based pricing and is required for collaborative, long running business relationships.

To get to a positive sum game one must begin by optimizing for value to customer (V2C) and aligning V2C and the target value ratio. The value ratio is Lifetime Value of a Customer (LTV) over V2C.

LTV / V2C where V2C > LTV.

One makes pricing into a positive sum game by creating value, basing price on the value created and then sharing the value between buyer and seller. In our experience, the value ratio ranges from 10% for earlier stage solutions, where there is still some risk or uncertainty (this is also known as the risk discount) and 30% for more established solutions. Going above 30% will often open an opportunity for competitors and lead to push back from buyers.

See How to negotiate price (getting to positive-sum pricing).

Negative Sum Game

Played poorly, pricing rapidly becomes a negative sum game. In a negative sum game both players can lose. This happens in several ways.

  • The price is too low, the vendor cannot sustain value and begins to cut corners or underinvest.

  • The price is too high, value expectations cannot be met, and the buyer moves to a less capable solution, but at a lower price.

  • The offer is poorly configured, with a high cost to serve relative to the value being created. Often, small changes can reduce the cost to serve while maintaining value, unwinding the negative sum game.

Price wars are the classic example of a negative sum game in pricing. These are most common in markets where there is high cross price elasticity (the tendency of buyers to change vendors purely on the basis of price). They result in a race to the bottom, one that can drain a category of the ability to sustain innovation and value creation for customers. Customers may think they are benefitting from the low prices created by a price war, but over time service, innovation and value all decline.

Is pricing a game of perfect information?

A game of perfect information is one like Chess or Go. All players can see all information and no one is in a privileged position. The market for equity in publicly traded companies is meant to be a game of perfect information, in which all parties have access to all relevant information. Anything else is considered insider trading and is punished.

Economists like games of perfect information as they are relatively tractable. AI developers like them as it is now clear that a deep learning AI can learn how to win any game of perfect information, and it can even learn to do this by playing itself.

B2B pricing is seldom a game of perfect information. For most solutions we do not know exactly how much value will be created and it can be difficult to compare solutions and prices from different vendors.

A game of hidden information is one in which each party has some of the information relevant to value and pricing, no party has all the information, but together all the information exists. By cooperating the players can change a game of hidden information into one of perfect information. But they will only do so if it is in their interest to collaborate. The well studied game of Prisoner’s Dilemma is an example of a game of hidden information. In this game, the outcome depends on the choice made by the other party, each party knows their own choice, but not the other party’s.

Investopedia has an excellent article for those interested. They summarize as follows…

  • A prisoner's dilemma describes a situation where, according to game theory, two players acting selfishly will ultimately result in a suboptimal choice for both.

  • The prisoner’s dilemma also shows us that mere cooperation is not always in one’s best interests.

  • A classic example of the prisoner’s dilemma in the real world is encountered when two competitors are battling it out in the marketplace.

  • In business, understanding the structure of certain decisions as prisoner's dilemmas can result in more favorable outcomes.

  • This setup allows one to balance both competition and cooperation for mutual benefit.

In a game of chance, there are unknowable factors that can impact the outcome. Humans seem to be drawn to games of chance, perhaps because so much was out of our control early in our evolution. Our brains are designed to operate under conditions of uncertainty, and as Prospect Theory has shown, we have a bias to avoid risk in situations where we are not sure what will happen.

Good pricing takes advantage of prospect theory to frame value in terms of risk reduction. The player with the most information and ability to manage risk is likely to win. In most cases, SaaS companies have more information than buyers on the potential impact of their solution and can use this information asymmetry to take on more risk and claim a higher price.

Will AIs change the information dynamics in pricing? The open question today is how AI will change information dynamics. We can already see this happening in two ways.

AIs reveal hitherto hidden information and convert games of hidden information into games of near perfect information. AIs are good at revealing hidden strategies and teasing out the implications often lost in the data. This will change how we develop pricing strategy and execute pricing tactics. Assume that the other player has an AI that is every bit as good as your own.

AIs reduce uncertainty and make the world more predictable, thereby reducing the risk discount that drives down prices, especially prices of early stage innovations. Some of the best work on the impact of AI can be found in the books Prediction Machines and Power and Prediction by Ajay Agrawal, Joshua Gans and Avi Goldfarb. They argue that AIs are basically prediction machines, they make prediction cheap and scalable. As a result, many things that were difficult or expensive to predict before will be much easier and cheaper to predict. The result will be that we move from rules based behavior to decision based behavior with the decisions predicated on predicted outcomes. It will be easier to track the outcomes of decisions and course correct. This will transform pricing in many ways.

  • AI will enable outcomes or results based pricing - rather than price based on estimates of value pricing engines will predict value, estimate risk and help players decide how to share the rewards.

  • AI will help us configure functionality, data and services in the ways that deliver the most value at the lowest cost, making more and more pricing games over into positive sum games.

  • AI will eliminate the need for price optimization based on estimates of willingness to pay (WTP) as WTP will be determined by collaboration to increase value for both parties.

Pricing is a game played by many players over time

Pricing is not a once and done thing. Every pricing action plays out over time in response to actions taken by other players. We know from the Prisoner’s Dilemma game mentioned above that the best play is different when one will only encounter the other player once from when there will be repeated, open ended, interactions. In the latter case, cooperation is rewarded.

The four most common pricing actions are

  • What functions, data and service to package together

  • What to price, which of the value metrics should be used as pricing metrics

  • How much to price, pricing levels and pricing curves

  • When to discount, who to offer the discounts to, how much to discount

Each of these actions leads to responses from customers and competitors. A new package will often be met with a corresponding package from a competitor. A price change will be noticed and exploited if possible. Discounts come to be expected by buyers who game the system to optimize the discounts they can receive (as one would expect them to).

Good pricing plans for how the game will be played over time.

Pricing is a team game, with three teams engaged. Like any team sport, people play different roles.

The internal vendor team: Pricing engages all of the critical business functions from R&D and product development, through product marketing, sales, implementation and professional services and customer success. Finance also has a big stake in pricing and its impacts. Pricing may report to product, to sales, to finance or directly to the CEO, but no matter where it reports it has to consult and inform the other functions on the decisions being made and how they are expected to play out.

They buyer team: Just as the vendor has a team engaged on pricing, most B2B buyers are also making a decision as a team, and the interests of the different players should be taken into account.

The buyer-vendor team: The most successful pricing games are played when the two parties form a combined team that shares goals and has clear processes for deciding on configurations, services and pricing. This is a key to moving from pricing as a zero sum game to pricing as a positive sum game.

Pricing is an open game

Perhaps the most interesting thing about pricing is that it is an open game. An open game is one in which the players can change the rules as part of the game play. Open games will be especially hard for AIs as the rules keep changing. Even the definition of winning can change and different parties may have different understanding of what it means to win.

For one exploration of open games see here (this one using tic-tac-toe as an example).

Some of the key ways pricing is played as an open game is to change the basic pricing metrics and packaging strategy. Changing a pricing metric can be especially effective.

Value Metric: The unit of consumption by which a buyer gets value.

Pricing Metric: The unit of consumption for which a buyer pays.

There are generally more value metrics than pricing metrics and part of pricing design is to choose the pricing metric that best tracks value. This is a choice and because it is a choice a different choice can be made. One of the most powerful ways to change the pricing game, and set the board in your favor, is to change the pricing metric for a category.

You can also change the game by changing or adding packages. Packages are what customers buy, and they compare your packages with each other and with the alternatives. Changing packaging can change how the comparison is made.

Another way to open the pricing game is to add value paths for additional buyers.

Value Path: A sequence of actions taken by a user that results in something of value.

Adding a new player to the buyer team can change the dynamics in your favor. One way to plan this is to create a grid that has the members of the buyer team on the X-axis and the value drivers (positive and negative) on the Y-Axis. Then see which member of the buyer team cares about which value driver. Make sure that every buyer has positive value drivers and that the positive outweighs the negative.

The most compelling way to change the pricing game is to change what it means to win.

The best way to do this is to shift the conversation from price to value. Be able to define and demonstrate how much value you are creating for a customer and how you are sharing that value. A conversation based on value to customers (V2C) and the value ratio (VR) is likely to lead to a higher customer lifetime value (LTV). A customer lifetime value, not short-term transactional gains, should define what it means to win the pricing game.

Implications for pricing AI

The nature of pricing poses some extreme challenges for pricing AI.

Pricing is

  • An open game

  • Of hidden information

  • And sometimes of random events

  • With partial (sparse) and high dimensional data (many different types of data can impact the game)

  • Played over time

  • By multiple parties

  • Where each party is a team

  • And each party has different, multiple, and often conflicting goals for optimization.

Given this, how do we build pricing AIs?

They will need to be able to

  • Deal with high dimensional and sparse data

  • Be able to optimize for multiple, conflicting goals

  • Adapt rapidly to changing conditions (external events)

No one approach to AI or one set of models (including Large Language Models like GPT-4) will be able to do all of this. Pricing AI will need to combine several different approaches to AI into one solution. These approaches include

  • Pattern recognition and social network analysis (for segmentation)

  • Supervised learning (for optimization problems)

  • Reinforcement learning (to develop game playing strategies)

  • Generative Adversarial Networks or GANs (to improve adaptability)

  • Large Language Models or LLMs like GPT-4 or LLaMA (to support conversations, inference and concept integration)

The first generation of pricing AI, which was generally used for revenue management and price optimization, will not be able to support pricing AI going forward.

This is a huge opportunity to innovate and create new systems that make pricing more effective for all parties and that shift us from zero, or even negative, sum games to positive sum games.

But pricing AIs will need to adhere to strict ethincs.

  1. Is it clear how we set prices? Can we explain this to ourselves? Can we explain this to our customers? Is our website clear on this? Can sales communicate it?

  2. “Is our pricing unbiased?” Do we treat all of our customers in a way that respects their backgrounds and needs and does not take advantage of their situation or weakness.

  3. Are we creating differentiated value? Do we use this to set prices? Do we adjust prices based on the value we are creating? Are we sharing that value with our customers?

See The ethics of pricing AI.

Read other posts on pricing design