The Wrong Game

John Nash proved something counterintuitive in 1950 that took business about seventy years to not understand: rational players in a competitive game will reach a stable equilibrium, but stable and optimal are not synonyms. The equilibrium they reach depends entirely on what game they think they're playing.

This matters for ecommerce in ways most operators haven't quite articulated yet.

What Nash Actually Said

The Nash Equilibrium is a state where no player can improve their outcome by changing strategy alone. Everyone is playing their best response to everyone else. The result is stable. It can also be terrible.

The prisoner's dilemma is the canonical example. Two suspects, interrogated separately. If both stay silent, both get light sentences. If one defects and the other stays silent, the defector goes free and the other gets the maximum. If both defect, both get medium sentences. Rational self-interest produces a Nash Equilibrium where both defect, both get medium sentences, and both would have been better off staying silent.

The equilibrium is stable. The outcome is worse than what silence would have produced. Nobody can improve by changing strategy unilaterally. The only exit is changing the game.

Here is the part that doesn't get enough attention: the prisoner's dilemma is a zero-sum adjacent structure. The payoffs are arranged so that cooperation feels risky, defection feels safe, and the rational actor chooses the worse outcome because they're solving for the wrong objective.

Now think about your marketing team and your merchandising team.

The Ecommerce Version

Marketing owns acquisition. They're measured on traffic, cost per click, return on ad spend. A good quarter for marketing is a quarter where they brought in a lot of qualified traffic at favorable costs.

Merchandising owns product selection and presentation. They're measured on sell-through rates, margin, inventory turns. A good quarter for merchandising is a quarter where the right products sold at the right prices with minimal markdown.

These objectives are not actually in conflict. Qualified traffic buying the right products at good margins is the outcome both departments want. But the departments are structured as if they're playing a zero-sum game -- as if the budget one gets the other loses, as if the attribution one claims the other forfeits, as if the credit for a successful quarter is a fixed sum to be divided.

So they behave accordingly. Marketing runs promotions without telling merchandising, which creates inventory stockouts on advertised products. Merchandising runs clearance events without telling marketing, which wastes ad spend driving traffic to items that are about to be discontinued. Each is rationally optimizing their own position. Each is making the other's job harder. The Nash Equilibrium they reach -- separately optimized, poorly coordinated -- is stable, measurable, and substantially worse than what cooperation would produce.

Nobody can improve by changing strategy unilaterally. Marketing can't fix the stockout problem by changing how they buy ads. Merchandising can't fix the ad waste problem by changing how they manage inventory. The only exit is changing the game.

The Misidentification Problem

Zero-sum thinking spreads through organizations in a specific pattern. It starts with budget cycles, where money genuinely is a fixed pool and each department's allocation comes at the cost of another's. This produces adversarial behavior that is, in the budget context, rational. The problem is that the behavior generalizes. Departments learn to treat information as a competitive resource, data as a proprietary asset, credit as something to be hoarded.

This is where the real damage happens.

Information, unlike budget, is not depleted by sharing. If marketing shares its customer acquisition data with merchandising, marketing still has the data. If merchandising shares its sell-through projections with marketing, merchandising still has the projections. The game they're actually playing is positive-sum: shared information allows both departments to make better decisions, producing better outcomes for both.

But they've been playing zero-sum so long they've internalized the logic. Data becomes a bargaining chip. Sharing it feels like conceding something. The rational actor in a zero-sum game doesn't concede anything voluntarily. So the data stays siloed, the decisions stay uncoordinated, and the outcome stays suboptimal.

This is what most ecommerce technology debt actually is. Not legacy code or inadequate platforms, but organizational Nash Equilibria encoded in system architecture -- separate tools for separate departments, no shared data layer, no joint decision-making infrastructure, because the humans who specified the tools were solving for departmental optimization, not system-level outcomes.

The tools reflect the game the humans thought they were playing. And the game was wrong.

When Systems Don't Talk

There's a more literal version of this in ecommerce operations that almost no one discusses clearly: the systems problem is a downstream symptom of the people problem.

When marketing's ad platform doesn't talk to the inventory management system, it's possible to spend significant money driving traffic to out-of-stock products. This happens constantly. It is treated as a systems integration problem. It is actually a coordination problem that manifested as a systems integration problem after the fact, because nobody who specified either system thought they were specifying a shared resource.

When the customer service platform doesn't talk to the order management system, the customer service agent has to ask for an order number that the customer shouldn't have to provide, because the systems that contain that information are in separate organizational fiefdoms. Each fiefdom specified its own tools. Nobody specified the relationship between the tools, because the relationship between the departments was also unspecified.

When the campaign goes live on a landing page nobody asked development to build properly, because marketing briefed the agency, the agency briefed the designer, the designer sent a file to a developer who had three days and no context, and somewhere in that chain nobody thought to talk to the ecommerce developers who actually know the platform, have the performance data, understand what converts, and were, this whole time, sitting by the phone. The strategists too. Both groups holding exactly the information that would have made the campaign work, neither of them called, the brief never arriving because the org chart said this was a marketing project and marketing handles marketing projects. The landing page loads slowly, the mobile experience is an afterthought, and the post-mortem blames the channel.

When development freezes code for peak season but marketing is still A/B testing landing pages and the CRO team is still pushing conversion experiments, three teams are each behaving rationally within their own constraints -- and collectively creating exactly the instability the code freeze was designed to prevent.

In each case, the departments are playing individually rational strategies in what they perceive as a zero-sum (or at least non-cooperative) game. In each case, the Nash Equilibrium they reach is stable, measurable, and substantially worse than the cooperative outcome would be.

The Exit Condition

Nash's mathematics pointed toward a solution: you change the payoff structure. If you can make the cooperative outcome the dominant strategy -- the one rational self-interest would select -- cooperation happens without requiring agreement or trust.

In organizational terms, this means measuring departments on joint outcomes rather than individual metrics. Marketing and merchandising measured on revenue together, not on traffic and sell-through separately. Development and product measured on site performance together, not on deployment velocity and feature count separately.

It also means building the information infrastructure that makes cooperative strategy visible. You cannot coordinate on data you don't share. You cannot optimize a joint outcome you can't measure. The technology stack has to reflect the cooperative game the organization is trying to play, not the competitive game the org chart implies.

This is not an argument for dismantling specialization. Division of labor is real and valuable. The point is that the game is positive-sum -- shared information, coordinated decisions, and joint metrics produce better outcomes for everyone -- and the most expensive thing an ecommerce operation can do is run a positive-sum game as if it were zero-sum.

The equilibrium you reach depends entirely on what game you think you're playing.

Make sure you're playing the right one.