Everyone Was Rational. The Customer Still Lost.

The bad experience nobody chose is a coordination failure, not a failure of effort.

If you've traced a bad customer experience back through your own organization and found nobody to blame, you did the analysis correctly. Every team hit its target. Nobody was negligent, nothing was concealed, and the customer still told the same story to four different people over three weeks. That result is not evidence that you looked in the wrong place. It's the signature of a specific and well-studied failure, and the absence of a culprit is the most reliable symptom it has.

It's worth knowing this has been reproduced in a laboratory for decades, because it takes the question of who is at fault off the table entirely.

The experiment that removes the culprit

At MIT, students have played the beer distribution game since the 1960s. Four players form a simple supply chain: a retailer, a wholesaler, a distributor, and a factory. Each week every player makes exactly one decision, which is how much to order from the person upstream. That's the whole game. There's no competition between them, no hidden agenda, and no way to be lazy at it.

The chain fails anyway, and it fails the same way nearly every time. Small changes in customer demand amplify as they travel upstream, and John Sterman, who published two decades of data from the game in 1989, recorded amplification as large as 700% across the four positions. Players order too much, then panic and order nothing, and the factory ends up whipsawed by a customer whose behavior barely changed.

The reason is almost boring, and that's what makes it useful. Players lose track of the supply line, meaning the orders they've already placed but not yet received. Feedback is delayed, so nobody can see the consequence of their own decision until several turns later, and by then it's tangled with everyone else's. Each player is behaving sensibly with the information in front of them. The system still produces an outcome none of them would choose.

That's the finding to carry into your own building. A bad outcome does not require a bad actor. It requires only local information, delayed feedback, and separate scorecards, and most organizations supply all three by design.

Why your teams reproduce it without meaning to

The management literature reached the same conclusion from a different direction.

W. Edwards Deming put it in one line: to optimize the whole, you have to sub-optimize the parts. A system performs on the strength of how its parts interact, not on how each performs alone, so tuning every department to its own peak degrades the thing that runs between them. Russell Ackoff stated it structurally: a system is never the sum of its parts, it's the product of their interaction.

Game theory describes the same trap through incentives. Your service team is measured on how fast it closes tickets, so it closes the ticket and passes the underlying issue along. Your product team is measured on shipping, not on the seam between its feature and the next one. Each is best-responding to its own payoffs, and no team can improve the outcome by acting alone. Economists call that an equilibrium, and its defining property is stability. It holds precisely because everyone in it is being rational.

The one thing no team is measured on is the customer's experience of the whole, so that's the thing that rots. It's a commons: a shared resource everyone draws from and nobody tends.

And the seams aren't randomly placed. They land on the boundaries between teams, because an organization ships its own communication structure, which Melvin Conway wrote down in 1968. Your customer never sees your org chart. They feel it at every handoff.

What the walk shows that the dashboards can't

This is where the fieldwork earns its place, because none of the above is visible from a status report. Every team's dashboard reads healthy while the experience running across all of them degrades. That's not a reporting failure. Each dashboard is accurately measuring its own part.

You see it by walking the path the customer walks, one stage at a time, sitting with the people doing the work. At a past pharmaceutical client, that walk surfaced five parallel submission channels for the same request, running with no single owner: formal form, email, attached spreadsheet, photocopy, screenshot. Five channels existed because five teams existed. Each had built a locally sensible route for its own work. The customer of that process experienced it as five different companies.

Nobody had decided on five channels. It was the beer game, run in an office, with forms instead of cases of beer.

What the companies that escaped it changed

None of this is new, and the field has started to measure the cost. Forrester's 2024 US Customer Experience Index found customer-experience quality at its lowest point in a decade, down for a third straight year, with only 3% of companies rated customer-obsessed and 39% of brands declining. One cause Forrester names is the failure to deliver experiences that hold together across functions. Reading that seam as the org chart surfacing in the experience is my interpretation rather than Forrester's phrase, but it's where the decline keeps coming from.

Call it the unowned journey: the path the customer takes across teams that belongs, as a whole, to nobody.

The companies pulling out of it didn't issue a values statement. They changed who was accountable. Amazon's version, documented in Working Backwards, is the single-threaded owner: one leader whose entire job is a single outcome, with authority over the teams that outcome crosses. When one person owns the experience end to end, coordination is built into the structure of the work instead of depending on goodwill.

Three levers that change the payoffs

You don't escape a bad equilibrium by asking people to be less rational. You change what the rational move is. Three levers do that, in rough order of power.

Give the journey an owner. Name one person accountable for the end-to-end experience, with real authority across the teams it crosses, rather than a seat at a coordination meeting. This has the most effect, because it creates someone for whom closing the seam is finally the rational thing to do.

Change what each team is measured on. If the service team is measured only on speed, you get speed, even when it costs the customer the resolution they called about. Put one shared metric, the journey's actual outcome, into every relevant team's scorecard, weighted heavily enough to change behavior.

Shorten the feedback delay. This is the beer game's own lesson and it's the cheapest of the three. Players fail because they can't see the consequence of their decision in time. Any change that lets a team see the downstream effect of its own handoff, that week rather than that quarter, reduces the amplification without reorganizing anything.

Reorganizing around the journey is the most durable fix and the most expensive, which is why it's rare and why it lasts.

Start with the worst handoff

You can run the diagnostic today, and it takes one question.

Find the worst handoff in your customer's experience, the point where people reliably fall through, and ask: who is rewarded for fixing it? If the honest answer is nobody, you've found your equilibrium, and no workshop or culture campaign will move it, because none of them changes a payoff.

That single question locates the problem. The redesign is the same move made deliberately: give the seam an owner, put its outcome on the scorecards around it, shorten the delay before anyone sees the damage. The same rational people then start making different choices.

The seams were never a failure of your people. They're built into the game, and the game is yours to redesign.

---