The Shape of the Kitchen

Picture one kitchen with five chefs. One is baking. Another is cleaning fish. Someone has just spilled oil on the floor. Everyone can cook, and with enough coordination everyone does. But nobody moves freely, and eventually the kitchen itself becomes the bottleneck.

Most teams don’t start there. They start small, with a tight team and a system that does just enough. One shared kitchen is the right choice. You’re shipping, you’re learning, you’re moving.

Then the business grows, and more chefs squeeze into the same kitchen. Releases slow down. Teams step on each other. What once felt nimble now feels stuck.

The instinctive answer is technical debt, or process. Often the real issue is structural: the architecture no longer matches the shape of the organisation.

You feel it first as deployment friction. A team wants to ship a small change, but it has to wait. Wait for another team to finish testing. Wait for someone else’s changes to stabilise. Wait because even a tiny change means releasing everything.

Imagine a platform offering payments and loyalty features for small businesses. The loyalty team finds a small change to how customers earn points that could lift revenue by a couple of per cent. It only touches existing users, and they want to test it quickly. But it lives in the same monolith as a compliance update for new customer onboarding. The two have nothing to do with each other. Because they share a codebase, they ship together.

The loyalty change is a finished plate that can’t leave the kitchen until someone else’s dish is done. So the experiment waits. Each day of delay costs more than projected revenue. No test runs, no feedback arrives and the team can’t learn. The problem isn’t the work. It’s the structure that forces unrelated changes to move together.

Giving each team its own kitchen changes that. With its own service, the loyalty team could release on its own terms, run the test and improve the feature before the week is out. It would also own the result from end to end.

Structure changes behaviour. When teams can move without waiting, they test more, listen sooner and improve faster. Not because anyone mandated it, but because the system allows it.

But separate kitchens aren’t free. Each one needs its own tools, its own cleanup, its own supply of fresh ingredients. In software, that means good testing, clear ownership, reliable observability and solid deployment pipelines. If any of those fall short, a service breaks, or worse, breaks others without anyone noticing.

A monolith concentrates complexity in one place. Microservices distribute it. You don’t reduce complexity; you move it. If the teams aren’t ready, you don’t get autonomy. You get drift, inconsistency and a lot of mess to clean up.

So when do you split? You don’t need five kitchens on day one. One shared kitchen works fine until it doesn’t. If teams ship smoothly, ownership is clear and feedback loops are tight, stay where you are.

But when a change that takes a day to make waits a week to ship, the queue to cook has grown longer than the cooking itself. That’s when it’s worth asking whether the kitchen still fits.

A team rarely moves faster than its kitchen allows.


Further reading:

Martin Fowler’s Monolith First and Microservice Premium, on when one system is enough and what splitting it costs.