Spotify itself never fully ran it.
In 2012, two Spotify staff members described it in a paper and two videos: squads, tribes, chapters and guilds. Small autonomous teams, each run like a startup. Coordination roles were deliberately left out, and teams were trusted to coordinate themselves.
Then, in 2020, a former Spotify engineer set the record straight. The model had been an aspiration. One of its creators admitted that even as they wrote it down, they were not working that way.
What broke was specific. Maximum autonomy with minimal coordination produced duplicated work, weak alignment and technical drift.
This is the oldest question in organisation design. Henri Fayol answered it in 1916.
Fayol named Centralisation as one of his fourteen principles of management. The part most skip is his conclusion. Absolute centralisation and absolute decentralisation are both impossible. It is always a matter of degree.
Fayol meant it to describe one thing: where a single decision should sit. Most organisations miss that and turn it into an identity. "We are a decentralised company."
McKinsey surveyed more than 1,200 managers on decision-making. 61% said that at least half the time they spend deciding is used ineffectively. The fault is rarely too much central control, or too little. It is decisions made at the wrong level.
Procurement has run this argument for decades. Centralise the categories in which buying power and compliance determine the result. Leave the ones where speed and local knowledge matter closer to the ground. Decide it category by category, not once for the whole function.
Real decentralisation depends on what you keep central. Hold the few things that keep the rest aligned, and the rest moves safely.
Which decision in your organisation is sitting one level too high, or one level too low?
