Dunbar's Number

The famous number 150 is the outer limit of a group humans can hold in their heads as a coherent social unit. For product work, the smaller numbers inside it are usually more useful:

— about 5 people for deep trust (the squad you'd actually ask for weekend help)
— about 15 for close working relationships
— about 50 for the wider group where you can still know what everyone's roughly up to

Teams that push past these numbers without restructuring hit predictable problems. Standup takes forever. Decisions stall because no one knows who owns them. Two engineers accidentally work on the same thing for a week. The communication overhead eats up the productivity gains that motivated the merger in the first place.

When a squad hits 15 people, your job is to notice that things which used to be implicit — who owns what, how conflicts get resolved — now need to be explicit and written down.

Two product squads that were each running well at 10 and 12 people get merged 'for efficiency and to reduce duplication' into a single squad of 22. The org chart looks cleaner.

Within three sprints:

— Standup has stretched to 38 minutes and nobody speaks up because everyone's waiting for their turn.
— A decision about the shared data model has been in 'still discussing' status for ten days.
— Two engineers discover they've been writing parallel implementations of the same service because the squad is now too big for the implicit 'Slack about what you're picking up' norm to work.
— M3 velocity is lower than either squad was delivering independently.

The leadership response is to hire a scrum master and add more ceremonies. The right response is to split the squad back into two.