Cognitive Load

Cognitive load, in the sense used for team design, is the total amount a team has to hold in mind to do its work: the domain it serves, the systems it owns, the tools it operates, and the interfaces it must coordinate across.

Team Topologies by Matthew Skelton and Manuel Pais brought the term into wide use in software organizations, arguing that a team’s capacity for it is finite and that organizations routinely exceed it without noticing. The book separates three kinds: load intrinsic to the problem itself, load added by the environment and tooling, and load spent on learning something worth learning.

That separation is what makes the concept actionable rather than merely descriptive. Intrinsic load is what the team is for and should not be reduced. Environmental load — the deployment machinery, the credentials dance, the four monitoring systems — is the portion that can be taken away without taking away any of the team’s actual purpose, and it is what an internal developer platform exists to absorb, usually by making one supported route — a golden path — the cheapest one to take.

Read as a budget, several familiar problems become one problem. A team owning too many services, a team responsible for a domain nobody has bounded, a team that must understand five neighbouring systems to ship a change — these are all overspending, and the symptom is the same: things take longer than anyone can explain, and the explanations are individually reasonable.

Read the other way, it is also the measure a platform is judged by: a platform run as a product asks what its users no longer have to know, not how many capabilities it offers. The practical use is as a test applied before adding responsibility. Asking what a team will stop holding in mind when it takes on something new is a harder question than whether it has capacity, and a more useful one, because capacity is usually assessed in hours and the constraint is not hours.

The platform’s role in reducing the environmental portion, and where that responsibility sits between data and developer platforms, is worked through in Data Platforms and Developer Platforms.