← All Book Reviews Book Review

Team Topologies

Matthehew Skelton, Manuuel Pais

Team Topologies cover

The book that made org design a first-class engineering concern

Here is an uncomfortable truth: your organisational structure is your architecture. Not a reflection of it. Not a consequence of it. Your teams - how they're shaped, how they interact, where the boundaries sit - directly determine the systems you build, the coupling you create, and the speed at which you can change.

Matthew Skelton and Manuel Pais didn't invent this idea. Mel Conway articulated it in 1967. But Team Topologies is the first book to take Conway's Law seriously as a design tool rather than a warning - and to give engineering leaders a practical, coherent model for deliberately shaping their organisations to achieve fast flow.

If you've ever watched a reorg make delivery worse, or felt the drag of a team structure that no one chose deliberately, this book will give you language, models, and a clear path forward.


Why this book matters

Most organisations design their teams around headcount, budget, hierarchy, or historical accident. They then wonder why their systems are tangled, their delivery is slow, and their engineers are frustrated.

Team Topologies makes the case that team structure is a product decision - one that deserves the same intentionality, iteration, and feedback loops as any other part of your system design. Get it right and you unlock fast flow. Get it wrong and no amount of agile process, DevOps tooling, or re-platforming will save you.

This is not a book for HR. It is a book for CTOs, heads of engineering, and technical architects who are willing to accept that their team design choices have technical consequences - and act accordingly.


Key insights

1. Cognitive load is the hidden constraint killing your teams

Skelton and Pais introduce cognitive load as the central lens through which to evaluate team design. Their argument is this: modern software systems are too complex for any one team to hold in their heads. When teams are asked to own too much - too many services, too many technologies, too many responsibilities - quality degrades, morale drops, and incidents multiply.

The solution is not to hire smarter people or demand more hours. The solution is to design teams around what they can realistically hold in their cognitive space - and build the surrounding structures that reduce unnecessary complexity.

This reframes every conversation about team capacity from "are we working hard enough?" to "have we designed our teams so that excellence is achievable?"


2. There are only four team types - and most organisations are using them wrong

The book defines four fundamental team types, each with a distinct purpose:

  • Stream-aligned teams - the default team type, aligned to a flow of work from a customer or business segment. These are your value-delivery engines. Everything else exists to serve them.
  • Platform teams - build and maintain internal platforms that reduce cognitive load for stream-aligned teams. The key shift: a platform team should operate like a product team, with stream-aligned teams as its customers.
  • Enabling teams - temporary or semi-permanent teams that help stream-aligned teams acquire new capabilities. They exist to teach, not to do. When the capability is embedded, the enabling team moves on.
  • Complicated subsystem teams - focused on a narrow, deep area of technical complexity (maths-heavy algorithms, real-time processing, specialist security domains). Their interface should be simple even if their internals are not.

The insight that stings: most organisations have none of these team types operating with clarity. Platform teams that build platforms nobody uses. Enabling teams that never stop doing. Stream-aligned teams that are anything but - responsible for five different product areas and three internal tools simultaneously.


3. Interaction modes are design decisions, not communication preferences

The book introduces three intentional interaction modes:

  • Collaboration - two teams work closely together on a novel, uncertain problem. Powerful but expensive. Should be time-bounded.
  • X-as-a-Service - one team consumes another's capability through a clear interface, with minimal back-and-forth. Scalable and autonomous.
  • Facilitating - an enabling team helps a stream-aligned team improve a capability, then steps back.

The revolution here is treating these as deliberate choices - not defaults. When you default to constant collaboration, you create coupling. When you default to X-as-a-service too early, you miss the iteration needed to define the right interface. Knowing which mode to apply, and when to change modes, is a core leadership capability.


4. Conway's Law is not a law to be avoided - it is a lever to be pulled

Skelton and Pais introduce the concept of the "inverse Conway manoeuvre": design your team structure first, and let the architecture follow. If you want a loosely coupled, independently deployable architecture, you need teams that are themselves loosely coupled and independently deployable.

This is a profound shift. Architecture review boards, platform standards, and technology strategy all matter - but the single most powerful architectural decision you will make is how you structure your teams. If your teams are tangled, your systems will be tangled. There is no amount of good intentions, documentation, or governance that overrides this.


5. Team-first thinking changes everything about how you make organisational decisions

The book argues that the team - not the individual - is the fundamental unit of delivery. This sounds obvious until you examine how most organisations actually behave: performance reviews that assess individuals in isolation, promotions that remove your best engineers from the teams that needed them, reorganisations that scatter established team dynamics without a second thought.

A team that has worked together, established trust, built shared context, and developed its own ways of working is genuinely worth protecting. Disrupting it has a cost - often an invisible one - that rarely appears in the decision to restructure.


Thought-provoking takeaways

  • If you drew your team boundaries today, which ones would you have chosen deliberately? Which ones just happened?

  • Your platform team builds a product. Who are its customers? What do those customers say about it? When did you last ask?

  • How many teams in your organisation are genuinely stream-aligned - owning a meaningful slice of value from end to end, able to deploy independently, with clear purpose? How many are actually just squads of people working on whatever arrives in the backlog?

  • The most expensive thing in your delivery system might not be infrastructure or tooling. It might be the cognitive load tax your engineers pay every day because your teams are too big, own too much, and receive too little platform support.

  • When did your last reorganisation improve delivery performance? How do you know?


Actions - do these now

  1. Map your current team types honestly. For each team, ask: what type are they operating as? What type should they be? The gap between those answers is your improvement backlog.

  2. Measure cognitive load on your highest-value teams. Ask them directly: what do you own that you wish you didn't? What takes time you can't justify? What slows you down most? The answers will tell you more than any architectural review.

  3. Identify your enabling team needs. What capabilities are your stream-aligned teams consistently weak in - testing, security, cloud-native patterns, data engineering? Do you have teams whose job is to transfer that capability? If not, you're either not investing in it or creating permanent dependency.

  4. Define your platform team's product roadmap. If your platform team doesn't have one - a roadmap driven by the needs of the teams it serves - it is not operating as a platform team. Start one. Present it to stakeholders. Measure adoption.

  5. Review your next reorg decision against team topology principles. Before you move people around, ask: which team types will this create? What will the interaction modes be? What will the cognitive load be for the resulting teams? Make the design intent explicit.


"Team design is not just an org-chart problem. It's a technical and strategic problem."

  • Matthew Skelton & Manuel Pais