Key Takeaways
Design team communication around the architecture you want to build
Structure shapes software. Conway’s law describes how a system tends to reflect the communication paths of the people building it. If database changes must pass through one central team, for instance, a shared database is likely to become part of the design, whether or not that was the original plan. The reverse Conway maneuver uses this force deliberately: organize teams around the architecture the business needs, rather than expecting an org chart to produce it by itself.
Adidas changed its pattern. The sportswear company shifted most engineering work into cross-functional teams aligned with business needs, while a smaller central group supported platforms and technical evolution. The company reported a sixtyfold increase in release frequency. The lesson is not to copy its exact staffing split, but to treat team boundaries as architectural decisions.
What's useful here is the shift from seeing organization charts as administrative diagrams to seeing communication as a design constraint. Architecture plans often fail not because the technical idea is incoherent, but because no team can change the system without negotiating across many boundaries. That makes team design a form of product design. The caution is that organizational reshaping cannot guarantee a desired architecture: skills, incentives, and engineering practices still matter. Conway’s law is best treated as a powerful tendency to account for, not a mechanical law that removes the need for technical judgment.
Keep teams small, stable, and responsible for one manageable domain
Capacity is finite. The book uses teams of roughly five to nine people as a practical default, drawing on research about the limits of trust and coordination in groups. Bigger groups can work in unusually high-trust settings, but size alone does not create more effective capacity. Stable teams need time to develop shared understanding, and frequent reassignment repeatedly imposes the cost of forming new working relationships.
Bound the work, too. Cognitive load means the mental effort required to understand and act on a team’s responsibilities. At OutSystems, an eight-person engineering productivity team had accumulated work across builds, testing, delivery, and infrastructure. Splitting it into smaller teams aligned to distinct domains reduced context switching and gave each group a clearer mission. The goal is not to minimize responsibility, but to keep it within the team’s ability to own and improve it.
The strong idea is that a team has a collective attention budget, not just a headcount. Every extra system, tool, and stakeholder consumes some of that budget, often invisibly until delivery slows or people lose confidence in their own expertise. This connects organizational design to human factors: work becomes safer when people can maintain a usable picture of the system they are changing. Team-size thresholds should remain heuristics, however. Trust, task interdependence, and communication practices can matter as much as the number of people in the room.
Make stream-aligned teams the default, then support them deliberately
Organize around value flow. A stream-aligned team owns a continuing flow of work tied to a business area, product, service, or user need. It should be able to deliver and operate useful changes without routine handoffs. The other three team types exist to help it stay focused and capable:
1. Enabling teams help other teams learn a skill or adopt a practice, then step back.
2. Complicated-subsystem teams own areas that require scarce, deep specialist knowledge.
3. Platform teams offer reusable internal services that reduce other teams’ technical overhead.
Support should remove friction. Amazon’s small service teams were designed to own their services from development through operation. The principle is broader than copying Amazon’s structure: most teams should be close to the work and users, while specialist teams support them only where specialization or shared services genuinely reduce the burden.
This model is compelling because it makes team purpose legible. Instead of debating whether a group is a DevOps, infrastructure, or component team, leaders can ask what outcome it owns and how it relates to teams delivering value. The stream-aligned team is the center of gravity, but that does not mean every specialist belongs inside every team. The design challenge is deciding when expertise should be embedded and when a dedicated service or specialist group lowers total coordination cost. A four-type vocabulary clarifies that decision without making it automatic.
Split systems where business, risk, or change naturally separates
Find real seams. A fracture plane is a natural boundary where software can be separated into parts that teams can own independently. Business-domain boundaries are often a strong starting point: each part can have a coherent model and language. Other useful seams include different regulatory obligations, release frequencies, user groups, risk levels, performance demands, or team locations. These boundaries can be combined when one division is not enough.
Beware hidden monoliths. A system can be split into services yet remain tightly joined through a shared database, one all-encompassing build, or releases that must be coordinated together. That is a distributed monolith in practice. Poppulo used business domains such as email, people, and surveys to organize product teams, while also recognizing security and reporting needs that crossed those domains. A boundary is useful when it enables more independent ownership, not simply when it creates another service.
The important distinction is between decomposition that looks modular and decomposition that changes how work flows. A service boundary that requires synchronized releases or constant negotiation is a new label over old coupling. Domain boundaries are especially powerful because they connect technical ownership to the way an organization understands its business. Yet boundaries are hypotheses, not permanent truths. Markets, regulation, and product use change, so teams should assess whether a split actually reduces dependencies and cognitive load rather than celebrating the number of services created.
Choose collaboration for discovery and services for routine delivery
Match interaction to the work. Team Topologies describes three ways teams can relate. Collaboration means working closely together when discovery or joint problem-solving is valuable. X-as-a-Service means one team provides a usable service, tool, or API that another can consume with little coordination. Facilitating means one team helps another remove obstacles or build a capability.
Do not confuse contact with progress. Collaboration can uncover problems quickly, but sustained collaboration costs attention and blurs ownership. Once a solution is understood, a stable service boundary can reduce repeated coordination. The book also points to research in which groups with intermittent interaction found solutions whose average quality matched groups that interacted constantly, while still producing some of the strongest solutions. The practical test is whether the current interaction mode suits the uncertainty of the work and gives both teams clear expectations.
These modes turn vague complaints such as “communication is poor” into a more precise question: what kind of communication is needed for this work? Discovery benefits from rich exchange because assumptions are still being tested. Repeated execution benefits from a dependable interface that protects attention. Facilitation fills a different need by lending expertise without permanently taking over the work. The framework also makes friction diagnostic: excessive coordination may signal a weak service boundary, while silence during a discovery effort may indicate missing trust, skills, or shared purpose.
Treat internal platforms as products, not infrastructure dumping grounds
Make shared services easy to use. A platform team provides internal capabilities that let stream-aligned teams work more independently. Its users are other teams, so the platform needs reliable services, clear documentation, usable interfaces, and a roadmap shaped by actual needs. The thinnest viable platform is the smallest set of shared capabilities that meaningfully speeds delivery; it may be a simple guide to existing services or a more substantial technical product.
Avoid building for its own sake. Sky Betting & Gaming first created a shared configuration-management system to address operational problems. As the system expanded, shared dependencies and competing priorities made it difficult for teams to use and change independently. The response was to rethink the group as a product-oriented provider of services, including build environments and logging. A platform succeeds when it reduces users’ work, not when its feature list grows.
A platform is an organizational interface as much as a technical layer. When it works, it hides low-value complexity and lets product teams spend attention on what distinguishes their work. When it becomes a gatekeeper, it simply relocates the bottleneck. Product thinking helps counter this by asking who uses the platform, what problem it solves, and what level of reliability is worth paying for. The phrase “smallest viable” also guards against a common temptation in engineering: treating a growing internal system as inherently more valuable than a modest solution that users can readily adopt.
Use friction and production signals to evolve the organization
Treat difficulty as information. Organizational sensing means using teams’ experiences and operational data to notice changes in customer needs, technical constraints, and coordination costs. If a team repeatedly waits for another group, struggles to use an internal service, or can no longer understand its system, the discomfort may reveal a misplaced boundary or missing capability. Slowing delivery, growing work in progress, and recurring dependencies are useful warning signs.
Keep operations connected to design. TransUnion used a staged transition to bring development and operations closer: system-build and platform-build groups first collaborated, then merged, and later returned their capabilities to development and operations teams. The transition took longer than planned, but the changing responsibilities helped teams improve automation, reliability, and operational awareness. Production is not merely the destination for software; its behavior is evidence that should shape the next design decision.
This is a cybernetic view of management: an organization needs feedback channels, a way to interpret what they report, and permission to adjust its structure. Team interactions provide qualitative signals, while telemetry and delivery measures can show where those signals have operational consequences. Neither source is sufficient alone. Metrics can reward local speed while hiding downstream strain, and anecdotes can reflect temporary discomfort rather than a structural problem. The most useful response is a small, explicit experiment in team responsibilities or interaction, followed by observation of whether flow and service quality improve.
Start with the work stream, not another reorganization
Begin with the team’s work. The book’s practical starting point is to understand what teams need to own software, serve users, and keep their cognitive load manageable. Then identify the important streams of change, align stream-aligned teams to them, and determine the smallest platform that removes recurring obstacles. Finally, look for capability gaps in coaching, service management, documentation, and technical practice.
Change for a reason. At uSwitch, the team adopted Kubernetes not simply to adopt a fashionable technology, but to create a useful platform boundary and change how teams worked. That distinction matters: tools should serve an organizational purpose. Team Topologies is not a universal formula, and its structure cannot compensate for poor engineering, unhealthy culture, weak funding practices, or confused business priorities. Treat the model as a set of adaptable design choices, not a template to impose everywhere.
The sequence helps prevent a reorganization from becoming theater. A useful redesign starts with customer-facing work and the obstacles that keep teams from delivering it, then changes boundaries only where the evidence points. This resembles iterative design: make a hypothesis, alter one part of the system, and observe the consequences. The approach also implies a leadership responsibility often overlooked in org-chart exercises: explain why teams interact differently and what success should look like. Without that shared purpose, even well-chosen structures can be experienced as another unexplained rearrangement.
Analysis
Team Topologies is best understood as a practical theory of organizational architecture. Its central move is to treat team boundaries, communication patterns, and software boundaries as parts of one design problem. This is more useful than treating organization charts as neutral containers for technical work: real dependencies and coordination habits shape what systems can be built and changed.
The framework’s strength is its focus on flow and cognitive capacity. Small, durable teams can own outcomes; support groups have explicit purposes; and collaboration is treated as a costly mode suited to uncertainty rather than a permanent virtue. The interaction modes also give organizations a vocabulary for diagnosing problems. A team waiting on another team is not merely experiencing a process delay. It may be exposing an unsuitable service boundary, missing capability, or mistaken allocation of ownership.
The book is strongest when it presents the model as adaptable. Its case studies show different routes rather than one standard reorganization: Adidas shifted toward in-house product delivery, Auto Trader evolved an enabling effort into a platform function, and TransUnion changed team relationships in stages. That variety supports the claim that design depends on context, maturity, and system scale.
There are limits. Conway’s law is a tendency, not a guarantee, and the reported organizational outcomes do not isolate team design from concurrent changes in tooling, investment, leadership, or engineering discipline. Likewise, rules about team size and cognitive load are useful prompts, not precise measurement instruments. A team can be small and still overloaded, while a larger one can function well when responsibilities and working relationships are clear.
The most transferable lesson is therefore not to install a named topology wholesale. It is to make responsibilities explicit, observe where work waits or becomes cognitively expensive, and test whether a different boundary or interaction reduces that friction. Team design becomes continuous engineering: a hypothesis about how people and systems can deliver value together, revised through operational feedback.
Review Summary
Team Topologies offers insights on organizing IT teams for high performance, emphasizing a team-centric approach critical for DevOps and Agile success. It presents four fundamental team types and three interaction modes as tools for building agile organizations. While praised for its comprehensive model and practical advice, some reviewers found it repetitive and lacking in certain areas. The book is recommended for leaders in software development, though opinions vary on its depth and applicability across different organizational contexts.
People Also Read
Glossary
Conway’s law
Communication shapes system designThe observation that systems tend to reflect the communication structures of the organizations that create them. If teams are divided by technical specialty or must coordinate through a central group, the resulting software often mirrors those divisions. The law is a tendency that organizations can account for when designing teams and architectures.
reverse Conway maneuver
Shape teams for desired architectureAn organizational design approach that structures team responsibilities and communication paths to encourage a desired software architecture. Instead of assuming that existing teams can simply implement any technical blueprint, leaders align teams to the system boundaries they want to build. The approach influences the likely architecture but does not guarantee a particular result.
cognitive load
Mental effort required for workThe amount of mental effort needed to understand and perform work. In a team context, it includes the complexity of the domains, systems, tools, and responsibilities the team must keep in mind. Team Topologies uses cognitive load to help decide whether a team’s remit and software boundaries are manageable.
stream-aligned team
Team owning a value streamA cross-functional team aligned to a continuing flow of work associated with a business area, product, service, user journey, or other valuable outcome. It aims to deliver and operate changes with minimal handoffs. This is the primary team type in the model; other team types exist largely to reduce its cognitive burden or increase its capabilities.
enabling team
Team that builds others’ capabilitiesA specialist team that helps other teams acquire skills, adopt practices, or investigate technologies they do not yet know well. It typically works closely with a team for a limited period, offering guidance and learning rather than permanently taking over delivery. Its goal is to make the supported team more capable and independent.
complicated-subsystem team
Team for deep specialist systemsA team responsible for a part of a system that requires substantial specialist knowledge, such as a complex mathematical model or a video-processing component. It exists when embedding the necessary expertise in every team that uses the subsystem would be impractical. Its purpose is to contain specialist complexity and reduce the burden on those teams.
platform team
Team providing internal servicesA team that creates and operates internal services, tools, APIs, and guidance that help stream-aligned teams deliver with greater autonomy. The platform is treated as a product for internal users, with attention to reliability, usability, documentation, and service evolution. A platform should reduce work for its users rather than become a new approval gate.
fracture plane
Natural software separation pointA seam in a software system where it can be divided into parts that are more independently owned and changed. Business-domain boundaries are common fracture planes, but regulation, release cadence, risk, performance, user groups, location, or technology may also matter. The best boundary reduces dependencies and makes team ownership more practical.
team API
How others work with a teamThe complete interface through which other teams interact with a team. It includes software and services, documentation, versioning, working practices, communication channels, priorities, and expectations for requests or changes. Thinking of these elements as an API encourages teams to make collaboration predictable and straightforward for their internal users.
X-as-a-Service mode
Consume a service with little coordinationAn interaction in which one team provides a usable service, component, tool, or platform and another team consumes it with minimal ongoing collaboration. It works best when the boundary is well designed and the provider manages usability, reliability, documentation, and change. The model supports predictable delivery by reducing the need to understand implementation details.
organizational sensing
Detect change through team signalsThe ability to notice shifts in customer needs, technology, operational health, and internal coordination by using teams and their communication as sources of information. Signals include production behavior, delivery delays, recurring handoffs, and difficulties using shared services. Organizations use these signals to adjust responsibilities, boundaries, and interactions.
thinnest viable platform
Smallest useful shared platformThe smallest set of shared capabilities that meaningfully helps delivery teams work faster or with less cognitive burden. It may be simple documentation around existing services or a more developed internal product. The concept discourages building a large platform without clear user value and encourages additions only when they improve the work of consuming teams.
FAQ
What's Team Topologies about?
- Focus on Team Structures: Team Topologies by Matthew Skelton and Manuel Pais explores how to organize business and technology teams to enhance software delivery and operational efficiency.
- Framework for Team Design: It introduces a framework that emphasizes team interactions and structures, aligning them with business goals and customer needs.
- Adaptability and Flow: The book advocates for a dynamic approach to team organization, allowing teams to adapt to changing contexts and maintain a steady flow of work.
Why should I read Team Topologies?
- Improve Team Performance: The book provides insights into optimizing team structures, leading to higher performance and better alignment with business objectives.
- Real-World Case Studies: It includes numerous case studies and examples from various organizations, illustrating the practical application of the concepts discussed.
- Guidance for Modern Challenges: As organizations face increasing complexity in software delivery, this book offers strategies to navigate these challenges effectively.
What are the key takeaways of Team Topologies?
- Four Fundamental Team Types: The book identifies four essential team types: Stream-Aligned, Enabling, Complicated-Subsystem, and Platform teams, each serving a specific purpose.
- Conway’s Law Importance: It emphasizes the significance of Conway’s Law, which links organizational communication structures to system design.
- Cognitive Load Management: The authors stress managing cognitive load on teams to ensure effective operation and high performance.
What are the four fundamental team types described in Team Topologies?
- Stream-Aligned Teams: Focused on delivering value aligned with a specific business stream, operating independently and responding quickly to changes.
- Enabling Teams: Help stream-aligned teams acquire new capabilities and improve practices, acting as a support system without becoming a bottleneck.
- Complicated-Subsystem Teams: Handle complex areas of the system requiring specialized knowledge, allowing stream-aligned teams to focus on core responsibilities.
- Platform Teams: Provide internal services that reduce cognitive load on stream-aligned teams, enabling efficient software delivery.
How does Team Topologies define Conway’s Law?
- Communication Structures Influence Design: Conway’s Law states that system design is constrained by the communication structures of the organization that creates it.
- Strategic Implications: Understanding this law is crucial for effective team design, helping organizations avoid pitfalls in software development.
- Reverse Conway Maneuver: The authors introduce structuring teams to align with the desired software architecture rather than the other way around.
What is the significance of cognitive load in Team Topologies?
- Cognitive Load Management: The book stresses managing cognitive load to ensure teams can handle responsibilities without being overwhelmed.
- Team Size and Responsibilities: Suggests limiting team sizes and responsibilities to match cognitive load capacity, maintaining high performance and morale.
- Impact on Software Delivery: Managing cognitive load improves work flow and reduces bottlenecks, leading to faster and more reliable outcomes.
How do the interaction modes work in Team Topologies?
- Collaboration Mode: Involves two teams working closely to achieve a shared goal, allowing rapid learning and innovation but increasing cognitive load.
- X-as-a-Service Mode: One team provides a service to another with minimal collaboration, characterized by clear ownership and responsibilities.
- Facilitating Mode: Involves one team helping another overcome obstacles or learn new practices, typically used by enabling teams.
How can organizations implement the concepts from Team Topologies?
- Assess Current Team Structures: Evaluate existing team structures and communication patterns to identify areas for improvement.
- Adopt the Four Team Types: Implement the four fundamental team types, ensuring each team has a clear purpose and responsibilities.
- Encourage Continuous Evolution: Foster a culture of adaptability where teams can evolve structures and interactions based on changing needs.
What is the Thinnest Viable Platform (TVP) in Team Topologies?
- Definition of TVP: Refers to creating a platform just large enough to meet the needs of stream-aligned teams without unnecessary complexity.
- Evolving Over Time: The TVP should evolve as technology changes and organizational needs grow, ensuring relevance and usefulness.
- Supporting Stream-Aligned Teams: Provides a well-defined platform to reduce cognitive load, allowing focus on delivering user value.
What are some common anti-patterns in team design mentioned in Team Topologies?
- Ad Hoc Team Design: Forming teams without a clear purpose or structure, leading to inefficiencies and communication breakdowns.
- Shuffling Team Members: Constantly changing team compositions disrupts dynamics and hinders performance.
- Siloed Functional Teams: Teams based solely on functional expertise can lead to bottlenecks and delays, disconnecting from overall work flow.
What are some real-world examples from Team Topologies?
- Poppulo Case Study: Transitioned from a single development team to multiple product teams, emphasizing team autonomy and alignment with business domains.
- TransUnion Example: Merged development and operations teams to improve collaboration and operational awareness, leading to safer production changes.
- Sky Betting & Gaming: Platform Evolution team transformed into a product team, improving engagement and reducing organizational friction.
What are the best quotes from Team Topologies and what do they mean?
- "The architecture of the system gets cemented in the forms of the teams that develop it.": Highlights the relationship between team structures and software architecture, emphasizing intentional team design.
- "Organizations should design teams intentionally by asking these questions...": Encourages thoughtful consideration of team structures and alignment with business goals.
- "A good platform is 'just big enough'.": Suggests platforms should meet team needs without becoming overly complex, promoting efficiency and usability.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.