Module 1.1: Building Platform Teams
Цей контент ще не доступний вашою мовою.
Discipline Module | Complexity:
[ADVANCED]| Time: 55-65 min
Prerequisites
Section titled “Prerequisites”Before starting this module:
- Required: Engineering Leadership Track — Stakeholder communication, ADRs, mentorship
- Required: Systems Thinking Track — Understanding feedback loops and emergent behavior
- Recommended: SRE: What is SRE? — Team structures for operational disciplines
- Recommended: Experience working on or with infrastructure teams
What You’ll Be Able to Do
Section titled “What You’ll Be Able to Do”After completing this module, you will be able to:
- Design a platform team structure with the right mix of skills across infrastructure, developer tooling, and SRE
- Build hiring criteria and interview processes that identify strong platform engineering candidates
- Implement team rituals and working agreements that foster collaboration with application teams
- Evaluate team effectiveness using DORA metrics, developer satisfaction, and platform adoption rates
Why This Module Matters
Section titled “Why This Module Matters”In 2019, a fintech company hired 12 platform engineers in three months. They had budget, executive sponsorship, and clear technical vision. Within a year, half the team had quit and the other half was running a glorified ticket queue.
The problem was not technical. They had hired backend engineers who were great at building distributed systems but had no interest in developer tooling. They organized the team as a service bureau — you file a ticket, we build it. Developers hated waiting, platform engineers hated being order-takers, and nobody was doing the strategic work of designing self-service abstractions.
Building a platform team is an organizational design problem, not a hiring problem. You need the right people, in the right structure, doing the right work. Get any one of those wrong and you’ll spend millions building infrastructure nobody uses. This module gives you the frameworks to get all three right.
Did You Know?
Section titled “Did You Know?”The 2024 State of Platform Engineering report found that 78% of organizations have some form of platform team, but only 34% rate their platform as “successful.” The gap is almost entirely about team design and organizational positioning, not technology choices.
Conway’s Law — “Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations” — was published in 1967, but most companies still design their org chart without considering its impact on their architecture.
Spotify’s “model” was never meant to be copied. Henrik Kniberg, who documented squads, tribes, and guilds, has said publicly that the model was a snapshot of one moment in time, and that Spotify itself moved away from it. Yet companies still reorganize around “the Spotify model” without understanding the context.
According to Team Topologies research, the optimal cognitive load for a single team is roughly what 7-9 people can own end-to-end. Platform teams that try to own “all infrastructure” for 500 developers inevitably collapse under the weight.
Team Topologies for Platform Organizations
Section titled “Team Topologies for Platform Organizations”The Four Team Types
Section titled “The Four Team Types”Matthew Skelton and Manuel Pais defined four fundamental team types. Understanding these is essential for platform leaders because your platform organization will contain all four:
| Team Type | Purpose | Platform Context |
|---|---|---|
| Stream-aligned | Delivers value in a business domain | Application teams that consume your platform |
| Enabling | Helps stream-aligned teams adopt new capabilities | Platform evangelists, developer advocates |
| Complicated subsystem | Owns deep specialist knowledge | Database team, networking team, security team |
| Platform | Provides self-service capabilities | Your core platform team |
How These Map to Platform Organizations
Section titled “How These Map to Platform Organizations”┌──────────────────────────────────────────────────────────┐│ PLATFORM ORGANIZATION ││ ││ ┌─────────────┐ ┌─────────────┐ ┌─────────────────┐ ││ │ Platform │ │ Enabling │ │ Complicated │ ││ │ Team │ │ Team │ │ Subsystem │ ││ │ │ │ │ │ Teams │ ││ │ • IDP core │ │ • DX team │ │ • Database │ ││ │ • CI/CD │ │ • Onboard │ │ • Networking │ ││ │ • Infra │ │ • Docs │ │ • Security │ ││ │ • Self-svc │ │ • Training │ │ • ML infra │ ││ └──────┬──────┘ └──────┬──────┘ └────────┬────────┘ ││ │ │ │ ││ └────────────────┼───────────────────┘ ││ │ ││ ┌─────▼─────┐ ││ │ Thin API │ ││ │ Layer │ ││ └─────┬─────┘ ││ │ │└──────────────────────────┼────────────────────────────────┘ │ ┌──────────────┼──────────────┐ ▼ ▼ ▼ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ Stream- │ │ Stream- │ │ Stream- │ │ aligned │ │ aligned │ │ aligned │ │ Team A │ │ Team B │ │ Team C │ └──────────┘ └──────────┘ └──────────┘Interaction Modes
Section titled “Interaction Modes”Teams don’t just exist — they interact. Team Topologies defines three interaction modes:
| Mode | Description | When to Use |
|---|---|---|
| Collaboration | Two teams working closely together | Discovery phase, new capabilities |
| X-as-a-Service | One team consumes another’s API | Mature, well-understood capabilities |
| Facilitating | One team coaches another | Adoption of new tools or practices |
Critical insight for platform leaders: Most platform capabilities should evolve through these phases:
Collaboration → Facilitating → X-as-a-Service (build it (teach them (they self-serve together) to use it) independently)If you jump straight to X-as-a-Service, you build something nobody wants. If you stay in Collaboration forever, you don’t scale.
Conway’s Law and Organizational Design
Section titled “Conway’s Law and Organizational Design”The Law You Cannot Break
Section titled “The Law You Cannot Break”“Any organization that designs a system will produce a design whose structure is a copy of the organization’s communication structure.” — Melvin Conway, 1967
This is not a suggestion. It is a force of nature. Your architecture will mirror your org chart whether you plan for it or not.
The Inverse Conway Maneuver
Section titled “The Inverse Conway Maneuver”Smart organizations use Conway’s Law deliberately. Instead of letting org structure accidentally shape architecture, they design their org structure to produce the architecture they want.
| If You Want This Architecture… | …Organize Teams Like This |
|---|---|
| Microservices | Small, autonomous teams owning individual services |
| Shared platform + independent apps | Central platform team + stream-aligned teams |
| Modular monolith | Teams aligned to bounded contexts, shared codebase |
| Multi-cloud abstraction | Platform team that owns the abstraction layer |
Real Example: The Accidental Monolith
Section titled “Real Example: The Accidental Monolith”A company had 4 development teams and 1 infrastructure team. The infra team owned all shared services: database, message queue, cache, logging.
The result? Every application used the same database cluster, the same Redis instance, the same Kafka topics. Not because it was the right architecture — because it was the only architecture their org chart could produce.
When they split the infrastructure team into embedded platform engineers (one per dev team) plus a thin shared-services layer, the architecture naturally evolved toward independent service ownership.
Lesson: If you want loosely coupled services, you need loosely coupled teams. If your platform team is a monolith, your platform will be too.
Hiring Platform Engineers
Section titled “Hiring Platform Engineers”What Makes Platform Engineers Different
Section titled “What Makes Platform Engineers Different”Platform engineers are not backend engineers who got bored. They are not ops engineers who learned to code. They are a distinct specialization that requires a specific combination of skills and mindset.
| Trait | Why It Matters |
|---|---|
| Empathy for developers | Your users are internal — you must care about their experience |
| Systems thinking | Platform decisions cascade across the entire organization |
| Comfort with ambiguity | ”Build the right thing” is harder than “build the thing right” |
| Product mindset | You are building a product, not running a service |
| Teaching ability | You succeed when others can do things without you |
| Deep technical skill | Kubernetes, CI/CD, IaC, networking, security — you need breadth and depth |
| Patience with adoption | Change happens slowly; you cannot force it |
The Interview Framework
Section titled “The Interview Framework”Traditional engineering interviews test algorithm skills and system design. For platform engineers, you need to also test for empathy, product thinking, and teaching ability.
Round 1: Technical depth (standard)
- System design: “Design a multi-tenant CI/CD platform”
- Coding: Infrastructure automation, API design
- Debugging: “This Kubernetes deployment is failing — walk me through diagnosis”
Round 2: Platform thinking (unique to platform roles)
- Scenario: “Three teams want conflicting things from the platform. How do you decide?”
- Design: “How would you build a self-service database provisioning system?”
- Trade-offs: “When would you NOT build a self-service feature?”
Round 3: Empathy and communication
- Role play: “I’m a frustrated developer. My deploy has been broken for 2 hours. Help me.”
- Writing: “Write documentation for a feature you just designed”
- Teaching: “Explain Kubernetes networking to someone who’s never used containers”
Where to Find Platform Engineers
Section titled “Where to Find Platform Engineers”| Source | Quality | Notes |
|---|---|---|
| Internal transfers from SRE/DevOps | High | Already know your context |
| Internal transfers from product engineering | High | Bring user empathy, may lack infra depth |
| External hires from platform companies | High | Proven experience, expensive |
| External hires from consulting | Medium | Broad experience, may lack depth |
| Junior engineers with platform interest | Medium | Need investment, high loyalty |
| Rebranded ops engineers | Low | Often lack product mindset — but not always |
The best platform engineers often come from product teams. They have the empathy and the frustration. They know what’s broken because they’ve lived with it. Pair them with someone who has deep infrastructure skills and you have a powerful combination.
Team Size and Skill Mix
Section titled “Team Size and Skill Mix”The Two-Pizza Rule (Adapted for Platforms)
Section titled “The Two-Pizza Rule (Adapted for Platforms)”Amazon’s “two-pizza team” (6-8 people) applies to platform teams, but the skill mix is different:
Minimum viable platform team (5-7 people):
┌─────────────────────────────────────────────┐│ MINIMUM PLATFORM TEAM ││ ││ ┌──────────┐ ┌──────────┐ ┌──────────┐ ││ │ Platform │ │ Platform │ │ Platform │ ││ │ Engineer │ │ Engineer │ │ Engineer │ ││ │ (Infra) │ │ (CI/CD) │ │ (DX) │ ││ └──────────┘ └──────────┘ └──────────┘ ││ ││ ┌──────────┐ ┌──────────┐ ┌──────────┐ ││ │ Platform │ │ Tech Lead │ │ Product │ ││ │ Engineer │ │ / Manager │ │ Manager │ ││ │ (Security)│ │ │ │ (shared) │ ││ └──────────┘ └──────────┘ └──────────┘ ││ │└─────────────────────────────────────────────┘Scaling the Team
Section titled “Scaling the Team”| Org Size | Platform Team Size | Structure |
|---|---|---|
| 20-50 devs | 2-3 platform engineers | One team, broad scope |
| 50-150 devs | 5-8 platform engineers | One team with specializations |
| 150-500 devs | 10-20 across 2-3 teams | Split by domain (infra, CI/CD, DX) |
| 500-1000 devs | 20-40 across 4-6 teams | Platform organization with enabling teams |
| 1000+ devs | 40+ across 6+ teams | Full platform division with sub-teams |
Rule of thumb: Platform teams should be roughly 10-15% of your engineering organization, though this varies by platform maturity and ambition.
Embedding vs Centralized
Section titled “Embedding vs Centralized”This is one of the most consequential decisions a platform leader makes.
| Model | Pros | Cons | Best For |
|---|---|---|---|
| Fully centralized | Consistent standards, efficient specialization | Disconnected from user needs, bottleneck | Early platform, small org |
| Fully embedded | Deep user understanding, fast response | Inconsistent practices, “captured” by teams | Mature org, high autonomy |
| Hub and spoke | Standards + context, scalable | Complex coordination, dual reporting | Most organizations at scale |
| Rotating embed | Fresh perspectives, knowledge sharing | Disruptive, relationship-building overhead | Medium orgs wanting best of both |
The hub-and-spoke model is the most common pattern for mature platform organizations:
┌─────────────────────────────────────┐│ Platform Hub (Central) ││ • Architecture decisions ││ • Shared tooling & standards ││ • Cross-cutting concerns ││ • Career development │└──────────────────┬──────────────────┘ │ ┌─────────────┼─────────────┐ ▼ ▼ ▼┌─────────┐ ┌─────────┐ ┌─────────┐│ Spoke A │ │ Spoke B │ │ Spoke C ││ (Infra) │ │ (CI/CD) │ │ (DX) ││ │ │ │ │ ││ Sits │ │ Sits │ │ Sits ││ with │ │ with │ │ with ││ infra │ │ release │ │ product ││ users │ │ eng │ │ teams │└─────────┘ └─────────┘ └─────────┘Case Studies
Section titled “Case Studies”Spotify: Squads, Tribes, and the Platform They Actually Built
Section titled “Spotify: Squads, Tribes, and the Platform They Actually Built”What people think Spotify did: autonomous squads with no coordination.
What Spotify actually did: built a sophisticated internal platform (Backstage) to make squad autonomy possible.
Key decisions:
- Autonomous squads needed shared infrastructure — you cannot have 200 teams each building their own CI/CD
- Backstage (now open source) started as Spotify’s internal developer portal, solving the “where is everything?” problem
- Golden paths provided opinionated defaults — squads could deviate, but most chose the easy path
- Platform was optional — squads were never forced to use it, which meant the platform had to be genuinely good
Lesson: Autonomy requires a platform. Without shared infrastructure, autonomy becomes chaos.
Netflix: The Paved Road
Section titled “Netflix: The Paved Road”Netflix does not have a “platform team” in the traditional sense. Instead, they have multiple infrastructure teams that each own a piece of the developer experience:
- Cloud infrastructure: AWS abstraction
- Build/deploy: Spinnaker (which they open-sourced)
- Data platform: Real-time and batch processing
- Developer productivity: IDE tools, local development
Key decisions:
- “Paved road” philosophy: We build the best path. You can go off-road, but you own the consequences
- High freedom, high responsibility: Teams choose their own languages, frameworks, and patterns — but they operate what they build
- Context over control: Instead of mandates, Netflix provides rich context (memos, documentation, data) so teams make informed choices
- Senior-heavy hiring: Netflix hires experienced engineers who need less guidance, which enables more autonomy
Lesson: If you hire senior enough people and give them excellent tooling, you need less governance.
Google: The Original Platform Organization
Section titled “Google: The Original Platform Organization”Google’s infrastructure is arguably the first modern platform organization, though they did not call it that:
- Borg (precursor to Kubernetes): Centralized cluster management
- Blaze/Bazel: Unified build system
- Stubby/gRPC: Standard RPC framework
- Bigtable, Spanner, etc.: Managed storage services
Key decisions:
- Mandate standard infrastructure: At Google’s scale, heterogeneity is unmanageable
- Invest in deep abstractions: Engineers do not manage servers, they describe workloads
- Dedicated infrastructure teams: Thousands of engineers work on internal platforms
- Monorepo: A single codebase enforces consistency and enables tooling
Lesson: At sufficient scale, mandated platforms become necessary — but only if the platform is excellent.
The Pattern Across All Three
Section titled “The Pattern Across All Three”| Spotify | Netflix | ||
|---|---|---|---|
| Adoption model | Optional | Optional (paved road) | Mandatory |
| Team autonomy | High | Very high | Medium |
| Platform investment | High | High | Very high |
| Developer count | ~2,000 | ~2,500 | ~30,000+ |
| Key insight | Portal for discoverability | Context over control | Abstractions at scale |
Common Anti-Patterns
Section titled “Common Anti-Patterns”Anti-Pattern 1: The Ticket Queue
Section titled “Anti-Pattern 1: The Ticket Queue”Symptom: Platform team operates as a service bureau. Developers file tickets, platform team executes.
Why it happens: Easier than building self-service. Feels productive (look at all these tickets we closed!).
Why it fails: Does not scale. Platform engineers become order-takers instead of building leverage. Developers wait instead of moving fast.
Fix: For every ticket type that comes in more than 3 times per month, build a self-service solution.
Anti-Pattern 2: The Ivory Tower
Section titled “Anti-Pattern 2: The Ivory Tower”Symptom: Platform team builds technically impressive infrastructure that nobody asked for.
Why it happens: Engineers love interesting problems. Building the perfect abstraction is more fun than talking to users.
Why it fails: Nobody adopts it. Developers work around it. Money wasted.
Fix: Mandatory user research before every major initiative. No project starts without talking to at least 5 developers.
Anti-Pattern 3: The Reorg Carousel
Section titled “Anti-Pattern 3: The Reorg Carousel”Symptom: Platform team restructures every 6-12 months. New leader, new strategy, new priorities.
Why it happens: Lack of clear metrics for success. Each new leader has a different vision.
Why it fails: No initiative reaches maturity. Institutional knowledge is lost. Trust erodes.
Fix: Define success metrics (adoption rate, developer satisfaction, time-to-production) and evaluate against them before restructuring.
Anti-Pattern 4: The Accidental Platform Team
Section titled “Anti-Pattern 4: The Accidental Platform Team”Symptom: A team that started as “DevOps” or “infrastructure” gradually becomes a platform team without acknowledging the transition.
Why it happens: Organic growth. Somebody built a CI/CD pipeline, then a deployment tool, then a portal…
Why it fails: No product management. No user research. No intentional design. Just accumulated tools.
Fix: Acknowledge the transition. Hire or appoint a product manager. Define your platform’s scope and users.
Hands-On Exercises
Section titled “Hands-On Exercises”Exercise 1: Team Topology Mapping (45 min)
Section titled “Exercise 1: Team Topology Mapping (45 min)”Map your current organization using Team Topologies:
Step 1: List all teams that interact with infrastructure or platform services.
Step 2: Classify each team:
┌──────────────────────────────────────────────────────────┐│ Team Name: ││ Type: [ ] Stream-aligned [ ] Platform ││ [ ] Enabling [ ] Complicated Subsystem ││ ││ Current interaction modes: ││ With [Team X]: [ ] Collaboration [ ] X-as-a-Service ││ [ ] Facilitating ││ ││ Desired interaction mode (in 6 months): ││ With [Team X]: [ ] Collaboration [ ] X-as-a-Service ││ [ ] Facilitating ││ ││ Cognitive load (1-5): ___ ││ Biggest pain point: │└──────────────────────────────────────────────────────────┘Step 3: Identify mismatches:
- Teams doing platform work but classified as stream-aligned?
- Collaboration modes that should have evolved to X-as-a-Service?
- Teams with cognitive load > 3 that need enabling support?
Step 4: Draft a target topology for 6 months from now.
Exercise 2: Platform Engineer Job Description (30 min)
Section titled “Exercise 2: Platform Engineer Job Description (30 min)”Write a job description for your next platform engineering hire. Include:
- Role context: What the team builds, who they serve, what stage they are at
- Key responsibilities: 5-7 bullets that emphasize product thinking, not just technical skills
- Required skills: Technical depth + empathy + communication
- Interview process: At least one round that tests user empathy
- What success looks like: 30/60/90 day milestones
Compare your draft against this checklist:
- Does it mention “developer experience” or “user experience”?
- Does it include product/empathy skills, not just technical?
- Would a talented backend engineer understand this is a different role?
- Does it honestly describe challenges, not just opportunities?
Exercise 3: Conway’s Law Audit (30 min)
Section titled “Exercise 3: Conway’s Law Audit (30 min)”Analyze Conway’s Law in your organization:
- Draw your org chart (teams and reporting lines)
- Draw your system architecture (services and dependencies)
- Overlay them: Where does org structure mirror system structure?
- Identify friction: Where does your architecture fight your org chart?
- Propose changes: Would reorganizing one team improve your architecture?
Document your findings:
Conway's Law Audit - [Date]═══════════════════════════
Current org structure creates: [+] Good alignment: [examples] [-] Poor alignment: [examples]
Architecture changes blocked by org structure: 1. [example] 2. [example]
Proposed org change: [What] → [Expected architecture improvement]War Story: The Platform Team That Built Itself Into a Corner
Section titled “War Story: The Platform Team That Built Itself Into a Corner”Company: Mid-size SaaS company, ~300 engineers, Series D
Situation: The “Developer Productivity” team (6 people) had been building internal tools for 2 years. They had built a custom CI/CD system, a deployment tool, an internal Kubernetes abstraction, and a developer portal. All technically impressive. All tightly coupled to each other.
The problem: When they tried to hire their 7th and 8th engineers, nobody could onboard. The system had grown organically with no documentation, no API boundaries, and no clear ownership model. The original 6 engineers each had their own domain in their head, but it was not written down anywhere.
Timeline:
- Month 1-2: New hires shadow existing team members. Learning is slow because there is no documentation and the architecture is not modular enough to work on independently.
- Month 3: First new hire makes a change to the CI system that breaks the deployment tool. Nobody predicted this coupling.
- Month 4: Team lead realizes they need to stop building features and invest in platform architecture. Leadership pushes back: “We hired new people to go faster, not slower.”
- Month 6: One of the original 6 engineers burns out and leaves. Critical knowledge walks out the door.
- Month 8: After painful rearchitecting, the team splits into 2 sub-teams with clear API boundaries. Onboarding drops from 3 months to 3 weeks.
Business impact: 6 months of near-zero feature delivery. One key departure. $800K+ in lost productivity.
Root cause: They built the team like a startup (move fast, everyone knows everything) but tried to scale it like a mature organization. The transition point — around 6-8 people — is where informal knowledge sharing breaks down and you need explicit architecture, documentation, and boundaries.
Lessons:
- Document before you need to: If only one person understands a system, you have a bus-factor problem
- Design for replaceability: Every component should be ownable by someone who did not build it
- Modularize at 5-6 people: This is the inflection point where informal coordination stops working
- Invest in onboarding: The time you spend on onboarding material pays back every single hire
Knowledge Check
Section titled “Knowledge Check”Question 1
Section titled “Question 1”Your organization has 200 developers and a 4-person platform team that’s overwhelmed. They spend 80% of their time on tickets. What’s your first move?
Show Answer
Analyze the ticket queue to find the top 3-5 most common request types. Then build self-service solutions for those specific request types. The goal is to shift from executing requests to eliminating them. You also need to grow the team — 4 people for 200 developers is understaffed (2% vs the recommended 10-15%). But hiring alone will not fix the problem if the team model is a ticket queue. Fix the model first, then scale.
Question 2
Section titled “Question 2”What are the four team types in Team Topologies, and which ones would you expect to find in a platform organization?
Show Answer
The four types are: Stream-aligned (delivers business value), Platform (provides self-service capabilities), Enabling (helps teams adopt new skills), and Complicated Subsystem (owns deep specialist knowledge).
A mature platform organization contains Platform teams (core IDP, CI/CD), Enabling teams (developer advocates, onboarding), and Complicated Subsystem teams (database, networking, security). Stream-aligned teams are the platform’s users, not part of the platform org.
Question 3
Section titled “Question 3”A VP of Engineering wants to adopt “the Spotify model” for your platform organization. How do you respond?
Show Answer
Push back constructively. Explain that: (1) The “Spotify model” was a snapshot, not a prescription — Spotify itself evolved past it. (2) What actually made Spotify effective was their investment in developer tooling (Backstage), not the org chart. (3) Copying an org structure without understanding the context that created it rarely works (Conway’s Law works both ways). Instead, propose identifying your specific organizational problems and designing a structure that addresses them. Use Team Topologies as a more rigorous framework.
Question 4
Section titled “Question 4”You’re hiring your first platform engineer. Two candidates: one is a brilliant infrastructure engineer with deep Kubernetes expertise but has never worked on developer-facing tools. The other is a solid (not exceptional) backend engineer who previously built internal developer tooling and has strong communication skills. Who do you hire?
Show Answer
Hire the backend engineer with developer tooling experience. Platform engineering is fundamentally about building for other engineers, and user empathy is harder to teach than Kubernetes. The infrastructure expert may build technically impressive systems that nobody wants to use, while the tooling engineer will build things developers actually adopt. You can pair them with infrastructure depth through documentation, pairing, and training. Note: this answer changes if you already have strong DX skills on the team and specifically need infrastructure depth.
Question 5
Section titled “Question 5”Scenario: Your platform team has been in “collaboration mode” with the payments team for 8 months, jointly developing a new deployment pipeline. The payments team loves it, but two other teams are waiting for the same capability. What do you do?
Show Answer
It is time to transition from Collaboration to X-as-a-Service. Eight months of collaboration means you should have learned enough to generalize the solution. Steps: (1) Extract the payments-specific parts from the general capability. (2) Build a self-service interface that does not require platform team involvement. (3) Move to Facilitating mode with the next two teams — help them adopt the self-service tool rather than building it with them. (4) Set a deadline for fully transitioning payments to the self-service model too.
Question 6
Section titled “Question 6”What is the “Inverse Conway Maneuver” and when would you use it?
Show Answer
The Inverse Conway Maneuver is deliberately designing your org structure to produce the architecture you want. Instead of letting communication patterns accidentally shape your systems, you shape your teams to match your target architecture. You would use it when: (1) Your current architecture does not match your goals. (2) You are starting a new platform initiative and can choose team structure. (3) You are reorganizing and want architectural outcomes. Example: If you want independent microservices, organize small autonomous teams. If you want a unified platform, organize a dedicated platform team with clear API boundaries.
Question 7
Section titled “Question 7”Why does a “hub and spoke” model typically work better than fully centralized or fully embedded for platform teams at scale?
Show Answer
Hub and spoke combines the benefits of both: the hub provides consistent standards, shared tooling, and career development (benefits of centralized), while spokes provide deep user context and fast response (benefits of embedded). Fully centralized teams become bottlenecks and lose touch with user needs. Fully embedded teams lose consistency and can be “captured” by product priorities. Hub and spoke lets you maintain standards while staying close to users. The trade-off is coordination complexity — you need strong communication between hub and spokes.
Question 8
Section titled “Question 8”Your platform team just launched a new feature, but only 2 out of 15 teams are using it after 3 months. The feature works correctly. What went wrong?
Show Answer
Most likely one or more of: (1) No user research — you built what you assumed developers needed, not what they actually needed. (2) Poor discoverability — teams don’t know the feature exists. (3) High switching cost — the existing workaround is “good enough” and migration effort is not justified. (4) Missing docs/examples — teams don’t know how to adopt it. (5) Not solving the right problem — the feature addresses a problem that only 2 teams actually have. The fix depends on root cause: if it is discoverability, invest in documentation and internal marketing; if it is a poor product-market fit, go back and do user research.
Summary
Section titled “Summary”Building effective platform teams requires three things: the right people (empathetic engineers with product mindset), the right structure (team topologies matched to your architecture goals), and the right work model (self-service over ticket queues).
Key principles:
- Conway’s Law is real: Your org chart shapes your architecture whether you plan for it or not
- Team Topologies provides the vocabulary: Platform, Enabling, Complicated Subsystem teams each play a role
- Interaction modes evolve: Collaboration to Facilitating to X-as-a-Service
- Hire for empathy: Technical skills can be taught; caring about developer experience cannot
- Scale intentionally: The informal patterns that work at 5 people break at 10
What’s Next
Section titled “What’s Next”Continue to Module 1.2: Developer Experience Strategy to learn how to measure and improve the experience your platform provides.
“You don’t build platforms for engineers. You build platforms with engineers, for the organization.”