Platform Engineering

1. Embrace Platform Engineering to Conquer Complexity

Platform engineering is the discipline of developing and operating platforms.

Taming the "over-general swamp." For 25 years, software organizations have struggled with managing shared code, tools, and infrastructure, often leading to either clunky centralized systems or chaotic decentralized approaches. The proliferation of public cloud primitives and open-source software (OSS) has exacerbated this, creating an "over-general swamp" where application teams drown in "glue" code, custom configurations, and operational complexity. This glue, while initially enabling rapid delivery, becomes a massive tax on productivity during maintenance and upgrades.

Leverage through abstraction. Platform engineering offers a strategic escape from this swamp by creating robust, user-friendly platforms that abstract away underlying complexity. By limiting the diversity of OSS and cloud primitives and encapsulating them behind stable APIs, platforms significantly reduce the amount of per-application "glue" code. This approach centralizes the cost of migrations and allows application developers to focus on business value rather than infrastructure intricacies.

Beyond traditional roles. This discipline goes beyond traditional infrastructure, DevOps, or SRE roles, which often struggle with a narrow focus. Platform engineering empowers dedicated teams with a broader mission to build software-based abstractions, balancing developer experience with production support, and optimizing for the widest possible audience. It's a deliberate organizational reset to manage complexity and deliver leverage, ensuring that the work of a few platform engineers amplifies the productivity of the entire organization.

2. Build Platforms as Curated, Customer-Centric Products

Developing platforms as compelling products means that we take a customer-centric approach when deciding on the features of a platform.

Product mindset is paramount. Treating a platform as a product is fundamental to its success, shifting focus from purely technical concerns to understanding and meeting customer needs. This involves a "curated product approach," where offerings are deliberately designed and opinionated, balancing specific interaction patterns with a clear scope of what the platform will and will not provide. This curation prevents platforms from becoming feature shops that merely react to every request.

Two types of platform products. Successful platforms often manifest as "paved paths" or "railways." Paved paths layer multiple offerings into easy-to-use workflows, hiding complexity for common use cases (e.g., a streamlined deployment pipeline). Railways, conversely, address significant gaps not covered by existing products, often generalizing prototypes built by application teams (e.g., a global application configuration system). Both require deep customer understanding and a willingness to say "no" to outliers.

Empathy over mandates. Internal customers, though colleagues, present unique challenges: a smaller, often captive audience, conflicting incentives, and a tendency to forget past improvements. A product culture fosters customer empathy, moving beyond simply fulfilling explicit requests to understanding "revealed preferences"—how users actually work. This means:

• Setting customer-focused goals (adoption, satisfaction, engagement).
• Bringing users into team meetings for feedback.
• Having engineers participate in support rotations.
• Prioritizing strategic work that enables customers to self-service, rather than building bespoke features.

3. Assemble Diverse Teams with Balanced Engineering Mindsets

The way you build great platform teams is by hiring people with diverse strengths and creating a culture where each is enabled to succeed.

Beyond the software/systems divide. Effective platform engineering demands a blend of skills, moving past the traditional dichotomy of software developers (focused on new code) and systems administrators (focused on operations). Teams overly skewed towards either extreme risk building systems that are either technically brilliant but operationally unstable, or robust but lacking in innovative abstraction. The goal is to integrate these perspectives into a cohesive unit.

Four key engineering roles. A successful platform team typically comprises four distinct engineering roles, each bringing unique value:

• Software Engineers: Strong coders with a desire to understand underlying systems, comfortable with on-call duties, and a deliberate pace.
• Systems Engineers: Broad generalists in systems, adept at automation, infrastructure integration, scaling, reliability, and observability.
• Reliability Engineers: Deeply focused on reliability, excelling in incident management, SLOs, and proactive operational practices.
• Systems Specialists: Experts in specific domains like networking, kernel, or performance, brought in for deep, targeted expertise when needed.

Cultivating an inclusive culture. Recognizing and rewarding these diverse roles is crucial, as traditional promotion paths often favor pure software development. Leaders must:

• Allow role-specific titles while maintaining unified level matrices where possible.
• Adapt interview processes to assess systems knowledge and customer empathy, not just algorithmic coding.
• Actively foster a culture where all contributions are valued, preventing "us versus them" mentalities and ensuring that operational and maintenance work is seen as equally critical as new feature development.

4. Prioritize Operational Discipline to Build Trust

Trust is like the air we breathe—when it’s present, nobody really notices; when it’s absent, everybody notices.

Foundational stability is non-negotiable. Platforms must be rock-solid foundations. Without operational discipline, even the most innovative platform will erode customer trust, forcing users to become experts in its flaws. This leads to "operational hell," where neglected issues consume engineering time, stall new features, and ultimately undermine the platform's value proposition.

Essential operational practices: Three core practices are vital for sustainable operations:

• On-Call: A merged DevOps model with 24x7 coverage, aiming for fewer than five business-impacting pages per week. This ensures deep expertise is available and fosters shared ownership of operational health. Eliminating false alarms and prioritizing stability are key to sustainability.
• User Support: Platform engineers must engage directly with support issues to understand user pain points and build better products. This may involve formalizing support levels, separating non-critical support into business-hours rotations, and, at scale, hiring support specialists or creating expert networks.
• Operational Feedback: Proactive measures like Service Level Objectives (SLOs), rigorous change management, and extensive synthetic monitoring are crucial. SLOs, when used internally, drive continuous improvement, while synthetic monitoring actively simulates user interactions to detect issues before customers do.

Leadership's role in operational excellence. Leaders must champion operational excellence, making it a measurable objective (e.g., via OKRs) and empowering experienced managers to drive it. This involves providing cover for teams to prioritize stability over new features when necessary, and optimizing the onboarding of new use cases based on their tolerance for operational risk. Trust is built through consistent, reliable operation, not just promises.

5. Plan Long-Term Projects with Incremental Delivery

The best laid schemes o’ mice an’ men / Gang aft a-gley.

Navigating long timelines. Platform projects often span months or years, requiring a different planning approach than typical application development. Optimistic assumptions and a lack of clear milestones can lead to "long slogs" that frustrate teams and stakeholders. Effective planning clarifies goals, manages expectations, and ensures continuous value delivery, even for multi-year initiatives.

Structured project proposals. Every significant project should begin with a comprehensive proposal document, covering:

• Background, tenets, and guidelines: Context and core principles.
• Details of the problem: A deep dive into the issue, avoiding solution-first thinking.
• Overview of possible solutions: Exploring alternatives with pros and cons.
• Proposed solution and rationale: Justifying the chosen path, including constraints.
• Plan of action: High-level implementation roadmap, early milestones, and success metrics.
  This document fosters early buy-in and prevents misaligned expectations.

Bottom-up roadmap for realism. Beyond product roadmaps, platform teams need a high-fidelity "bottom-up" roadmap that balances new features with essential operational work. This includes:

• Keep the Lights On (KTLO): Non-discretionary operational tasks like on-call and critical incident remediation (ideally no more than 40% of workload).
• Mandates: Top-down executive directives, carefully estimated for impact.
• System Improvements: Proactive investments in reliability, efficiency, and security (e.g., reducing toil, improving testing, performance tuning, security hardening).
  Merging these ensures a realistic plan, prevents burnout, and allows for strategic allocation of resources.

6. Rearchitect Existing Platforms, Don't Just Build "v2s"

If you don’t end up regretting your early technology decisions, you probably overengineered.

Evolution over revolution. As platforms scale, their initial architectures inevitably become bottlenecks for features, efficiency, reliability, or security. The "second-system effect" often leads teams to attempt a complete "v2" rewrite, which frequently fails due to overreach, scope creep, and a mismatch between the "pioneer" mindset needed for greenfield development and the "settler/town planner" mindset of mature platform teams. Rearchitecting, or "changing the plane's engine as it's flying," is generally a more successful approach.

Mindsets and architectural maturity. Platforms evolve through stages, each requiring a different engineering mindset:

• Pioneers: Build scrappy, agile solutions for ambiguous needs, prioritizing speed.
• Settlers: Turn prototypes into scalable, useful products, building trust and understanding.
• Town Planners: Industrialize systems for efficiency, robustness, and cost-effectiveness.
  Rearchitecting allows a platform to transition through these stages, adapting its architecture incrementally without disrupting existing users or requiring a complete, risky rewrite.

Strategic rearchitecture planning. Successful rearchitectures require a deliberate, multi-year plan:

1. Think big on final goals: Envision a 3-5 year future, addressing all four system capabilities (features, efficiency, reliability, security) and potentially subsuming adjacent systems or making big bets on new OSS/vendor tech.
2. Factor in migration costs: Crucially, estimate the effort required for users to migrate, as this can dwarf the value of the new offering.
3. Determine major 12-month wins: Identify high-value, incremental deliverables that showcase the new architecture's capabilities and secure early adopter commitments.
4. Get leadership buy-in: Secure long-term executive sponsorship, as rearchitectures are expensive and require sustained commitment, often necessitating patience and iteration to find the right balance of costs and benefits.

7. Engineer Migrations for Seamless User Transitions

A platform is meant to be the stable thing that provides an enduring surface to build on, like a foundation.

Migrations as value opportunities. The constant evolution of underlying systems (security patches, vendor changes, hardware updates) necessitates frequent platform migrations. Instead of viewing these as an unavoidable tax on application teams, great platform teams see them as an opportunity to prove their value by minimizing user effort. The goal is to make migrations as transparent and painless as possible, ideally requiring no work from the customer.

Engineering for ease. Platform teams should prioritize engineering solutions to simplify migrations:

• Minimize "glue": Good platform abstractions reduce the custom integration code application teams need to maintain, making underlying changes less impactful.
• Architect for transparency: Leverage containerization, autoscaling, and advanced deployment techniques (canary, blue/green) to run multiple platform versions simultaneously, allowing for gradual, operator-driven transitions.
• Track usage metadata: Automatically collect data on who uses what, and how, to understand dependencies and target communications effectively. Centralized ownership metadata is crucial for accountability.
• Automate to avoid "clipboards": Replace manual coordination and tracking with software-driven workflows, reducing human "glue work" and making migrations more predictable and efficient.

Coordinating with care. Even with automation, some coordination is inevitable:

• Scope and prioritize changes: Limit overlapping migrations and align them with broader business initiatives.
• Communicate early and publicly: Provide ample warning, clear requirements, and dedicated support channels.
• Push through the final 20%: Acknowledge that the last, most complex applications will require sustained effort and potentially mandates, but ensure the platform team bears the bulk of the burden.
• Document on-ramps and off-ramps: Provide clear guides and tools for users to transition, and dogfood the process internally.

8. Master Stakeholder Relationships and Budget Realities

The painful truth of building internal-focused platforms is that slow delivery and an indirect tie to business value means your stakeholders have an outsized influence on your success.

Navigating the political landscape. Stakeholder management is crucial for platform success, as their leaders' perceptions can outweigh actual delivery. Misaligned stakeholders can undermine investments, divert resources, or even push for shadow platforms. This isn't just "politics"; it's a natural consequence of organizational growth and diverging priorities.

The power-interest grid. Map stakeholders based on their power and interest in your work:

• High Power, High Interest: Manage closely (e.g., CPO, major engineering leads). These are critical for buy-in and can be your biggest advocates or detractors.
• High Power, Low Interest: Keep satisfied (e.g., CTO, CFO). Engage judiciously, highlighting strategic value without overwhelming them with detail.
• Low Power, High Interest: Keep informed (e.g., individual engineering teams). These are your direct customers; product management handles most engagement.
• Low Power, Low Interest: Monitor (e.g., peripheral teams). Minimal engagement unless issues arise.
  This mapping guides where to invest your communication and negotiation efforts.

Strategic communication and compromise.

• Right transparency: Avoid oversharing minutiae, which can lead to micromanagement or disengagement. Tailor communication (e.g., Wins and Challenges) to the audience, focusing on impact.
• "Yes, with compromises": Be flexible. Accommodate high-profile or urgent requests by negotiating scope and timing, building political capital.
• "No" with rationale: When declining, provide clear, business-impact-driven reasons (product strategy, technical feasibility) and offer alternatives or guidance.
• Managing shadow platforms: Understand their drivers (urgency, novel demand, poor relationships, engineer preference). Break down silos, partner on urgent needs, and be patient, sometimes playing "cleanup crew" to integrate successful shadow platforms.

Budget management in tough times. During downturns, platforms are often seen as cost centers. Roadmaps and indirect metrics may be dismissed. To survive:

1. Tie work to protected business outcomes: Realign projects with high-profile initiatives.
2. Group work by teams: Justify headcount by project teams, not individuals, to preserve focus.
3. Propose cuts with strong opinions: Proactively suggest areas for reduction while fiercely defending essential, strategic investments.

9. Align Platform Purpose, Strategy, and Execution

The purpose of a team is not goal attainment but goal alignment.

Beyond individual team success. While individual platform teams may excel, overall organizational success hinges on their collective alignment. Misaligned teams create a "swamp" of overlapping, incompatible products that confuse users and waste resources. Alignment ensures platforms work together harmoniously, amplifying their collective leverage.

Three pillars of alignment:

• Purpose: All platform teams must share a holistic mindset, embracing the core pillars of platform engineering (product, development, breadth, operations). This means prioritizing customer experience and managing complexity, not just technical excellence.
  • Achieved by: Hiring diverse talent, fostering a collaborative culture, and adopting common practices (e.g., blameless postmortems, operability reviews).
• Product Strategy: Platforms must avoid operating in silos, where each team builds for its narrow scope. This leads to duplicated efforts and difficult cross-platform use cases.
  • Achieved by: Independent product management reporting to a cross-area leader, senior ICs advocating for cross-platform architecture, and leveraging customer surveys to identify broader needs. Judicious restructuring may be necessary when misalignment costs are high.
• Planning: Teams must support each other's critical projects, avoiding conflicts and miscommunication about timelines.
  • Achieved by: Aligning only on significant, large-scale projects (e.g., >1 developer year), not every detail. Leaders must be forthright in confronting misalignment, using a "Disagree and Commit" approach to ensure collective buy-in and prevent future bottlenecks.

Leadership drives alignment. Senior leaders play a crucial role in fostering alignment by:

• Creating a shared vision and objectives that transcend individual team goals.
• Facilitating collaborative planning processes that identify and resolve conflicts early.
• Making deliberate, transparent decisions based on overall company benefit, even when it means prioritizing one platform's needs over another's.

10. Relentlessly Manage Complexity, Not Just Eliminate It

We must design for the way people behave, not for how we would wish them to behave.

Complexity is inherent, management is key. Platforms don't eliminate complexity; they manage it to generate leverage. A successful platform team constantly addresses complexity in all its forms, recognizing that human behavior and organizational dynamics are as critical as technical solutions. Over-reliance on "human glue" or unchecked growth can inadvertently create new layers of complexity.

Avoiding accidental complexity. Platforms can inadvertently create new "accidental complexity" by over-relying on manual coordination or failing to provide adequate tooling.

• Human glue: Manual workarounds, extensive documentation, and constant coordination to bridge gaps between systems. This is inefficient and prone to error.
• Solution: Invest in automation and diagnostic tools (e.g., exposing platform metrics, synthetic monitors) to reduce the need for human intervention in routine tasks. Treat TPMs as specialists for truly complex, non-automatable scenarios.

Wrangling shadow platforms. Shadow platforms, built by application teams to solve local problems, increase overall system complexity.

• Strategy: Don't try to stop all shadow platforms; instead, be aware of them. Build trust to stay informed, embed engineers, and set expectations for future integration. When taking over a shadow platform, focus on reducing its pioneer-driven complexity while corralling new complexity within your platform team.

Controlling growth for efficiency. Unchecked team growth can exacerbate complexity by:

• Reducing the incentive to automate, leading to more manual toil.
• Removing pressure to be strategic about what's built, fostering empire-building.
• Creating a sprawling portfolio of half-baked products.
  Good platform leaders understand that leverage means not needing to grow linearly with the overall engineering team. They prioritize smart efficiency, ensuring new work in established areas is funded by existing teams, and making thoughtful investments above a baseline KTLO capacity.

11. Cultivate Platforms Your Users Genuinely Love

What’s love got to do, got to do with it? What’s love, but a second-hand emotion?

Love as a proxy for productivity. While "love" might seem an unusual metric for internal platforms, it's a powerful proxy for user productivity and satisfaction. Loved tools "just work," are enjoyable to use, and bring pleasure to getting the job done. This goes beyond simplistic metrics like adoption or efficiency, which can lead to platforms that are easy for the platform team to control but frustrating for users.

Characteristics of loved platforms:

• Intuitive interfaces and automation: Like Amazon's Apollo deployment platform, loved tools have UIs that clearly reflect system state and robust APIs/command-line tools that enable easy scripting, avoiding "clickops."
• Strong, pierceable opinionation: They offer clear "paved paths" for common use cases, making 80% of tasks effortless. Crucially, they also provide "pierceable abstractions" or escape hatches for the remaining 20% of users with special needs, preventing them from feeling trapped.
• Eliminating user friction (even with "hacks"): Sometimes, the most beloved features are those that make a complex problem "go away" for the user, even if the underlying implementation is technically "hacky." These pioneer-driven solutions prioritize user productivity above architectural purity.
• Obvious value proposition: Platforms like an internal S3 succeed when they tap into existing user awareness, offer high compatibility with existing tools, demonstrate strong engineering quality, and deliver quickly.

Beyond simplistic metrics. While CSAT scores and other surveys can provide valuable insights, they must be rigorously deployed with good sample populations and a genuine willingness to act on feedback. True "love" is often revealed through qualitative feedback and sustained, enthusiastic usage, rather than just numbers. Ultimately, cultivating platforms that users love means consistently delivering value, managing complexity on their behalf, and making their work awesome and autonomous.

Last updated: December 16, 2025