Book summary “An Elegant Puzzle”: summary

Introduction

You often find yourself overwhelmed when moving from writing code to leading engineering teams. Engineering management brings constant friction, whether you face hypergrowth, tangled technical debt, or misaligned organizational boundaries. You might wonder how to scale systems without burning out or losing your technical edge.

Engineering management is not about outworking problems but about building reliable systems to handle organizations, tools, approaches, culture, and careers.

Will Larson provides a practical framework for these challenges in An Elegant Puzzle. The book examines organizational design, technical migrations, leadership approaches, culture building, and career development.

As you read, you will find answers to key operational questions:

• How do you structure growing teams to avoid communication overhead?
• When should you tackle technical debt through migrations versus rewrites?
• How do you balance inclusion and opportunity while scaling team culture?
• What systems keep your own career and time sustainable?

1. Organizations

Engineering teams must be sized according to strict constraints and managed through systematic states rather than reactive ad-hoc support.

Direct-support managers should handle six to eight engineers, and managers-of-managers should oversee four to six managers. Teams smaller than four individuals fail to function as proper abstractions and become too fragile due to vacations and departures. Furthermore, a two-tier 24/7 on-call rotation requires at least eight engineers.

Will Larson notes that he has sponsored teams of one or two people and regretted it every single time because they function indistinguishably from individuals and easily collapse back into maintaining technical debt.

You must follow concrete sizing rules:

• keep teams to six to eight members during steady state;
• to create a new team, grow an existing team to eight to ten, and then bud into two teams of four or five;
• never create empty teams;
• never leave managers supporting more than eight individuals.

Staying on the path to high-performing teams

Teams exist on a continuum of four distinct performance states, and each state requires a specific systemic fix rather than tactical ad-hoc support. The four states are falling behind, treading water, repaying debt, and innovating. Managers must apply the correct system solution for a given state rather than burning out trying to fix things tactically.

You can steer teams using specific structural interventions:

• identify whether the team is falling behind, treading water, repaying debt, or innovating;
• when falling behind, hire net new people to increase overall capacity until reaching treading water;
• when treading water, consolidate efforts and reduce concurrent work to limit work in progress;
• when repaying debt, add time to allow compounding value to grow without panicking stakeholders;
• when innovating, maintain enough schedule slack to build quality into the work and prevent backtracking.

A case against top-down global optimization

Disassembling high-performing teams to reallocate surplus staff to other priorities destroys productivity and incurs severe re-gelling costs. Global optimization assumes workers are interchangeable parts, but high-performing teams take years to gel. Because teams have high fixed costs, removing even a single person can break them, while shifting scope preserves the team unit and avoids restarting the gelling clock.

You can maintain operational momentum without breaking units:

• preserve gelled teams instead of breaking them apart for global optimization;
• shift scope incrementally to teams with spare capacity rather than moving people around;
• rotate individuals for a fixed period into an area that needs help while retaining their home team identity.

Productivity in the age of hypergrowth

Hypergrowth rapidly destroys engineering productivity unless managers deliberately combat entropy through concentrated hiring, minimized interruptions, and flexible software design. As headcount doubles every six months, training obligations, interview loads, on-call fragmentation, and deployment outages consume the capacity of trained engineers. Left unmanaged, three new hires can yield less output than a single trained engineer due to training overhead and support drag.

Larson describes his experience at Uber, where headcount grew from 200 to 2,000 over two years, doubling every six months, leading to the maxim that process is always six months behind headcount.

You protect team output during rapid expansion through deliberate guardrails:

• concentrate hiring periods for teams rather than spreading hiring evenly;
• rotate engineers off interviewing every three or four months if they exceed three interviews a week;
• funnel interruptions into a small area like ticket queues and utilize ownership registries;
• design software to fail open, layer policy, and keep interfaces generic to skip migrations.

Where to stash your organizational risk?

Large organizations inevitably accumulate systemic organizational debt and risk, which leaders must explicitly prioritize rather than trying to fix everything at once. Organizational debt includes problems like biased interview processes and toxic team cultures. Because these risks linger as low-priority items until they explode, leaders must explicitly select a few areas for progress with their manager, document an agreement, and deliberately accept doing the rest poorly.

You can manage accumulated risk systematically:

• identify a few organizational risk areas to improve in agreement with your manager;
• write up an explicit plan defining what reasonable progress looks like;
• keep unfixable or low-priority risks close rather than delegating doomed responsibilities onto healthy teams.

Succession planning

Succession planning requires systematically auditing one's own daily responsibilities, identifying operational gaps, and closing them to build an enduring organization. Leaders frequently become single points of failure by filling dozens of small operational holes. By examining calendars, meeting roles, recurring processes, support channels, and external relationships, a leader can map out their true scope and mentor others to take over critical tasks.

You execute sustainable handovers through structured reviews:

• audit calendar meetings and note explicit or diplomatic roles;
• review non-meeting tasks like interviewing and candidate closing;
• look back over six months for recurring processes and document your role;
• analyze team member dependencies and inbound chat requests;
• filter identified gaps into easy gaps closing in under four hours and riskiest gaps requiring months of effort.

Select project leads

To improve organizational health, execution capability, and inclusivity, companies must democratize access to critical projects through a structured and transparent application process rather than relying on a small group of favorites. Critical projects are scarce and heavily desired, while other projects are abundant. Leaders should define project scope in a short document specifying time commitment, application requirements, and selection criteria, then announce it publicly, allowing private applications and giving at least three days to apply.

You run transparent project selection processes step by step:

• define the project’s scope and goals in a short document including time commitment, requirements to apply, and selection criteria;
• announce the project to a public email list, at an all-hands, or over Slack;
• allow folks to apply in private so applicants do not see who else has applied;
• give at least three working days for people to apply;
• nudge folks to apply who you think would be good candidates but who might not self-select;
• select a project leader based on the selection criteria and privately confirm their ability to commit;
• sponsor the project leader by finding someone who has successfully completed a similar project to be their advisor;
• notify other individuals who applied that they were not selected and provide feedback;
• kick off the project, notifying the same folks who received the application announcement about the leader, sponsor, and plan;
• record the project, selected leader, and sponsor into a public spreadsheet linked to a project brief.

2. Tools

Leaders can drive broad organizational change and maintain alignment without micromanagement by employing structured tools like systems thinking, strategies, metrics, migrations, and controls.

Systems thinking uses stocks, flows, and feedback loops to reveal the hidden causal structures behind organizational bottlenecks and developer velocity. Complex organizational issues are driven by accumulations called stocks and changes called flows.

• Model difficult operational problems as systems with stocks, inflows, and outflows.
• Label every flow as a rate and every stock as a quantity.
• Use modeling tools like Insight Maker or Stella to test hypotheses quickly without performing the underlying work.

Product management operates as an iterative elimination tournament structured around problem discovery, problem selection, and solution validation. When you must cover product management responsibilities, you should avoid jumping straight to execution.

• Populate the problem space using user pain, user purpose, benchmarks, cohorts, competitive advantages, moats, and compounding leverage.
• Narrow down problems by considering tournament survival, time frames, industry trends, return on investment, and learning experiments.
• Validate solutions by writing a customer letter, identifying prior art, finding reference users, preferring experimentation over analysis, and justifying switching costs.

Visions and strategies

As an organization grows past 50 people, standard project roadmaps fail to maintain alignment, requiring strategies and visions as primary coordination tools. Strategies are grounded documents addressing specific challenges through trade-offs and actions, while visions are aspirational documents describing a future where those trade-offs are no longer mutually exclusive.

An effective strategy recommended by Richard Rumelt consists of three strict sections: diagnosis, policies, and actions. The diagnosis is a thorough problem statement describing the challenge and constraints.

• Write a diagnosis describing the challenge at hand, calling out factors and constraints.
• Identify guiding policies that explain the general approach and trade-offs between competing goals.
• Apply the guiding policies to the diagnosis to determine specific actions.

A good vision document uses vivid details to illustrate a future dream without prescriptively constraining its possibilities. Visions help teams think beyond local maxima by outlining an aspirational endpoint that makes trade-offs unnecessary.

• Write a vision statement summarizing the document in one or two sentences.
• Define the value proposition for users and the company.
• List the required capabilities for the product, platform, or team.
• Identify solved constraints that will no longer limit you in the future.
• List future constraints you expect to encounter.
• Include reference materials in an appendix.
• Synthesize details into a short one-page narrative following the vision statement.

Metrics and baselines

Good goals combine four distinct kinds of numbers to provide clear targets, context, velocity, and bounds. Relying purely on numbers without baseline context makes it impossible to judge ambition or difficulty.

• State a target for where you want to reach.
• Identify a baseline for where you are today.
• Describe the current velocity as a trend.
• Set bounds for change using a time frame.

Investment goals describe a future state to reach, while baseline or countervailing metrics describe present properties to preserve, such as cost limits or alert loads.

• Define investment goals for future capabilities or performance.
• Pair investment goals with baseline metrics to narrow the solution space and guard against regressions.

At Stripe and Uber, the author managed infrastructure costs as a baseline metric to drive widespread efficiency changes. By exploring data, diving deep, attributing costs to actual teams, contextualizing via benchmarking, and sending targeted nudges, you can prompt voluntary behavioral changes.

• Explore data in a warehouse or spreadsheet to locate levers for change.
• Dive deep into the major contributors to build a mental model.
• Attribute company-level metrics to the secondary teams actually driving the performance.
• Contextualize data using benchmarking across cohorts.
• Nudge teams by sending push notifications when metrics or benchmark performance change.
• Establish baseline metrics with key teams if nudges are insufficient.
• Run monthly or quarterly reviews for teams not sustaining agreed baselines.

Migrations and controls

Software and system migrations are the sole scalable mechanism to manage technical debt as codebases and companies grow. Because small tech debt items are picked off immediately by engineers, remaining tech debt requires coordinated multi-team migrations.

• De-risk the migration by writing a design document, shopping it with edge-case teams, and embedding into the most challenging teams first.
• Enable the organization by building tooling to programmatically migrate the easy 90 percent and providing intuitive self-service documentation.
• Finish the migration by stopping the bleeding with linters, generating tracking tickets for management, and executing the long tail of unstaffed work yourself.

Uber migrated from Puppet-managed services to a fully self-service provisioning model where engineers could spin up a service in two clicks. Previously, provisioning a new service required two weeks of clock time and two days of engineering time.

Reorganizations should solve concrete structural problems rather than avoid interpersonal issues, and they must be grounded in realistic headcount projections and appropriate manager-to-engineer ratios.

• Validate that organizational change is the right tool using a checklist.
• Project headcount a year out using optimistic, natural, and historical hiring rates.
• Set target ratios of management to individual contributors, typically targeting six to eight engineers per manager.
• Identify logical teams and groups, writing crisp mission statements for each.
• Staff teams using internal candidates, internal transfers, and external hires.
• Commit to moving forward with executive buy-in.
• Roll out the change by explaining reasoning privately to heavily impacted individuals first, documenting changes, and showing empathy.

An example calculation used to determine team counts and second-degree managers involves 35 engineers and 7 engineers per manager. Controls are mechanisms like metrics, visions, strategies, roadmaps, or headcount planning used to align with other leaders.

• Identify the appropriate controls needed for your organization.
• Agree on the degree of alignment for each control using levels such as 'I'll do it', 'Preview', 'Review', 'Notes', 'No surprises', or 'Let me know'.

Narratives, sharing, and time management

Career narratives bridge your personal goals with business priorities, helping you escape artificial competition on standard promotion ladders. Writing a career narrative focusing on skill gaps and personal strengths allows you to find abundant opportunity in your current environment.

• Write down as many goals as you can for what you would like to accomplish in the next one to five years.
• Prioritize the list and pick a few goals to focus on for the next three to six months.
• Share and iterate on your goals with your manager during one-on-ones to find the intersection with business needs.
• Refresh and review the document together approximately every quarter.

Rather than issuing top-down mandates that require bandwidth and coordination, you can measure a baseline, run an experiment locally, document the solution, and share it transparently for others to adopt voluntarily.

• Model: Measure health and throughput baselines, then run an experimental practice locally with your team.
• Document: Record the problem solved, the learning process, and detailed adoption instructions in a canonical document.
• Share: Send a short email sharing your documented approach and rollout experience without demanding compliance.

Executive communication demands concise structures that respect leadership's tendency to skim, dive into details, and debug problems. The author once bombed a technical presentation at Yahoo! because he structured it like an academic talk and failed to start from the conclusion.

• Tie the topic directly to business value in one or two sentences.
• Establish a historical narrative explaining current context and next steps.
• State an explicit ask from the audience.
• Provide a data-driven diagnosis of current constraints.
• Explain the decision-making principles and mental models applied.
• Outline what is next and when it will be done.
• Return to your explicit ask to ensure you receive necessary guidance.

To fix time management, you should perform quarterly time retrospectives, prioritize long-term success over short-term quality, finish leveraged tasks, stop doing certain tasks with structured announcements, size activities backward by allocating fixed hours rather than forward by frequency, and delegate system exceptions.

• Quarterly time retrospective: categorize calendar from the past three months to figure out time investment and shuffle goal time allocation for the next quarter.
• Prioritize long-term success over short-term quality.
• Finish small, leveraged things to create future bandwidth.
• Stop doing things: identify critical work you won't do, recategorize it as organizational risk, and alert your team and management chain.
• Size backward, not forward: specify the hours you are able to dedicate to an activity and perform as many as possible within that time.
• Delegate working in the system and hand off the responsibility to handle exceptions.
• Decouple participation from productivity by evaluating meeting attendance value.
• Hire until you are slightly ahead of growth using clear organizational design.
• Calendar blocking: add three or four two-hour blocks scattered across the week.
• Get administrative support once commitments reach a high level of complexity.

3. Approaches

Effective leadership approaches rely on working through opinionated policies, explaining core constraints instead of just saying no, partnering with management, and shifting from outworking problems to systematic delegation.

Good policy is opinionated and built by identifying goals and constraints that narrow allowed actions. When people ask for exceptions, granting them undermines fairness and creates exception debt. Instead of handling one-off exceptions, you should collect escalations as test cases to update and revise constraints during scheduled policy refreshes.

The author developed an opinionated policy for team staffing across remote employees and distributed offices, establishing constraints such as keeping offices first-tier and restricting teams to at most one office, using a 60-minute commute threshold as an illustrative office constraint.

You can apply these operational practices:

• Draft opinionated policies grounded in clear goals and strict constraints.
• Collect every escalation as a test case for reconsidering constraints rather than granting one-off exceptions.
• Declare a future time for refreshing the policy to evaluate its effect before revision.
• Batch all exception requests until the scheduled policy refresh date.

Saying no

Saying no effectively as an engineering leader requires explaining team constraints regarding velocity and priorities rather than just being firm or using the word itself. For velocity disagreements, you can use a kanban board to expose delivery steps, identify core constraints, and share data on how time is spent. For priority disagreements, you should document incoming asks, establish guiding principles, and filter tasks accordingly.

The author once sat in a room with his manager and CTO discussing a massive production incident caused by alert fatigue, velocity context gaps, and centralized database limits, where he said no to firing the on-call engineer to preserve team morale during a 12-hour on-call shift duration mentioned in the context of engineer phone battery drain.

You can apply these operational practices:

• Build a kanban board describing workflow steps and who performs them to explain team velocity constraints.
• Translate execution problems into data using story points or time-spent sampling profilers.
• Document all incoming asks by blending existing planning artifacts and tickets.
• Develop guiding principles for task selection based on top-level plans and team mission.
• Test incoming asks against guiding principles and prioritize at regular planning sessions.

Your philosophy of management

Management is fundamentally an ethical profession where actions have a profound impact, and leaders should rely on strong relationships, prioritize people over process, and tackle hard problems immediately. An effective management philosophy evolves with experience and reality.

The author reflects on his early management philosophy built on the Golden Rule and leading from the front, which later expanded as he encountered complex edge cases.

You can apply these operational practices:

• Debug internal problems starting from the relationship angle.
• Ask honestly if process failures require a new process or if you are just moving food around on your plate.
• Engage difficult situations and poor relationships head-on without postponement.
• Guide decision-making by doing the right thing for the company, the team, and yourself, in that order.
• Audit your practices to separate true best practices from cargo-culted habits.

Managing in the growth plates

Leadership behaviors must match the organizational context, requiring an intense focus on execution in high-growth areas and iterative improvement and relationship building in slower-growth areas. Midsize rapidly growing companies feature growth plates where new problems emerge rapidly, demanding execution currency rather than new ideas.

The author experienced instant conflict when moving from execution at Digg during a crisis into a different organizational context after acquisition.

You can apply these operational practices:

• Treat working in growth plates or stable areas for the first time as a brand-new role.
• Provide concrete resources and reduce the existing backlog of ideas rather than adding new ideas in growth plates.
• Build rich relationships, gel the team, and focus on career development outside the growth plates.

Ways engineering managers get stuck

Engineering managers can get stuck in various traps at different career stages, such as only managing down, mistaking team size or title for impact, and disconnecting from ground truth. New managers might exclusively manage up or down, optimize locally, ignore hiring, or neglect relationships.

You can apply these operational practices:

• Reread career and management stuckness lists annually or at performance review season.
• Audit your daily activities to ensure you are not falling into traps like mistaking team size or title for impact.

Partnering with your manager

Successful upward partnership requires keeping your manager informed about your problems, progress, workload, and goals while also understanding their priorities, stress levels, and managerial pressures. You can maintain communication via shared documents, one-on-one updates, or quarterly self-reflections.

The author reflects on his first software job where he had zero one-on-ones with his remote manager for two years, leaving him without an initial mental model of how to partner with management.

You can apply these operational practices:

• Maintain a shared document detailing your current problems, progress, preferences, busyness, and goals, and keep it updated for your manager.
• Use one-on-ones to fill information gaps and support your growth areas.
• Write a regular quarterly self-reflection or career narrative and share it with your manager and peers.
• Ask your manager about their current priorities, stress levels, and how you can help lighten their load.

Finding managerial scope and direction

Career progression in management should be measured by taking responsibility for broad, complex organizational scope rather than simply accumulating more direct reports. Leaders should pursue scope by taking ownership of crosscutting, complex projects such as cost-accounting initiatives, on-call frameworks, or recruiting processes involving 50 engineering managers involved in crosscutting project initiatives.

The author became a Director of Engineering at Digg because the company and team kept shrinking, serving as a party favor rather than a true recognition of success within hypothetical title-inflation scenarios of 5 direct reports to be a manager, 20 to be a director, and 40 to be a VP.

You can apply these operational practices:

• Look for gaps in your organization or company that you can fill with broad, complex crosscutting projects.
• Cast a wide net for ideas during discovery without judgment before crafting a strategy.
• Distill complex strategy documents down to three to four core bullet points for effective communication.

Close out, solve, or delegate

Transitioning to managing managers requires abandoning the habit of outworking problems and instead evaluating incoming tasks through a strict triage framework. When faced with any incoming problem, email, dispute, or request, you must choose one of three distinct actions.

The author notes how moving into a role supporting managers generates ambient anxiety and detachment from direct team collaboration, requiring a shift in problem-solving habits.

You can apply these operational practices:

• Close out: resolve a specific ask completely and communicate it so it never comes back.
• Solve: design a norm, process, or coaching intervention to eliminate an entire category of tasks for the next six months.
• Delegate: redirect the ask to someone with specialized skills or someone who can work in the system.

4. Culture

Healthy engineering culture is built by balancing opportunity and inclusion, making your peers your first team, managing positive and negative freedoms, and eliminating hero programmer dynamics.

Opportunity and membership

Building an inclusive organization requires focusing on two distinct pillars: opportunity and membership. Opportunity ensures fulfillment through challenge via rubrics, transparent project lead selection, explicit budgets, and nudging involvement. Membership ensures psychological safety and belonging through recurring weekly events, coffee chats, Employee Resource Groups, and team lunches.

The author recalls having coffee with a coworker whose daily anxiety revolved around finding someone to eat lunch with. This illustrates how lack of membership drains creative energy.

You can apply these operational practices:

• Implement rubrics for every important people decision including promotions, hiring, and performance designations.
• Use structured approaches for selecting project leaders to distribute opportunity fairly.
• Establish explicit budgets for conferences and education rather than vague ongoing funds.
• Nudge uninvolved employees directly and show them distribution data regarding their budget usage.
• Host recurring weekly events like paper-reading groups, Employee Resource Groups, offsites, coffee chats, and team lunches during working hours.

Make your peers your first team

Managers should shift their primary team identity from direct reports to their cohort of peers. Jason Wong wrote an article titled Building a First Team Mindset exploring this theme. You can build a cross-organizational community of learning and practice senior leadership behavior by following these steps:

• Ensure awareness of each other's work through significant time investment and sharing weekly progress.
• Evolve from character to person by spending time together learning about each other, often at a team offsite.
• Referee defection by holding team members accountable for good behavior.
• Avoid zero-sum culture by recognizing impact, support, and development.
• Explicitly discuss the idea of shifting identities away from the team supported toward the team of peers.

Company culture and managing freedoms

Organizational culture can be dynamically managed by balancing positive freedoms to act and negative freedoms from interference. Each positive freedom strips a negative freedom and vice versa, known as the Paradox of Positive Liberty. In 1969, Roger Miller and Janis Joplin sang that freedom is just another word for nothing left to lose.

Tom DeMarco's book Slack suggests engineering teams generally follow standard operating procedures while always changing exactly one thing for each new project. You should ramp toward negative freedom to stabilize successful execution during steady periods. Shift toward positive freedoms to help the organization adapt during disruptions.

Kill your heroes, stop doing it harder

Relying on hero programmers who pull 70-hour weeks during crises creates a toxic environment that alienates the rest of the team. This do-it-harder mantra fails because it breeds discontent and disengages nonheroes who no longer know how to contribute. Projects fail slowly over time through accumulated choices.

Fixing broken systems requires taking responsibility, reaching closure, or stepping back to let unsustainable systems collapse. Isaac Asimov's Foundation series parallels the goal of accelerating and minimizing the collapse of an empire.

5. Careers

Career development and organizational staffing require humane interview processes, cold sourcing, rigorous performance management systems, and structured interview loops.

Working at a company consists of stable eras bridged by rapid changes such as management shifts or reorganizations. Transitions allow you to raise your skill floor, while stable periods allow you to raise your ceiling through mastery. Tracking past transitions and eras builds a rich career narrative and helps avoid lingering past the point of personal growth. You can apply specific practices to map these professional changes:

• Build a map of your past year by marking every change that meaningfully altered how you work as a transition.
• Identify the skills relied upon during transitions and the skills developed.
• Analyze the eras following transitions to identify changing values, expectations, and utilized skills.

As a rule of thumb, the author uses two years as a minimum stay duration and four years as a target stay duration.

Running a humane interview process

An effective interview process must be treated as a simple, candidate-centric system prioritizing candidate kindness, role alignment, signal detection, interviewer preparation, and continuous feedback loops. Interview processes should replace algorithmic whiteboard tests with prepared presentations, collaborative pair programming, product demos, or roleplays. You can guide this process with concrete rules:

• Be kind to the candidate by managing time and scheduling cleanly.
• Ensure all interviewers agree on the role's requirements and skill expectations.
• Identify specific signals for each interview slot using formats like prepared presentations, codebase debugging, product demos, or roleplays.
• Show up prepared to find signal.
• Deliberately express interest in candidates by having interviewers send personalized notes.
• Create feedback loops for interviewers and loop designers via pair interviews and candidate debriefs.
• Instrument and optimize the recruiting funnel like any conversion funnel.

Cold sourcing

When personal referral networks dry up, you can systematically find candidates beyond your immediate circle by executing cold sourcing on professional networks. Cold sourcing involves reaching out directly to unknown professionals through concise, polite messages. To maintain a functional cold sourcing pipeline, follow these practices:

• Join LinkedIn and build out your network by following individuals you actually know.
• Be patient with search rate limits and platform throttling.
• Use the search function to identify second-degree connections by job title or company.
• Grab email addresses from accepted profiles and send a short, polite note inviting them to coffee or a phone call with a job description link.
• Schedule and conduct chats to evaluate mutual fit and describe the process.

Around 600 network connections is the approximate number at which LinkedIn search throttling becomes easier to manage.

Hiring funnel

Operating a healthy hiring process requires instrumenting and optimizing an extended hiring funnel encompassing identification, motivation, evaluation, closing, onboarding, impact, promotion, and retention. The core hiring funnel flows through identifying candidates via inbound, sourced, or referrals; motivating them with team time and clear role definitions; evaluating them while balancing certainty, candidate experience, and efficiency; and closing them. You can manage this extended funnel using specific steps:

• Define a process for how candidates flow through the identification, motivation, evaluation, and closing phases.
• Instrument the funnel using a formal applicant tracking system to monitor conversion rates and benchmark against peer companies.
• Extend the funnel by tracking onboarding speed, employee impact, promotion velocity, and retention rates.

The P40 productivity percentile metric measures ramp-up time for new hires.

Performance management systems

Formal performance management systems are composed of career ladders, performance designations, and performance cycles. These combined systems focus company efforts on activities that drive success and provide explicit feedback to employees on how the company values their work. Career ladders describe expected evolution, designations rate performance against expectations for a period, and cycles occur periodically to assign consistent designations.

Calibration sessions ensure consistent and fair performance ratings across teams, but they require strict rules to avoid bias and politicization. Standard calibration happens at each level of the organizational tree to align ratings. To prevent calibration from turning into a bastion of bias, adopt these practices:

• Adopt a shared quest for consistency.
• Read, don't present.
• Compare against the ladder, not against others.
• Study the distribution, don't enforce it.

Career levels and designation momentum

Designation momentum is the natural tendency of a performance process to consistently produce the same evaluations for the same people despite changes in performance. This phenomenon excites high performers who keep getting good ratings, but it can demotivate them if they lack direct feedback for improvement. For those receiving poor designations, it creates frustration because it makes it difficult to determine if ratings are lagging indicators of past issues or current performance.

As a career level experiences level drift and upward mobility becomes evasive, companies perform a level split to separate the career level into two halves. Over time, level drift leads to distinct clusters of workers who reached the career level at its highest expectations versus those who reached it recently. Splitting the level allows these cohorts to inhabit distinct levels, extending career progression runways while solidifying the moat guarding access to post-career levels.

Creating specialized roles

Creating specialized roles like SREs or TPMs introduces unique challenges such as class systems and task offloading, but provides significant efficiency advantages. Specialized roles often face challenges like being treated as second-tier service roles or becoming single points of failure in a brittle organization. When managed with executive sponsorship, recruiting support, and a dedicated career ladder, specialists deliver specialized efficiency and solve capacity constraints effectively.

The author designed the initial iteration of Uber's SRE role. There is a rule of thumb threshold of 20 individuals to always create a new role if it immediately covers this many people, or reluctantly if it covers them within two years.

Designing an interview loop

Designing an interview loop requires a rigorous first-principles approach using metrics, skill identification, direct tests, and detailed rubrics.

Writing a great interview loop is almost identical to writing a great career ladder. Rather than relying on design by committee or hiring for potential, you should instrument your funnel, study role models, identify essential skills, test candidates by having them demonstrate skills rather than describe them, and avoid Boolean rubrics. You can execute this design process through specific steps:

• Metrics first.
• Understand the current loop's performance.
• Learn from peers.
• Find role models.
• Identify skills.
• Test for each skill.
• For each test, a rubric.
• Group tests into interviews.
• Run the loop.
• Review the hiring funnel.
• Schedule an annual refresh.

An interview where candidates described building a healthy team was replaced with one where they are given organizational health survey results to identify issues and propose remedies. Grouped tests occur in a typical 45-minute interview slot, and you should run a new loop 10 to 20 times before reviewing funnel metrics.

Conclusion

Engineering management requires you to connect organizational design, operational tools, leadership approaches, team culture, and career frameworks into a single coherent system. You cannot solve structural problems by simply working harder or relying on individual heroics. Instead, you must build reliable systems that scale gracefully as your organization grows.

Sustainable engineering leadership comes from replacing reactive chaos with predictable processes, clear constraints, and humane career support.

To apply these principles in your daily work, follow these rules:

1. Size your teams according to strict organizational limits and guide them through distinct maturity states.
2. Drive alignment and change using structured tools rather than micro-management.
3. Replace arbitrary refusals with transparent explanations of core technical and operational constraints.
4. Cultivate an inclusive culture that eliminates single points of failure and hero programmer dynamics.
5. Treat career management and hiring as rigorous, empathetic systems that protect long-term team health.

10 Key Ideas

1. Direct-support managers must handle six to eight engineers, while managers-of-managers oversee four to six managers, and teams smaller than four members collapse easily.
2. Hypergrowth rapidly destroys engineering productivity unless managers deliberately combat entropy through concentrated hiring, minimized interruptions, and flexible software design.
3. Complex organizational issues are driven by accumulations called stocks and changes called flows that can be modeled using systems thinking to reveal hidden bottlenecks.
4. Leaders can prompt voluntary behavioral changes across the organization by exploring data, attributing costs to actual teams, and contextualizing metrics via benchmarking.
5. Coordinated multi-team migrations and programmatically built tooling are the sole scalable mechanisms to manage growing technical debt when small items are gone.
6. Effective policies must be opinionated and built by identifying goals and constraints that narrow allowed actions while collecting escalations as test cases to update those rules.
7. Management is an ethical profession where actions have a profound impact, and leaders should rely on strong relationships, prioritize people over process, and tackle hard problems immediately.
8. Managers should shift their primary team identity from direct reports to their cohort of peers to build a cross-organizational community of learning and practice senior leadership behavior.
9. Operating a healthy hiring process requires instrumenting and extending an extended funnel encompassing identification, motivation, evaluation, closing, onboarding, impact, promotion, and retention.
10. Designing a great interview loop requires a rigorous first-principles approach using metrics, skill identification, direct tests, and detailed rubrics instead of relying on design by committee.