High Output Management

High Output Management

Manage teams the way Andy Grove ran Intel: a manager's output is not what the manager does — it is what their organization produces. This skill turns High Output Management into auditable practice: production principles for knowledge work, output indicators, managerial leverage, meetings as the medium of management, clean decisions, OKRs, and a management style matched to task-relevant maturity.

Core Principle

A manager's output = the output of their organization + the output of the neighboring organizations under their influence. Nothing a manager does — emails, meetings, reviews, decisions — counts in itself; it counts only through how it raises that combined output. Since managerial time is the scarce input, the craft reduces to one question asked relentlessly: of everything I could do right now, what creates the most output per hour spent? Choose high-leverage activities; eliminate negative-leverage ones.

Scoring

Goal: 10/10. Rate management practices, calendars, and processes 0-10 against the principles below. State the current score and the specific changes needed to reach 10/10.

• 9-10: Output indicators with quality pairs, subordinate-owned 1:1s on a TRM-based cadence, delegation with task-level monitoring, OKRs that stretch without driving pay, calendar built around forecasted key events
• 7-8: Process meetings run well, but a few activity metrics, ad hoc decision meetings, or skipped training sessions remain
• 5-6: 1:1s happen irregularly, indicators track busyness, delegation is all-or-nothing, planning produces documents instead of actions
• 3-4: Management by interruption: status theater, decisions made by rank, reviews as annual surprises
• 0-2: No 1:1s, no indicators, firefighting as the operating mode, output invisible and unmeasured

Framework

1. Production Principles for Knowledge Work

Core concept: Grove's breakfast factory — deliver a three-minute egg, buttered toast, and hot coffee simultaneously, at acceptable quality and lowest cost — contains all of production: build the flow around the limiting step (the egg), fix problems at the lowest-value stage, batch where setup costs dominate, and choose deliberately between building to forecast and building to order. Every team — engineering, support, recruiting — runs a production line, whether or not anyone has drawn it.

Why it works: Knowledge work hides its assembly line, and invisible flow invites firefighting. Production thinking makes flow visible: once you know the limiting step, everything else gets scheduled around it; once defects are caught at the egg stage instead of on the customer's plate, fixing them costs a fraction.

Key insights:

• Build around the limiting step: find the longest, hardest, or most expensive stage and offset everything else from it — often code review, staging access, or one overloaded specialist
• Fix problems at the lowest-value stage: kill a flawed spec in review, not after three sprints of building on it
• Batch work with high setup cost — interviews, code reviews, interrupt handling — so the setup amortizes across the batch
• Most knowledge work is built to forecast, not to order: staff the pipeline to the forecast and accept controlled risk, as the toast goes down before the customer orders
• You cannot watch all the work: treat it as a black box and cut windows into it with a handful of indicators

Applications:

Ethical boundary: Run systems hot, not humans — production thinking optimizes the work, never treats people as interchangeable machines.

See: references/indicators-and-production.md when finding a limiting step or building a dashboard — limiting-step analysis, worked indicator pairs, leading vs trend indicators, stagger charts, and how to run an operation review.

2. Indicators That Don't Lie

Core concept: Measure output, not activity — what the team shipped that survived, not how busy it looked. Pair every quantity indicator with a quality counterpart so neither can be optimized at the other's expense, favor leading indicators that buy time to act, and report forecasts in stagger charts that show how each forecast evolved.

Why it works: People do what management measures, so an unpaired indicator is an instruction to game it. The pair closes the loop: push throughput and the escape rate exposes the corner-cutting. Leading indicators and stagger charts convert measurement from autopsy to steering.

Key insights:

• Lines of code, hours logged, and tickets touched are activity; features alive in production and problems solved are output
• Pair quantity with quality: deploys/week with change-failure rate, ticket closes with reopens, velocity with incident count
• Leading indicators (review queue age, build flakiness, on-call page rate) warn before output drops; trend indicators compare output against your own history and forecast
• A stagger chart re-forecasts the same horizon every period; reading down a column shows whether forecasting is honest, optimistic, or sandbagged
• Administrative work measures like a factory: offers per recruiter-week, invoices processed per day — always with a quality pair

Applications:

Ethical boundary: Indicators measure the work, not the worker — used for surveillance, they teach people to optimize the number instead of the output.

3. Managerial Leverage

Core concept: Leverage is the output created per unit of managerial time. High-leverage activities affect many people at once (training, well-prepared decisions, information gathering) or redirect months of work with a small, well-timed nudge. The calendar is the manager's production system: forecast the key events, batch the rest, and say no at the source when capacity is full.

Why it works: Managerial activities differ by orders of magnitude in output per hour — ninety minutes preparing a review shapes a year of someone's work, while a day of meddling subtracts output. A manager who lets the calendar happen to them spends prime hours on whatever shouted loudest.

Key insights:

• Negative leverage is real: meddling (supervising an expert in detail), waffling (stalling a decision others wait on), and a manager's visible gloom all multiply downward through the team
• Delegate the tasks you know best — monitoring them costs you least — and remember that delegation without monitoring is abdication
• Monitor at the task level, not the person level: sample like incoming inspection, deeper at low task-relevant maturity, lighter as it rises
• Forecast your limiting steps: put 1:1s, staff meetings, reviews, and planning on the calendar first and let interrupts fill around them, not the reverse
• Run below 100% load: a fully booked manager turns every surprise into a delay for everyone downstream; saying no early is cheaper than failing late
• Batch interruptions with office hours and known checkpoints instead of letting them shred maker time

Applications:

Ethical boundary: Leverage means multiplying others' output, never hoarding information or approvals until you become the bottleneck everyone must visit.

See: references/leverage-and-calendar.md when auditing a calendar or setting up delegation — the weekly leverage audit, positive/negative-leverage catalog, delegation protocol with TRM-based monitoring depth, calendar-redesign procedure, and interruption management.

4. Meetings Are the Medium of Management

Core concept: A meeting is not a symptom of bad management; it is where managerial work — gathering information, imparting it, deciding, nudging — actually happens. Process-oriented meetings (one-on-ones, staff meetings, operation reviews) run on a regular cadence and should carry the bulk of that work, roughly a quarter of the calendar. Mission-oriented meetings are ad hoc and exist solely to produce a decision.

Why it works: Regularity makes meetings cheap — standing agendas, shared expectations, zero setup cost — and starves the expensive kind: issues get caught small in 1:1s and staff meetings instead of exploding into emergency decision meetings. Grove's malorganization test: ad hoc mission-oriented meetings eating more than about a quarter of managerial time means the process is broken.

Key insights:

• The 1:1 is the subordinate's meeting: they own the agenda and bring it; the supervisor's job is to listen and learn what is really going on
• Set 1:1 frequency by task-relevant maturity, not seniority or affection: new-to-task weekly, veterans every few weeks — never less than monthly
• Both sides keep a "hold" list of non-urgent items for the next 1:1 — it batches interruptions away
• The supervisor takes the notes: writing down agreed actions signals commitment and forces follow-up
• "One more thing": after the agenda is done, ask what else is on their mind — the real issue often surfaces in the last five minutes
• Staff meetings are controlled free discussion — the manager moderates as a Socratic prodder, not a lecturer; a recurring "ad hoc" meeting is a process meeting in denial

Applications:

Ethical boundary: Hijacking the 1:1 for status extraction teaches people to stop bringing real problems — status belongs in writing.

See: references/meetings-and-one-on-ones.md when designing a meeting cadence or running a 1:1 — the full 1:1 playbook with agenda templates, staff-meeting design, operation-review roles, and meeting-cost math for when to kill a meeting.