High Output Management

1. 1Name the environmental demand explicitly: what will the market, customer, or competitor require of your team next quarter or next year, stated as a specific output level (e.g., 'the Japanese DRAM producers can deliver aggressively priced, high-quality memory to the U.S. PC market').
2. 2State your present status against that demand in the same units, so the two are directly comparable.
3. 3Compute the gap between projected demand and present status — but attach the action to closing tomorrow's gap, not today's, because today's gap is already a failure of past planning.
4. 4Decide what you do TODAY to affect the future output, since the delivery time offset means action taken now lands weeks or months out; front-load energy because 'energy put in early in the process pays off tenfold and energy put in at the end pays off negative tenfold.'
5. 5Lead with your strength: when demand outstrips status in a contested area, redeploy resources toward the business you are genuinely best at rather than defending a position where you are 'second best.'
6. 6Set a focused hierarchy of objectives with paired output indicators — one for quantity of output, one for quality — so each objective is measurable and comparable (e.g., vouchers processed paired with errors found).
7. 7Plan the way a fire department plans: build a responsive team capable of handling the unanticipated, rather than pretending a plan can forecast events like globalization or the information revolution.

Intel founded itself on DRAMs and initially owned nearly the whole market. By the mid-eighties, environmental demand had changed: Japanese producers had built manufacturing prowess to deliver aggressively priced, high-quality memory into a booming U.S. PC market. Intel's present status against that demand was deteriorating — its share was whittled down and price cuts pushed the business into major losses. Rather than keep trying to close today's gap by tweaking the DRAM operation, Grove read the projected demand-vs-status gap and acted on tomorrow's: back out of the founding business entirely and refocus on the microprocessor business, where Intel judged it was strongest. Implementation meant redeploying employees, letting some go, and shuttering factories — costly action taken today for future output. The principle that governed the choice was 'lead with our strength,' because being second best in a tough environment is not good enough. Intel went on to become the largest semiconductor manufacturer in the world.

• If you are staring at today's gap and asking what decision closes it, then stop — that gap is a past planning failure; ask instead what today's action closes next period's gap.
• If projected demand exceeds your present status in an area where you are only second best, then redeploy toward the business you are best at rather than defending the weak position.
• If an objective has no paired quantity-and-quality indicator, then it is not yet a real objective — you cannot measure it, compare it, or focus effort on it.
• If you can invest effort early versus late in a program, then invest early, because early energy pays off tenfold and late energy pays off negative tenfold.
• If your plan assumes it can anticipate specific future events, then rebuild it as a fire-department plan: an energetic, responsive team ready for the unanticipated.

• Reacting to today's gap: the tell is a planning meeting spent deciding what to do about a shortfall that already happened, with no discussion of the future gap you are creating right now.
• Back-loading effort: the tell is a program that looks fine early and then absorbs frantic energy at the end — the negative-tenfold payoff of late action.
• Defending a second-best position: the tell is continued price-cutting and losses in a contested business while your genuine strength goes underfunded.
• Objectives without indicators: the tell is a list of goals no two groups could be compared on, and no one can say objectively whether the function performed well.
• Measuring only quantity: the tell is high output numbers (vouchers processed, square feet cleaned) with no paired quality check, so errors or poor work go invisible until suppliers or a senior manager complain.

• ▸Environmental demand (next period): the specific output the market/customer/competitor will require, in measurable units.
• ▸Present status: your current output in the same units, stated honestly.
• ▸Future gap: demand minus status for the period you are planning FOR, not the one just passed.
• ▸Today's action: the concrete step you take now that lands on the future gap.
• ▸Lead-with-strength check: is this the business/area you are genuinely best at? (yes/no + why).
• ▸Early-effort allocation: what energy you front-load this period to earn the tenfold payoff.
• ▸Objective (top of hierarchy): the single focused result this action serves.
• ▸Quantity indicator: the output count that measures the objective.
• ▸Quality indicator: the paired measure that guards the count against poor work.

1. 1Map your organization onto the two extremes: totally mission-oriented (each unit runs its own location, merchandising, personnel, purchasing, reporting only financials up) versus totally functional (all support pooled centrally). You will land somewhere in between.
2. 2Identify the one thing mission-oriented grouping buys you: units staying in touch with their business/product needs and initiating rapid change when those needs shift. Everything else favors functional. If a unit does not need that responsiveness, keep it functional.
3. 3Split the corporation into business divisions (mission-oriented, product marketing and product development) and functional groups (sales, manufacturing, finance, data processing) that operate as internal subcontractors serving all divisions at lower cost and greater responsiveness than outside vendors.
4. 4Assign division heads full freedom of action over their mission—like a fighting unit told to take a hill—while functional groups supply the shared 'blankets, paychecks, surveillance': manufacturing, finance, IT, sales. Roughly two-thirds of people can sit in functional units.
5. 5Test the balance against Sloan's rule: reconcile centralization and decentralization to get the best combination of responsiveness and leverage. Push toward mission when local response matters; push toward functional to capture economies of scale and spread expertise.
6. 6Stand up dual/matrix reporting so a person or unit answers both to a mission owner and to a functional owner, and add peer coordination groups (planning, advisory) so the internal-subcontractor relationships actually get negotiated rather than assumed.
7. 7Revisit the mix periodically—senior managers and middle managers at Intel re-examined it repeatedly and always returned to hybrid. Reorganize when the product line broadens and the number of things you track multiplies.

ABC Technologies, three years old, realigned into three product divisions—Super System, Ultra System, and Hyper System—each with a VP and General Manager (John Doe, William Smith, Robert Worker), all reporting to CEO Samuel Simon. The divisions were given product marketing and product development responsibilities: this is the mission-oriented half, letting each division stay close to its product area and move fast when needs change. But sales and manufacturing were deliberately kept at the corporate level under Sales VP Albert Abel and Manufacturing VP William Weary—the functional half, acting as internal subcontractors so no division rebuilds its own sales force or plant. The trigger was growth: as the product line broadened, the number of things to track multiplied, making a hybrid the only workable form. Note the pattern is identical to Intel's army analogy—the divisions are fighting units taking hills, while sales and manufacturing supply services to all of them.

• If a unit must stay in touch with its business/product area and initiate rapid change when needs shift, then group it as a mission-oriented division; otherwise keep it functional.
• If a capability shows obvious economies of scale or expertise worth spreading across the whole company (manufacturing, finance, data processing), then centralize it as a functional internal subcontractor.
• If an internal group can provide a service at less expense and with greater responsiveness than an outside vendor, then keep it in-house as a functional group.
• If you have re-examined organizational forms and no single extreme fits, then default to hybrid—there is simply no alternative for a reasonably large organization.
• If growth broadens the product line and multiplies what you must track, then it is time to split into product divisions and reassign shared functions upward.

• Going fully mission-oriented for the autonomy and losing all leverage—each unit rebuilds its own support groups, forfeiting economies of scale. Tell: duplicated finance, purchasing, or IT teams across units.
• Going fully functional for efficiency and killing responsiveness—units can no longer initiate rapid change when their product needs shift. Tell: divisions complaining they can't move without central sign-off while competitors move faster.
• Treating the compromise as a one-time decision. Tell: the org chart hasn't changed even though the product line has broadened and tracking has multiplied.
• Setting up dual reporting on paper but leaving peer coordination undefined, so functional 'subcontractors' and mission divisions never negotiate service terms. Tell: divisions bypass internal functions and hire outside reps at higher cost.
• Confusing individual output with managerial output—running your 'own' job well while the mission/functional interface goes unmanaged. Tell: a manager busy at personal tasks while neighboring teams' output stalls for lack of coordination.

• ▸Unit name — the team or division you are placing.
• ▸Responsiveness need (H/M/L) — does it must stay close to its product and change fast?
• ▸Mission grouping rationale — the one benefit: rapid, local response to its business.
• ▸Functional grouping rationale — the scale/expertise leverage gained by centralizing.
• ▸Internal-subcontractor test — can an in-house group serve cheaper AND more responsively than outside? (Y/N)
• ▸Dual reports — name the mission owner and the functional owner this unit answers to.
• ▸Peer coordination forum — the planning/advisory group where service terms get negotiated.
• ▸Sloan balance check — is the mix reconciling centralization and decentralization, or tilted? (note the tilt)
• ▸Re-examine trigger — the growth/product-line event that would force reorganization.

1. 1Map the environment's CUA (complexity, uncertainty, ambiguity) for the task at hand: is the work routine and predictable, or turbulent and fast-changing? The passages warn that globalization and the information revolution mean everything happens faster and anything that can be done will be done — this raises CUA and demands you develop a higher tolerance for disorder.
2. 2Read the individual's motivation: are they self-interested, bound by explicit agreements, or aligned with shared group values? Match the control mode to that motivation, not to your own preference.
3. 3When motivation is self-interested and the environment is stable/low-CUA, use free-market forces — measure output with hard, countable indicators (vouchers processed, square feet cleaned) so parties trade on price and results, and comparison across groups drives performance.
4. 4When mutual obligation exists but transactions can't be fully priced, use contractual obligations — spell out the exchange in advance, pair every quantity indicator with a quality counterpart (vouchers processed paired with errors found), so both sides know the terms.
5. 5When CUA is high and no contract can anticipate every case, use cultural values — shared goals and group interest — because rules and prices can't keep pace with turbulence; here you drive toward order but accept you cannot script every response.
6. 6Re-check the fit whenever the environment shifts. Intel's DRAM retreat shows that when the environment changed under attack, the old operating mode no longer fit and had to be abandoned: adapt or die.
7. 7Build slack for the mismatch you can't eliminate — as with inventory in the breakfast factory, deliberate slack lets you absorb change (higher CUA) while still meeting obligations.

Consider Intel's college recruiting as a controlled process running in a moderately predictable environment. Managers visit campuses, phone-screen candidates, then bring the strongest to the plant. The plant visit is the expensive limiting step, so control here is contractual and measured: you increase the ratio of accepted offers to invitations by screening first, minimizing use of the costly step per hire. Because CUA is moderate — graduation dates are known, but candidate acceptance is uncertain — you stagger and offset the steps working backward from graduation, building time-slack into the sequence. Contrast this with the DRAM business by the mid-eighties: aggressively priced Japanese competition raised CUA sharply, free-market forces turned against Intel, and being second best was no longer good enough. The fit broke, so Intel exited the founding business and redeployed people to microprocessors — a change of mode forced by a change of environment. The lesson: a control mode that fits a stable, staggered process fails when the environment turns turbulent.

• If the work is routine and output is a countable physical thing, then rely on free-market/measured control with paired quantity-quality indicators.
• If mutual obligation exists but the environment is stable enough to specify terms, then use contractual control and write the exchange down in advance.
• If CUA is high and no rule can anticipate the cases, then rely on cultural values and shared goals rather than prices or contracts.
• If the environment's speed or competition rises sharply, then re-test the fit immediately — a mode that fit yesterday may now leave you second best, which is not good enough.
• If you cannot eliminate unpredictability in either the demand or the supply flow, then build deliberate slack (inventory, time offsets) at the lowest-value stage.

• Keeping the old control mode after the environment changes. The tell: you're incurring major losses defending a business or method that once dominated, like Intel clinging to DRAMs under the Japanese onslaught.
• Using pure output measures with no quality counterpart, so people game the count. The tell: vouchers processed climbs while errors found by suppliers rise unnoticed.
• Applying tight, contractual, measured control to high-CUA work that can't be specified. The tell: rules keep breaking and you scramble to research problems because you have no archive of indicators.
• Accepting disorder instead of driving toward order when CUA rises. The tell: you tolerate turbulence but stop running your processes like a well-oiled factory, so nothing gets measured or improved.
• Automating or hard-wiring a mode for the sake of low cost while the environment still demands flexibility. The tell: the continuous egg-boiler produces only three-minute eggs and you can no longer serve the customer who asks for four.

• ▸Task/unit: name the activity you are controlling
• ▸CUA level: rate complexity/uncertainty/ambiguity as low / moderate / high
• ▸Individual motivation: self-interest / mutual obligation / shared values
• ▸Proposed control mode: free-market / contractual / cultural
• ▸Output indicator: the countable physical thing you measure
• ▸Paired quality indicator: the counterpart that guards against gaming
• ▸Slack built in: where and how much (inventory, time offset) you hold
• ▸Fit verdict: does mode match motivation AND CUA? yes / no
• ▸Re-check trigger: the environmental change that forces re-evaluation

1. 1Define the specific task first, not the person globally. Task-relevant maturity is per-task: the same subordinate can be high-maturity on one assignment and low on another, so re-assess each time the work changes.
2. 2Read the subordinate's TRM on this task from evidence: prior track record on similar work, demonstrated skill, and their willingness to take responsibility for the outcome — not from tenure or general competence.
3. 3Pick the matching style. Low TRM: be structured — give tight, task-oriented direction (what, when, how). Medium TRM: shift to communicating — two-way, emphasize support and mutual reasoning. High TRM: go minimal/monitoring — set objectives and check results, involve yourself little in the how.
4. 4Deliver the direction at the chosen density, then watch the response. Signs of confusion, missed steps, or stalled action mean you under-structured; signs of resentment or disengagement mean you over-structured.
5. 5Adjust in real time and re-match. As TRM rises with successful reps, deliberately withdraw structure so you don't smother a capable subordinate; if performance slips on a task, temporarily add structure back without treating it as a global demotion.
6. 6Keep the one-on-one as your instrument for reading TRM. It is the best source of organizational knowledge you have, and it's where you detect whether the current style still fits the task.

Take Grove's own use of the concept in interviews: the question he built from task-relevant maturity — 'Is it better to be a high-maturity person on a task or a low-maturity one?' — has no fixed right answer, because the correct management style depends on the match. Apply it to Intel's college recruiting. A recruiter running plant visits for the first time has low TRM on that task: you specify the sequence — phone-screen before inviting, work backward from graduation dates, offset each step — because errors here waste the most expensive step, the candidate's travel. A recruiter who has run several cycles successfully has high TRM: you hand them the objective (raise offers-per-visit) and monitor the ratio rather than dictating each screening call. Give the veteran the structured treatment and you get resentment; give the novice the minimal treatment and you burn money on unqualified plant visits. The task is identical; only the match to maturity changes what good management looks like.

• If TRM on this task is low, then be structured: direct what/when/how and inspect closely.
• If TRM is medium, then communicate: emphasize two-way exchange and support over instruction.
• If TRM is high, then monitor: set objectives, stay out of the method, check outputs.
• If the subordinate's TRM rises through successful reps, then withdraw structure deliberately — don't keep managing yesterday's maturity level.
• If performance drops on one task, then add structure to that task only; don't downgrade your whole relationship or their standing on other tasks.

• Rating the person globally instead of per-task — the tell is you use one style for everything they do regardless of the assignment.
• Anchoring style to tenure or seniority rather than demonstrated skill and willingness — the tell is a long-serving employee getting hands-off treatment on work they've never done.
• Over-structuring a high-TRM subordinate — the tell is visible resentment, disengagement, or the person going through the motions.
• Under-structuring a low-TRM subordinate — the tell is confusion, missed steps, stalled action, and expensive rework.
• Freezing the style after the initial read — the tell is you never notice maturity has climbed, so you keep micromanaging capable people and calling it thoroughness.

• ▸Task (specific): name the single assignment being delegated
• ▸Evidence of skill: prior results on similar work
• ▸Evidence of willingness: does the person take responsibility for the outcome?
• ▸TRM rating: low / medium / high on this task
• ▸Prescribed style: structured / communicating / monitoring
• ▸Match check: does chosen style fit the rating above? (yes/no)
• ▸Watch-for signal: confusion (under-structured) vs. resentment (over-structured)
• ▸Re-assess trigger: what event (rep completed, slip occurred) prompts a new read

1. 1Treat every review and every compensation decision as feedback: ask what it tells the subordinate about their progress against what the organization needs, not just what they earned.
2. 2Set the review's frequency to match the task's feedback horizon — like MBO, feedback must arrive soon after the activity it measures, so tie merit signals to short cycles (quarterly or monthly milestones), not just an annual verdict.
3. 3Judge output the way managers are judged: measure the subordinate against the results their group or influence produces, not against their individual busywork.
4. 4Calibrate the review to the person's task-relevant maturity — hands-on, corrective detail for someone immature in the task; delegated, higher-level assessment for someone experienced and proven on it.
5. 5Batch your performance reviews: set aside one block and do them one after another to reuse the mental set-up, so the quality stays consistent across people instead of degrading with fatigue.
6. 6Deliver the tough message directly — apply the friendship test before you build relationships that make you cringe when the review is harsh — because a softened, insincere review is feedback that lies.
7. 7Follow through: a review with no monitoring of the corrective actions afterward is abdication; check at the earliest stage that the subordinate is adjusting course, not after they've polished a failed direction.

Grove's MBO airport drive is the model for what accurate feedback must do. Your objective is to catch a plane in an hour; your key results are reaching towns A, B, and C at 10, 20, and 30 minutes. If you've driven 20 minutes and still haven't hit town A, that feedback tells you immediately that you're lost — soon enough to get off the highway and ask directions. A performance review or merit signal that arrives only annually is like learning you missed town A after the plane has departed: accurate, but useless for adjustment. So structure the review's key results as short, checkable milestones tied to organizational need, and deliver the read on them fast. That way compensation and appraisal function as course-correction, not as post-mortem.

• If the subordinate is immature in the specific task, then the review must be detailed and corrective (hands-on); if mature and proven on it, then assess at the delegated, results level.
• If you cringe imagining delivering a tough review to this person, then don't build the friendship — protect the honesty of the feedback.
• If your planning cycle is annual, then set the review/feedback cycle to at least quarterly or monthly, so feedback lands soon after the activity it measures.
• If you've delegated a task, then still monitor and review the outcome — delegation without follow-through is abdication and leaves the subordinate no accurate gauge.
• If a subordinate does poor work, then correct it — don't let them 'make their own mistakes' when the tuition is paid by their customers.

• Insincere or softened reviews to spare a friend — the tell: your stomach turns at the thought of a hard verdict, and the written review reads gentler than the results warrant.
• Feedback delivered too late to act on — the tell: the review covers a full year, and problems surface only after they've compounded past correction.
• Applying one review style to everyone regardless of maturity — the tell: you ask 'hands-on or hands-off?' as if there's a fixed answer, instead of 'it depends on the employee.'
• Judging the person by their individual effort rather than group output — the tell: praise for personal industriousness while the team or influenced organizations underperform.
• Reviewing only polished final results — the tell: you catch a fundamental problem after the subordinate has spent time refining the wrong thing, instead of checking the rough draft.

• ▸Subordinate & task: name the specific task being gauged, not the person in general
• ▸Organizational need: state what the org actually requires from this output
• ▸Objective: where does this person need to go (one line)
• ▸Key results / milestones: the short-interval checkpoints that show pace
• ▸Task-relevant maturity on THIS task: immature / mixed / mature
• ▸Review style dictated by maturity: hands-on-corrective or delegated-assessment
• ▸Feedback frequency: how soon after the activity the signal lands (target quarterly/monthly)
• ▸Merit signal: how compensation reflects progress against the need
• ▸Monitoring next step: what rough-stage check you'll do so the review isn't abdication

1. 1Treat training as your own task, not a delegated one. The passages are unambiguous: Grove titles the chapter 'Why Training Is the Boss's Job.' Put your name on the syllabus and stand at the front of the room yourself.
2. 2Identify job-relevant content by working backward from output. Just as the breakfast factory manager knows every step—boiling the egg, assembling, delivering—you teach the specific tasks that convert raw inputs into finished output for your team's actual work.
3. 3Build training around detecting and fixing problems at the lowest-value stage possible. Teach subordinates to catch the rotten egg on delivery, to reject a weak candidate at the campus interview, to find the defect at unit test rather than final test. Cheap correction is a teachable skill.
4. 4Run training as a continuing process, not an event. Grove learned management over twenty years of collaboration; you transfer knowledge the same way—recurring sessions tied to real problems your people are hitting now.
5. 5Teach the indicators you use to run the operation. Show subordinates the five daily numbers (forecast, variance, raw-material inventory, equipment status, manpower, quality log) so they learn to read the black box the way you do.
6. 6Model credibility by continuing to do the work while you teach. Grove wrote he kept 'running Intel' while articulating these ideas; a manager who still operates the process earns standing to instruct in it.
7. 7Use your own daily experiences as course material. Grove built his teaching from middle managers' examples 'from their daily lives'—harvest real incidents from your team and turn them into next week's lesson.

Grove himself is the worked example: he first learned the job hands-on—'getting the eggs and toast and pouring the coffee myself'—then, as the operation scaled to a national Breakfast Factory with regional warehouses and a corporate staff, the knowledge of how to run it had to be transmitted rather than performed personally. He treated teaching as a management task, convening a group of Intel middle managers who 'cheerfully accepted the role of guinea pig' and 'suffered through my first attempts to articulate these ideas.' He drew the curriculum from their real daily managerial experiences, not from abstract theory. He ran this as an ongoing collaboration over years, refining the material each pass. The output was systematic, job-relevant instruction—later distilled into the 'Why Training Is the Boss's Job' argument—delivered by someone still credibly running the operation he taught.

• If a subordinate keeps missing output, then before disciplining, ask whether you ever trained the task—training gaps look like performance gaps.
• If you can catch a defect earlier and cheaper, then make that the lesson: teach people to reject the rotten egg at delivery, not at the customer's table.
• If output is concentrated in the last week of the month rather than spread linearly, then the manager under you needs training on pacing manpower and equipment—teach the linearity indicator.
• If a task is being done well but inconsistently across your team, then standardize it through your own training rather than writing a memo—personal instruction transfers judgment, memos transfer only rules.
• If you would not personally demonstrate the task credibly, then you are not yet qualified to teach it—go do the work until you are.

• Delegating training to HR or a manual and calling it done. The tell: you cannot name what your people learned last quarter, and defects still surface at the final, most expensive stage.
• Treating training as a one-time onboarding event. The tell: the syllabus exists but hasn't changed as the work changed—no continuing process, no fresh examples from current daily problems.
• Teaching theory disconnected from the actual output. The tell: subordinates can recite principles but can't read the five daily indicators or fix a live production problem.
• Losing credibility by teaching what you no longer do. The tell: your examples are stale, people quietly discount your instruction, and the 'role model' is a figurehead who left the floor long ago.
• Skipping the messy first attempts. Grove's ideas only worked after guinea-pig managers 'suffered through' rough versions—managers who wait to teach until the material is polished never start, and the knowledge stays trapped in their heads.

• ▸Task being taught: name the specific job-relevant task, tied to real output.
• ▸Output it produces: state what finished value this task adds (the 'boiled egg' step).
• ▸Lowest-value catch point: where in the process should a defect here be detected and fixed.
• ▸Real example used: cite an actual recent incident from your team's daily work.
• ▸Who teaches: confirm you (the manager) deliver it personally, not a delegate.
• ▸Indicator taught: which measurement (forecast, variance, inventory, equipment, manpower, quality) they'll learn to read.
• ▸Demonstration step: what you will show yourself to prove current, credible competence.
• ▸Follow-up date: when this recurs—training is a continuing process, set the next session.
• ▸Learning check: how you'll verify the task is now done correctly at the right stage.

1. 1Diagnose the source of each subordinate's motivation against Maslow's hierarchy—identify whether they are driven by security, esteem, self-actualization, or the sub-forms of self-actualization (competence-driven vs. achievement-driven).
2. 2Recognize that lower needs extinguish once fulfilled while self-actualization does not; the man who reached vice president—a life-long goal—then had to find a new source of motivation. Watch for people who have satisfied esteem and now drift.
3. 3Install output indicators for the group's work (square feet cleaned, transactions processed, sales orders entered) so objectives are explicit, measurable, and comparable across groups doing the same function.
4. 4Publish those indicators so groups performing the same function in different locations can be compared against each other—the competitive spirit this engenders frequently has an electrifying effect on motivation and performance.
5. 5Match your control mode to the person's motivation and the CUA (complexity, uncertainty, ambiguity) factor of their environment, rather than defaulting to one style.
6. 6Match your management style to the subordinate's task-relevant maturity: structured/hands-on when immature in the task, two-way communication when medium, monitoring-only (objectives + monitoring) when mature.
7. 7Use operation reviews to reach people several levels apart—the presenting manager wanting to leave a good impression on the supervisor's supervisor and outside peers is itself a source of motivation.
8. 8Do not neglect training: motivation and training are the only two ways a manager impacts output, so if you are not training you are neglecting half the job.

Apply the sports analogy to a custodial group. You begin with an output indicator—square feet cleaned—which spells out the objective clearly and gives an objective measure of an administrative function. You then compare the custodial group in one major building against the group in a second building performing the same function. The published comparison engenders a competitive spirit that has an electrifying effect on the motivation each group brings to its work, with a parallel improvement in performance. The mechanism is Maslow's higher rungs: once security is met, output and competition motivate the group toward esteem and, for the achievement-driven, without limit—just as the athlete pursuing a 'personal best' works to do a little bit better this time than the time before.

• If the subordinate is immature in the task, then use hands-on structured training; if mature, then delegate and monitor only—the answer to hands-on vs. hands-off is 'it depends on the employee.'
• If motivation is self-actualization (competence- or achievement-driven), then set higher goals and competitive measures, because this drive has no limit and does not extinguish itself.
• If a person has just attained a long-sought goal (esteem satisfied), then help them find a new source of motivation, because fulfilled lower needs stop driving performance.
• If the environment is high-CUA and motivation is group-oriented, then rely on cultural values; if self-interested or the environment is simple, then use contractual/market-based control instead.
• If commonality of operational values is absent, then you cannot progress to a monitoring style or delegate—transmit shared values first, since the supervisor is accountable for the output.

• Letting a subordinate 'make his own mistakes to learn' when immature in the task—the tell is poor work whose tuition is paid by the customers, which is absolutely wrong; the supervisor, not customers, must bear the teaching.
• Assuming knowledge-economy employees are too smart to need training—the tell is customers flabbergasted by employees who fail at simple tasks like taking restaurant reservations.
• Managing motivation without indicators—the tell is you cannot compare groups or spell out objectives, so no competitive spirit forms and performance stalls.
• Applying one control mode everywhere (e.g., idealizing cultural values as utopian)—the tell is it fails where it doesn't fit, like using group-interest logic to buy tires or run a tax system.
• Reverting to the wrong style for the person's current maturity—the tell is either micromanaging a mature performer or leaving an overwhelmed person (like Mike, who quit unable to tolerate ambiguity) without structure.

• ▸Subordinate name / group: who you are assessing
• ▸Dominant Maslow need: security / esteem / self-actualization
• ▸Self-actualization type (if any): competence-driven or achievement-driven
• ▸Task-relevant maturity for this task: low / medium / high
• ▸Environment CUA factor: low / medium / high
• ▸Chosen management style: structured / two-way communication / monitoring-only
• ▸Output indicator in place: the measurable unit (e.g., square feet cleaned)
• ▸Comparison group for competitive spirit: same-function group to benchmark against
• ▸Training action this period: what you will teach, given motivation + training are the only two output levers

1. 1Assess each subordinate's task-relevant maturity before you decide how to manage them — separate skill on THIS task from general seniority; a senior person can be immature on a new task.
2. 2For an immature subordinate, go hands-on: use one-on-ones and direct instruction where you teach your skills, know-how, and ways to approach the problem, because their errors are 'paid for' by real customers.
3. 3For a mature subordinate, shift to a delegate approach — you monitor rather than instruct, reducing the frequency and duration of your intervention.
4. 4Deliver training deliberately as part of the manager's job — treat it as one of only two levers (motivation and training) you have to affect output; skipping it means neglecting half your job.
5. 5Use the one-on-one meeting specifically for mutual education and information exchange: talk about specific problems and situations so you transfer skill, while learning what the subordinate is doing and worried about.
6. 6Recognize that with e-mail and computer networks, subordinates already know what's going on and inform you minute-by-minute, so free one-on-one time from status updates and spend it on teaching and concerns.
7. 7Pair an output indicator with a quality indicator when you monitor whether a subordinate can do the job — e.g., completion date against capability — so raised competence isn't measured only by speed.

Grove's associate faced a subordinate who did poor work and reacted, 'He has to make his own mistakes. That's how he learns!' Grove judges this absolutely wrong because the subordinate's tuition — the cost of learning by mistake — is paid by his customers. The correct read is that the subordinate was immature in the task, which calls for hands-on training, not a delegate hands-off stance. Had the associate assessed task-relevant maturity first, he'd have taught the skill directly through instruction and close monitoring until the subordinate could do the job. Only once competence was demonstrated would stepping back and letting the person operate independently be warranted. The lesson: 'hands-on or hands-off' is never a fixed personality trait of the manager — it depends entirely on the subordinate's current capability on that specific task.

• If the subordinate is immature in the task, then go hands-on with training; if mature, then delegate and monitor.
• If letting the subordinate learn by mistake means the cost falls on customers, then do not use the hands-off approach — train first.
• If you are not training a subordinate, then you are neglecting half your job, because motivation and training are the only two ways to affect output.
• If subordinates already learn status through the computer network, then use one-on-ones for teaching and concerns, not for bringing them up to date.
• If you measure a subordinate's output (e.g., units completed), then pair it with a quality/capability indicator so competence isn't faked by speed.

• Treating 'hands-on vs hands-off' as your fixed style rather than a function of the subordinate — the tell: you manage every report the same way regardless of their task maturity.
• Letting an immature subordinate 'make his own mistakes' on live work — the tell: customers, not the company, absorb the learning cost, and quality complaints rise.
• Assuming smart knowledge-economy employees need no training — the tell: customers are flabbergasted or furious when a person is untrained on even a simple task like taking a reservation.
• Filling one-on-ones with status updates the subordinate already knows via the network — the tell: meetings recap what you both read on screen instead of building skill.
• Measuring only activity or output while ignoring quality — the tell: a subordinate looks productive on the count (calls made, units finished) but competence and error rates go unchecked.

• ▸Subordinate name: who you're assessing
• ▸Specific task: name the exact job, not the role
• ▸Task-relevant maturity (low/med/high): rate competence on THIS task only
• ▸Evidence of maturity: recent work sample or result that justifies the rating
• ▸Cost-of-mistake owner: who pays if they err (you / company / customer)
• ▸Chosen style (hands-on / delegate): follow from maturity + cost owner
• ▸Training action: specific skill or know-how you will teach and how
• ▸Output indicator: the countable thing you'll monitor
• ▸Paired quality indicator: the error/quality measure that guards it

1. 1When a subordinate underperforms, stop and diagnose against the two-cause test: he either can't do the job (capability) or won't do it (motivation). There is no third cause, so force the diagnosis into one of these two buckets before you act.
2. 2If the gap is capability, respond with training. Grove holds that training is the boss's job—not HR's, not a vendor's—so you personally build or transfer the missing skill.
3. 3If the gap is motivation, respond with feedback and incentives, not more instruction. Repeating the how-to on a subordinate who already knows it wastes both of you.
4. 4Deliver task-relevant feedback the way sport does: specific to the task, frequent, and tied to observable output, so the person can adjust and sustain a personal best.
5. 5Match your management style to the subordinate's task-relevant maturity on that particular task—high involvement and structure when maturity is low, delegation when it is high. The same person can be high-maturity on one task and low on another.
6. 6Run the one-on-one as your standing instrument for all of the above: it is where you detect the can't/won't split, deliver feedback, and calibrate style before problems calcify.
7. 7Judge your own success not by how much you know but by whether the individual's output rose—your knowledge is 'worth exactly nothing' unless it translates into their performance.

Ben Horowitz's illustration in the foreword makes the mechanism concrete. Picture Ms. Manager who knows the product's viral loop better than anyone in the company. On the surface that expertise looks like performance, but Grove's logic says it is 'worth exactly nothing unless you can effectively transfer that knowledge to the rest of the organization.' The individual-performance question is then applied to each of her people: is a given subordinate failing to work the viral loop because he can't (never learned it) or won't (isn't motivated)? If can't, she trains—she does the knowledge transfer herself. If won't, no amount of teaching helps and she must supply feedback and incentive instead. The whole point is that her individual brilliance only counts once it shows up as her team members' output.

• If the person is not doing the job, then classify it as can't or won't first—do not skip to a fix, because the wrong remedy (training a motivated person, motivating a skilled one) fails.
• If the cause is can't, then train personally; treat training as the boss's job one.
• If the cause is won't, then apply task-relevant feedback and compensation as feedback rather than repeating instructions.
• If task-relevant maturity is low, then increase structure and involvement; if high, then delegate—adjust per task, not per person.
• If you can't point to a rise in the subordinate's output, then your own knowledge on that topic has produced zero managerial value, regardless of how expert you are.

• Treating expertise as performance: the manager who knows the most assumes she is contributing, when un-transferred knowledge yields nothing. Tell: she can explain the work brilliantly but her team's output is flat.
• Applying one remedy to both causes—usually more explanation. Tell: you keep re-training someone who clearly already knows how, and nothing changes because the real gap was motivation.
• Using a single management style across all tasks. Tell: you micromanage a subordinate on work he's mastered, or hand off work he isn't ready for, because you ignored task-relevant maturity.
• Feedback that isn't task-relevant—vague praise or annual generalities instead of frequent, output-specific signals. Tell: the person can't tell you what to do differently tomorrow.
• Chasing today's performance gap with a new decision instead of recognizing it as a past planning failure. Tell: you scramble to close a gap that better upstream planning would have prevented.

• ▸Subordinate & task: name the specific task, not the whole job
• ▸Observed output vs. expected: state the concrete gap
• ▸Can't or won't: pick one, with the evidence that decides it
• ▸If can't — training action: what skill you will personally transfer, by when
• ▸If won't — feedback/incentive action: the task-relevant feedback or compensation lever you will apply
• ▸Task-relevant maturity (low/med/high): for this task only
• ▸Matched management style: structure & involvement level chosen to fit that maturity
• ▸Next one-on-one check: date and the output signal you'll re-measure

1. 1Redraw your boundary of ownership: list every organizational unit you supervise AND every unit you influence without formal authority (as a know-how source), because both count as your output.
2. 2Score each unit's current output not by activity but by value added — chips made, bills prepared, software designed, ad copy written — the finished, useful thing that leaves the unit.
3. 3For every hour in your day, ask which unit's output that hour increases; discard or delegate hours that only pass information along without adding value.
4. 4Rank your available tasks by leverage — the number of people whose output each task affects — and load your calendar toward the high-leverage few (training a team, a good one-on-one) over low-leverage busywork.
5. 5Elicit peak individual performance inside each unit, because a team only performs well when each member does; supply task-relevant feedback to sustain that level.
6. 6Cut windows into the black box of each unit — run regular one-on-ones as your primary intake of organizational knowledge — so you spot future high-priority trouble before it detonates.
7. 7Fix your span so the load is real: aim for six to eight subordinates (or their equivalent, counting know-how influence); if a structure leaves you only two direct reports, step in and 'act' as one manager rather than meddle or semi-retire.
8. 8Push energy toward the front of every program — early effort pays off tenfold, late effort pays off negative tenfold — so today's output gap is treated as a past planning failure, not a present scramble.

A plant manager oversees only an engineering section and a production section — just two direct reports, which is too thin a span to constitute meaningful output and pushes him toward either on-the-job retirement or meddling. Rather than leave it, he chooses to 'act' as his own engineering manager. Now the manufacturing manager still reports to him, and the five engineers who would have reported to a head of engineering report to him directly. His span becomes six: five engineers plus the manufacturing manager. The org chart keeps the engineers below the manufacturing manager's level, so the manufacturing manager takes no exception to being leveled with individual contributors. The manager's output is now the combined value-added output of all six under him — chips designed and produced — and his hours are spent raising that combined output rather than sitting idle over two reports.

• If an hour does not increase the output or the value of the output of people you're responsible for, then it doesn't belong in your day — you're a manager, so in principle every hour should.
• If you're merely passing information along rather than making things truly better in your unit, then you're not adding value and are a career casualty in the making — find the value-add or step aside.
• If you have only two direct reports, then choose to 'act' as one of the intermediate managers to reach a real span of six to eight rather than meddle or retire in place.
• If a person under you is ready to advance but you're in the way, then move — the responsibility to avoid blocking them is yours, not your boss's.
• If an output gap appears today, then treat it as a past planning failure and shift effort earlier next time, because late effort pays negative tenfold.

• Managing by activity instead of output — counting how busy the unit looks rather than the value-added thing it ships; the tell is a full team that produces nothing that leaves the door.
• Skipping one-on-ones — the manager then understands very little about what's happening in the organization; the tell is being surprised by 'emergencies' that were time bombs anyone plugged in would have heard ticking.
• Confusing your personal output with your unit's output — grinding on individual tasks while the team's combined output stalls; the tell is a productive-looking manager whose organization drifts.
• Running the job as a job shop of stops, starts, and interruptions instead of striving toward regularity; the tell is a day shredded by walk-ins and no blocks of protected, coordinated time.
• Waiting for principles to be imposed from the top ('nothing does until the Old Man decrees it') instead of acting as the micro CEO you are; the tell is a manager who blames PDQ's culture for their own group's flat output.

• ▸Unit / person I supervise OR influence — name each, including know-how relationships without formal authority
• ▸Value-added output this unit ships — the finished useful thing, not the activity
• ▸Current output level — high / adequate / below what peak individuals could produce
• ▸Highest-leverage task I can do here — the action affecting the most people's output
• ▸Feedback owed — the task-relevant feedback needed to lift an individual to personal best
• ▸Time bomb watch — the future high-priority trouble I can see through a 'window' now
• ▸Span check — direct reports or equivalents (target 6–8); note if I should 'act' as an extra manager
• ▸Value test — am I adding value or just passing information along? one honest line

1. 1Identify the limiting step in your administrative workflow — the most expensive, time-constrained, or hardest-to-reverse stage — and build the schedule around it, offsetting earlier steps in time so material arrives just before the limiting step needs it.
2. 2Choose objective indicators that characterize the output of your administrative unit, then track them on stagger charts so you see the trend from one forecast to the next and the variance against actual results.
3. 3Use the trend data to infer de facto standards (how many people produce how much output), then forecast future workload and staff the unit to match forecasted growth or decline — reassigning headcount rather than freezing it at peak.
4. 4Walk each step in the workflow and question why it exists; discard any step that survives only by tradition or formal procedure and that common sense says you can drop.
5. 5Replace 100 percent inspection with variable sampling: check at the lowest-value-added stage, sample against predetermined criteria, and raise or lower inspection frequency based on whether defects are appearing.
6. 6Keep inventory (work-in-process, applications, drafts) at the lowest-value stage possible to preserve production flexibility for a given inventory cost.

The American Embassy in London processes up to 6,000 visa applications a day with sixty people, holding 60,000 to 80,000 passports at any time, while lines of a hundred people wait outside. Its fixes — newspaper ads asking for early application, same-day drop boxes, and classifying applications into different processing tracks — only added logistical overhead and made the lines worse, because nothing addressed the basic issue of overall processing speed. About 98 percent of applicants are approved without question, so a 100 percent check is unnecessary. The production-discipline fix is a quality assurance test: sample the applications thoroughly against predetermined criteria rather than inspecting every one. This breaks the logjam without materially raising the chance an undesirable applicant slips through. It mirrors how the IRS induces compliance through checks and audits without an agent examining every single return.

• If for weeks you find no problems in a process, then reduce inspection frequency; if defects begin to develop, test more frequently until quality returns.
• If nearly all items pass a check (as 98 percent of visas do), then a 100 percent inspection is waste — switch to a criteria-based sample.
• If a step exists only by tradition or because a formal procedure ordains it, then throw it out.
• If a manager would have to examine everything a subordinate does to inspect an activity, then that's meddling — apply variable inspection (dig deep selectively) instead.
• If inventory can be held at a lower-value stage without hurting output, then keep it there to gain production flexibility per unit of inventory cost.

• Adding expediting or classification schemes that create logistical overhead without touching the limiting step — the tell: cost and effort rise but output and queue length don't improve, as at the embassy.
• Leaving administrative staffing at its highest level because you lack standards — the tell: Parkinson's law kicks in and work expands to fill the time available.
• Confusing activity with output in the 'soft professions' — the tell: managers list judgments, meetings, and mistakes-detected as their output instead of the finished result the group produces.
• Inspecting everything a subordinate does — the tell: subordinates stop feeling responsible for their own work because they know the supervisor checks it all.
• Continuing to inspect at 100 percent out of habit — the tell: 'we've always done it this way' is the only justification, even in widget manufacturing where variable inspection is available.

• ▸Unit output definition: state the finished result the unit produces (not the activities)
• ▸Chosen indicators: 2-4 measurable indicators that characterize this output
• ▸Limiting step: the most expensive/constrained stage the workflow must be built around
• ▸Time offsets: which upstream steps start when, relative to the limiting step
• ▸Lowest-value inventory point: where work-in-process is held to keep flexibility
• ▸Step audit: list each step and mark keep/discard with the reason it exists
• ▸Inspection plan: sampling criteria + rule for raising/lowering frequency
• ▸De facto standard: output-per-person inferred from trend/stagger-chart data
• ▸Forecast staffing: headcount matched to forecasted workload growth or decline

1. 1State the equation before choosing tasks: your output = the output of your own organization + the output of neighboring organizations under your influence. Every activity you weigh is measured against this, not against your personal productivity.
2. 2Score each candidate activity on three leverage dimensions from the definition: how many people it affects, how long it affects their behavior, and whether it supplies unique key knowledge to a large group. Rank the day's possible tasks by these.
3. 3Convert your unique knowledge into transferred knowledge. If you know more about the product or process than anyone, that knowledge is worth exactly nothing until you move it into the organization — build the teaching, review, or dissemination act that spreads it.
4. 4When output is lagging, diagnose before acting: a person either can't do the job or won't do it — not capable, or not motivated. Route to training if it's capability, to motivation if it's will; there are only these two levers.
5. 5Prefer changing what you do over doing the same thing faster. Working harder raises activity per hour; working smarter raises output per activity. High leverage comes from the second.
6. 6Front-load your energy. Energy spent early in a process pays off tenfold; energy spent late pays off negatively. Choose the early planning and design acts over late firefighting.
7. 7Pair every indicator you use to steer, so effect and counter-effect are both measured, and monitor the output of a work unit, not its activity, so you don't optimize a high-visibility task that produces little.
8. 8Recheck the ranking as the environment shifts. Because everything now happens faster and email disseminates information without you, information-passing is lower leverage than it once was; redirect the freed time to teaching, decision-making, and setting order.

Take Grove's own case: a manager who understands the product's viral loop better than anyone in the company. Left as private expertise, that knowledge is worth exactly nothing. The high-leverage move is not to apply the knowledge personally on one project but to transfer it — through a training session, a design review, or a documented standard — to the rest of the organization. One teaching act reaches many people (breadth), shapes how they build for months (duration), and delivers unique key knowledge to a large group (all three leverage dimensions at once). Compare that to the manager personally coding one feature: high activity, narrow output, no multiplication. The same logic explains why Grove treats training as the boss's job — it is a repeatable, high-leverage act that raises the output of the whole team, not just the trainer.

• If an activity touches many people, lasts over a long period, OR delivers unique knowledge to a large group, then treat it as high-leverage and schedule it first.
• If you hold unique key knowledge, then your job is to transfer it to the organization; if you only apply it yourself, its leverage is near zero.
• If an employee's output is low, then ask 'can't or won't' — if can't, train; if won't, motivate; do nothing else because there is nothing else.
• If you can either do the current work faster or change what work you do, then change the work (raise output-per-activity) before you push for more speed.
• If you steer by an indicator, then pair it with its counter-effect (inventory vs. shortages, completion date vs. capability) so you don't overreact and land at an extreme.

• Confusing personal brilliance with managerial value — the tell: a manager who is the smartest person in the room but whose team's output doesn't move because knowledge never leaves their head.
• Treating today's gap as a decision to make now — the tell: scrambling to close a shortfall that actually came from a planning failure months earlier; the fix belonged in the past.
• Measuring activity instead of output — the tell: praising a salesman for calls made rather than orders won, or a team for hours logged rather than delivered results.
• Steering on a single unpaired indicator — the tell: inventories driven so lean that shortages appear, or a compiler either endlessly perfected or rushed out inadequate.
• Over-inspecting subordinates' work — the tell: people stop owning their output because they know the boss checks everything, and the manager's own time is wasted on low-leverage review.

• ▸Activity: name the specific task you are considering
• ▸People affected: how many are touched by it
• ▸Duration of effect: how long behavior is shaped (days/months/years)
• ▸Unique knowledge supplied: what key knowledge does it move, and to how many
• ▸Leverage rating: high / medium / low based on the three tests above
• ▸Can't-or-won't check: if aimed at an underperformer, is the gap capability or motivation
• ▸Lever chosen: train (capability) or motivate (will)
• ▸Paired indicator: the output metric and its counter-effect you will watch
• ▸Timing: is this early-process (pays tenfold) or late-process (pays negatively)

1. 1Categorize every activity by what it accomplishes, not how it looks: as you move through the day, tag each interaction as information-gathering, information-giving, nudging, or decision-making. Grove found most activities carry multiple tags at once — a single mail annotation can gather, nudge, and decide.
2. 2Build breadth into the intake system: combine verbal exchanges, written reports and mail, scheduled reviews, and unstructured 'visits' (lunch in the cafeteria, employee orientation Q&A). Each channel captures something the others miss.
3. 3Run one-on-ones as the primary depth channel. Come prepared, let the subordinate carry the agenda, and use the time for mutual education — the subordinate teaches you the specifics you need to decide well, you teach skills and approach.
4. 4Treat verbal and casual channels as your fastest, richest source. Grove notes verbal information is faster and, from face-to-face contact, carries emotional coloring that reports strip out. Reserve reading mail and reports as self-discipline and backstop, not the front line.
5. 5Deliberately expose yourself to the 'visit' channel — settings like the orientation lecture — because the nature of questions reveals the concerns and understanding-level of employees you'd otherwise never reach.
6. 6When you conclude something, immediately map who else needs it and share deliberately down and across, so one input propagates. Grove ended a one-on-one by deciding to relay the conclusion to every group he supervised.
7. 7Use information-giving occasions to transmit values, not just facts: in a lecture or presentation, your handling of questions makes you a living role model for what the company prizes (e.g., the importance of training).

In a one-on-one, Grove's sales manager reviewed trend indicators of incoming orders. Grove was only vaguely familiar with them, but the subordinate laid out specific data showing business had stopped growing — and proved it was not merely the typical slow summer by relating the numbers to other industry indicators. After pondering the data together, they reached a congruent conclusion: business was in fact slowing, so they should be conservative on near-term investment. The subordinate left having decided to scale back growth in his own area; Grove left having decided to share the conclusion with all the business groups he supervised. One meeting, through shared information, moved every manager reporting to Grove — that is the leverage of gathering and then propagating information well.

• If a data point surprises or contradicts your standing assumption (business is 'just seasonal'), then withhold the decision until the subordinate walks you through the specifics and you've cross-checked against related indicators.
• If you reach a conclusion in a one-on-one that affects others, then decide explicitly who receives it and share it deliberately rather than assuming it will diffuse on its own.
• If a channel is verbal and face-to-face, then weight it for speed and emotional context; if it's a written report, then use it to confirm and to force the writer's self-discipline — not as your first alert.
• If your day shows only formal meetings and reports, then add unstructured 'visits' (cafeteria, orientation, tours) to reach concerns that never surface in scheduled forums.
• If you are giving information to a group, then treat your conduct as value-transmission and read their questions back as fresh information-gathering.

• Judging activity by appearance instead of output — dismissing mail-scribbling or a cafeteria lunch as trivial. Tell: you can't say what each interaction gathered, gave, nudged, or decided.
• Relying only on formal reports and missing the verbal channel's speed and emotional coloring. Tell: you learn about problems (like weak foreign-training support) late, and only when someone corners you.
• Gathering well but not sharing — hoarding a conclusion so it affects no one beyond the room. Tell: your good decisions produce no leverage because the people who must act never heard them.
• Skipping or shortening one-on-ones because they 'take time.' Tell: your decisions rest on vague familiarity rather than the detailed specifics only the subordinate has.
• Treating information-giving sessions as pure broadcast, ignoring that questions are data and your manner is values-signaling. Tell: you leave a lecture having transmitted facts but learned nothing about employees' understanding level.

• ▸Channel: name the source (one-on-one, staff meeting, mail, tour, complaint, visit)
• ▸Type: tag it — information-gathering / information-giving / nudge / decision
• ▸What it surfaced: the specific fact, concern, or trend captured
• ▸Verbal vs. written: note speed and whether emotional context was present
• ▸Cross-check: which other indicator or report confirmed or contradicted it
• ▸Conclusion reached: the congruent view or decision produced
• ▸Who else needs this: list every manager/group who must receive it
• ▸Propagation action: how and when you will share it downward/across
• ▸Value signaled: what preference or company value your handling transmitted

1. 1Classify each meeting before scheduling it: process-oriented (one-on-ones, staff meetings, operation reviews) run on a regular schedule to exchange information and educate; mission-oriented (a decision has to be made) are called ad hoc and require a chair.
2. 2Run one-on-ones as the primary channel for mutual education and information exchange: the supervisor teaches skills and suggests approaches, the subordinate reports what he is doing and what concerns him.
3. 3Set frequency and duration by how much you already know: with e-mail carrying developments to your screen minutes after subordinates post them, you can hold one-on-ones with more direct reports, less often, and for shorter durations.
4. 4Prepare before every one-on-one — preparation and holding both cost time, so build the meeting into your loaded schedule with slack so one unanticipated call does not blow up the day.
5. 5Reserve one-on-ones for what electronic channels cannot replace: the tête-à-tête to surface concerns and issues the subordinate is uncomfortable raising, not for routine status catch-up you already read online.
6. 6Treat the one-on-one as your best source of organizational knowledge — managers who skip them understand very little about what is happening in their organizations.
7. 7For mission-oriented meetings, install a chair who owns the outcome, because these exist to produce a decision, not to exchange routine information.

Grove contrasts the one-on-one with the 'proverbial Japanese employees sitting around the table.' Those employees don't need to meet their supervisor to be brought up to date — a lot of the one-on-one's information-exchange purpose is handled minute by minute around that shared table. But they still leave the table for a tête-à-tête to discuss concerns or raise issues they're uncomfortable with. Grove says working around the electronic equivalent of that table is no different: your subordinates inform you on your screen minutes after they choose to. So the update purpose shrinks, but the discomfort-surfacing purpose remains. That is precisely why he answers 'Absolutely' when asked if one-on-ones are still needed — you simply need them for fewer purposes, so you can cover more employees less frequently in shorter meetings.

• If information already reaches you via e-mail before the meeting, then cut the meeting's duration and frequency rather than cancel it — you still need it for the concerns that don't travel over the network.
• If you have ten direct reports instead of five, then hold one-on-ones less often — but do not drop them, because you don't need them as often, not because they stopped mattering.
• If a decision must be made, then it's a mission-oriented meeting and someone must chair it; if the goal is education and exchange, then it's process-oriented and runs on a schedule.
• If you find yourself meddling in subordinates' work, then you lack a raw-material inventory of discretionary projects to absorb your free time.
• If you don't hold one-on-ones, then assume you understand very little about what's actually happening in your organization.

• Cutting one-on-ones entirely because 'e-mail already tells me everything' — the tell: you learn about morale problems and uncomfortable issues only after they've become crises, because those never surfaced on the network.
• Using the one-on-one only for status catch-up — the tell: the meeting duplicates what you already read on your screen and adds nothing the subordinate couldn't have emailed.
• Skipping preparation and over-loading your schedule so one phone call wrecks the day — the tell: meetings get bumped and you run with no slack.
• Running a mission-oriented meeting with no chair — the tell: everyone talks, no decision emerges, and it reconvenes.
• Treating a written procedure as the value itself — the tell: you follow formal statements without the real thinking that led to them, so the practice degrades into ritual.

• ▸Subordinate name & role: who you're meeting
• ▸Cadence set by knowledge level: how often, given what e-mail already tells you
• ▸Duration budgeted: shorter if update purpose is already covered online
• ▸Items already known from network: skip these in the meeting
• ▸Concerns to surface (their tête-à-tête topics): issues they may be uncomfortable raising
• ▸Skills/know-how to teach this session: what you'll educate on
• ▸Open problems to review: specific situations to discuss
• ▸Follow-ups from last one-on-one: commitments to check
• ▸Slack margin in your day's load: buffer so one call won't cancel this

1. 1Push the decision down: identify the lowest level where people hold both the technical knowledge and the position authority to decide it, and route it there rather than up the chain.
2. 2Mix knowledge-power people with position-power people in the same room; the informal, egalitarian setting is a survival mechanism so both types speak freely, not affectation.
3. 3Run free discussion first — surface every relevant view and disagreement openly before anyone converges; a suppressed objection is a decision defect that surfaces later.
4. 4Force the trade-offs into a quantifiable set of relationships: name the factors being balanced (here manpower, capacity, inventory vs. delivery time and quality) so the group reasons from numbers, not opinion.
5. 5Reach a clear decision — state explicitly what was decided, so no one leaves the room reconstructing it differently.
6. 6Secure full support from everyone affected, including those who disagreed; support is not agreement, it is a commitment to back the decision.
7. 7Act early on the leading indicators, not the lagging ones — decide when the signal is faint, because energy spent early pays off tenfold and energy spent late pays off negative tenfold.

Intel's exit from DRAMs shows the process under maximum pressure. By the mid-eighties, aggressively priced high-quality Japanese DRAMs had whittled Intel's share and forced price cuts that turned the business into major losses. The company had to reason through a stark trade-off: stay in the founding business as second-best in a tough environment, or redeploy to the microprocessor business it judged it was best at. The clear decision was to lead with strength and back out of DRAMs. Full support meant carrying out what was extraordinarily difficult — moving and redeploying employees, letting some go, and shuttering factories. The result validated the process: Intel became the largest semiconductor manufacturer in the world. The lesson embedded in it: being second best is not good enough, and delaying the decision until the losses were undeniable is exactly the lagging-indicator trap that makes bad news act late.

• If a person on the team holds both the knowledge and the position authority to decide, then let them decide — don't pull it upward.
• If a participant has an objection but stays silent, then the discussion isn't free yet; stop and draw it out before converging.
• If someone disagrees but the decision is clear, then require full support (commitment to back it), not agreement — those are different bars.
• If you can't reduce the trade-offs to a quantifiable set of relationships, then you haven't done the thinking; there is a right answer, so keep working the numbers.
• If the signal is a leading indicator of bad news, then act now; if you wait for the lagging indicator, you're acting too late by design.

• Deciding at the wrong level — kicking it up because 'the Old Man hasn't decreed it,' when as a micro CEO you can and should decide within your group. Tell: people wait for imposition from the top instead of acting.
• Skipping free discussion and converging on the first plausible answer. Tell: no disagreement is ever voiced in the room, then objections leak out afterward.
• Position-power silencing knowledge-power. Tell: the technically expert people defer to the senior people and the real trade-off never gets debated — the formal-office, hierarchical setting suppresses candor.
• Acting only on lagging indicators of bad news. Tell: leaders reaffirm the business is strong, then miss their numbers because they waited for undeniable losses (Intel's own delayed DRAM exit; the CEOs who didn't see the bubble coming).
• Reaching a decision without securing full support. Tell: the clear decision exists on paper but the hard implementation — redeployment, layoffs, shutting factories — stalls because the affected people never committed to back it.

• ▸Decision to be made — state the single question in one sentence.
• ▸Lowest competent level — name the person/group with both knowledge and authority to decide.
• ▸Free discussion — list every relevant view and objection surfaced, including dissent.
• ▸Trade-offs quantified — the factors being balanced and their measurable relationships.
• ▸The clear decision — write exactly what was decided so no one reconstructs it.
• ▸Who must support it — everyone affected, including those who disagreed.
• ▸Support secured? — yes/no per name; note commitment to back it, not agreement.
• ▸Signal type — leading or lagging indicator; if leading and bad, decide now.
• ▸Implementation cost — the hard actions required (redeploy, cut, shut) and owner of each.