26High Output Management

Andrew S. Grove, High Output Management, Random House, 1983; Vintage edition with a new introduction, 1995

Linked to 37 chapters: supports 22 · contradicts 9 · mechanism 18 · same claim 6 · pairing 8

Grove wrote this book in 1983 after twenty years of running things at Fairchild and Intel, and its method is to treat every kind of work as production. A breakfast, a hire, a compiler, a criminal conviction and a middle manager's day are all flows with a limiting step, time offsets, inspection points and a rule about where to catch defects. The book's centre is a single sentence, which Grove later called the most important in it: a manager's output is the output of the organisational units under that manager's supervision or influence. Everything a manager does personally, the meetings, the reports, the decisions, is activity, and activity counts only through its leverage on other people's output. From that sentence the book derives its instruments. Indicators, paired so that no measure is gamed without its counterweight. Meetings as the medium in which managerial work is done, not an interruption to it. Decisions made by free discussion, a clear call and full support. Planning that produces today's tasks. An organisation that is always a hybrid of mission and function, held together by dual reporting and by culture. And, in the last part, the two levers a manager has on an individual, training and motivation, with the subordinate's maturity in the task at hand setting how closely to supervise. Grove's later introduction added a motto for the faster world of the 1990s: "Let chaos reign, then rein in chaos."

The breakfast factory

The book opens with a waiter who must deliver a three-minute egg, buttered toast and hot coffee at the same moment. Production's charter is to build and deliver to demand, on schedule, at acceptable quality, at lowest cost, and the way to plan it is to find the limiting step, the egg, and offset every other step backwards from delivery. Grove moves the same idea at once to recruiting a graduate, where the expensive plant visit is the limiting step and a phone screen raises offers per visit, and to writing a compiler, where each piece takes a unit test before assembly and the whole takes a system test before shipment. Every production has three operations, process, assembly and test, and a rule about rejects: catch the bad egg at delivery, the weak candidate on campus and the faulty module at unit test, because value added along the flow is what a late rejection throws away. When the waiter's toaster queue makes the egg a six-minute egg, the trade-offs appear, more staff, more equipment or standing inventory, and when volume justifies a continuous egg boiler the factory buys uniformity and low cost at the price of flexibility. His closing example is the criminal justice system, a funnel from report to jail in which a conviction costs over a million dollars and the limiting step is the number of cells, so that the largest investment is wasted at the last stage.

The factory manager's day begins with indicators, and Grove's rules for them follow from the fact that indicators steer attention as a cyclist steers where he looks. Any measure beats none. Measure output, not activity: orders, not sales calls. Measure something physical and countable. And pair every indicator with its counter-effect, inventory against shortages, vouchers processed against errors found, so that driving one down does not silently drive the other up. The organisation is a black box, and the manager cuts windows in it: a linearity chart that shows in April whether June's hiring target is reachable, trend charts against a standard, and the stagger chart, a forecast re-issued monthly and displayed against all its earlier versions, which Grove thought the best trend indicator he knew and wished on every economist. Control of future output means building to order or to forecast, and forecasting is done twice, by manufacturing and by sales, each held to its own prediction. Quality is inspected at receiving, in process and at the exit, and the choice between a gate, where everything waits for the test, and a monitor, where a sample is taken and the flow continues, is a choice between throughput and risk; for the same money one can monitor far more than one can gate, so Grove leans to monitoring wherever experience says big problems are unlikely, and to varying inspection with the quality observed. His example is the American embassy in London, checking a million visa applications a year of which 98 per cent are approved, where sampling would have cleared the queue.

Managerial leverage

Asked what a manager's output is, middle managers answer with activities: judgments, direction, training, planning. Grove gives outputs instead, processed wafers, working designs, educated students, a healed patient, and states the equation. Output equals the sum of activities each multiplied by its leverage, and productivity rises three ways: doing an activity faster, raising its leverage, or shifting time from low-leverage to high-leverage activities. He prints a day of his own calendar to show that a manager's work has no pattern, ends when he is tired rather than finished, and takes place two-thirds in meetings, which are therefore the medium and not the enemy. Information gathering is the base of everything else, and its best source is verbal and sketchy, like a headline; written reports exist to make the writer precise, which is why writing them matters more than reading them. Leverage is high when one person affects many, when a brief act affects someone for a long time, or when a person supplies knowledge that others depend on; it is negative when a manager arrives unprepared, waffles so that no green light becomes a red light, or spreads depression through an organisation. Delegation without monitoring is abdication, and the monitor should be applied, like inspection, at the lowest-value stage, the rough draft rather than the polished report. A supervising manager should have six to eight reports, and interruptions are handled by batching them into scheduled one-on-ones.

The meetings chapter separates process meetings, which are scheduled, from mission meetings, called to decide. The one-on-one belongs to the subordinate, who sets its agenda and prepares its outline; it lasts at least an hour, happens as often as the subordinate's maturity in the task requires, and its rule is to ask one more question until both are at the bottom of the problem. Ninety minutes every two weeks, Grove calculates, can lift the quality of two weeks of the subordinate's work. The staff meeting is where peers learn to decide together, and the operation review is where people several levels apart teach each other. A mission meeting costs money, ten managers for two hours at a hundred dollars an hour is two thousand dollars, and its chairman does most of the work before it begins; more than eight people and it cannot decide. Decisions themselves follow a model of free discussion, a clear decision and full support, where support means commitment rather than agreement. Grove's diagnosis of why groups fail to decide is the peer-group syndrome, peers circling for fear of going against the group until a senior person, the peer plus one, gives them the confidence to conclude; his remedy for surprise vetoes is six questions settled in advance, what is being decided, when, by whom, with whom consulted, who ratifies, and who is told. Planning is the same three steps as the factory: what the environment will demand, what the work already under way will deliver, and what to do to close the gap. Its output is not the bound plan but the tasks it causes, and management by objectives is the same process applied to the quarter, with few objectives and dated key results, set high enough that even hard effort has only even odds.

A team of teams

When the breakfast factory franchises nationally, every question of centralisation gets a different answer: eggs are bought regionally, machinery centrally, wages locally, tableware centrally with regional warehouses. Grove generalises. Every organisation can be drawn as wholly mission-oriented, each unit doing everything for itself, or wholly functional, each function serving every unit; the first has one advantage, responsiveness, and the second has all the others, economies of scale, re-allocable resources and the leverage of specialists. Every large organisation with a common business purpose ends up as a hybrid, a claim he calls Grove's law, and the task of the hybrid is allocating shared resources and settling the conflicts that allocation creates. That cannot be done at the top, which he illustrates from the Hungarian central planners of his youth who could not get high-contrast film to shops in winter, so it falls to middle managers through dual reporting. The founding case at Intel was plant security: the corporate security manager knew the standards, the plant manager knew whether the guards turned up, and the guards reported to both. As careers progress, a general manager comes to supervise functions he does not know, so technical supervision must come from a functional chief or, as often, from a council of peers, which requires each member to surrender some individual decision-making. That surrender requires trust, and trust is not an organisational principle but a matter of culture.

Culture, in turn, is one of three modes of control, and Grove introduces them through a drive to work. Buying tyres, you act on self-interest and the market controls the transaction. Stopping at a red light, you honour a contract that a policeman monitors. Arriving at an accident, you forget both and help strangers at some risk. Most work cannot be priced by the piece, so the employment contract buys general authority to monitor and correct; where the environment changes faster than rules can be written, only shared values will do. Which mode fits depends on two variables, whether the person's motivation runs to self-interest or group interest, and how complex, uncertain and ambiguous the environment is, a composite Grove calls the CUA factor. Self-interest in a simple environment is the market's domain; group interest in a simple one is the contract's; group interest in a complex one is culture's; and self-interest in a complex environment admits no control at all, which is why a senior outside hire brought in to fix a troubled unit so often fails. New employees should start in low-CUA jobs and be promoted into ambiguity as shared experience teaches them the values.

The players

A person who is not doing the job either cannot or will not, and Grove's test is whether they could do it if their life depended on it. That yields the manager's two levers, training for capability and motivation for the will, and the last part of the book takes them in turn. Motivation cannot be done to someone; it comes from within, and a manager can only build an environment in which motivated people flourish, judged by performance and not by anyone's report of feeling motivated. Grove takes Maslow's hierarchy because his observation of working life confirms it: physiological, safety and social needs get people to show up, esteem and self-actualisation make them perform, and only the last is limitless, whether it is driven by competence, the violinist who keeps practising, or by achievement, the ring-tosser who stands just far enough from the peg to make it hard. Money is a necessity low in the hierarchy and a measure high in it, and the test is whether a person cares about the absolute size of a raise or its size relative to others'. Fear, at the upper levels, becomes fear of failure, which makes achievers conservative, so objectives are set where even hard effort has a fifty-fifty chance. The sports analogy follows. People who are indifferent at the office run marathons to beat a clock, so give work the character of a competitive sport, with rules and a racetrack. Intel's facilities upkeep improved almost at once when a senior manager in each building began scoring it and the scores were compared; nothing else changed. The manager's role in this is the coach, who takes no credit, is tough on the team, and once played the game.

There is no best management style. Intel rotated managers between similar groups and found that neither the managers nor the groups kept their high or low output; results came from particular pairings. The variable that decides is the subordinate's task-relevant maturity, a combination of achievement orientation, readiness to take responsibility, education, training and experience, all specific to this task in this environment. A star sales manager moved into a plant became overwhelmed, because his general maturity had not changed and his task-relevant maturity was near zero, and Grove notes that this should have been predictable. When maturity is low the effective style is structured, telling the subordinate what, when and how; at medium it is individual-oriented, two-way communication and support; at high it is minimal, agreeing objectives and monitoring. None of these is nicer than another; they are effective or not. Structure moves over time from being imposed to being internal, as it does between a parent and a growing child, and the progression depends on shared operational values, which the supervisor is responsible for transmitting. A subordinate's mistakes are paid for by the organisation's customers, so letting him make his own is not teaching. Maturity can also fall: the sergeant of a quiet camp who reverts to barking orders when a jeep of the enemy comes over the hill is doing the right thing, and the mode managers most dislike is the one they most often reach for too late.

Managers misjudge their own style. Asked to rate their supervisors, subordinates described far more structure than the supervisors themselves reported; about nine in ten supervisors saw themselves as more communicating or delegating than their people did. Friendship helps the communicating style and makes the structured one harder to return to, and Grove's test is to imagine delivering a tough review to the friend.

The review is the most important form of feedback a manager gives and, he says, among the things managers do worst. Its purpose is to improve performance, first by finding missing skills and second by raising motivation at the same skill level. Assessing means deciding in advance what was expected, weighing output measures against internal measures whose proper ratio can shift month by month, and allowing for the time offset between a manager's work and the results it produces. Grove's own error was a manager whose superb year was the light of distant stars, work done years before, and whose collapse the following year came as his real performance improved. The trap he names is potential: a manager's rating cannot be higher than his organisation's, whatever his form, or the company learns that acting like a manager is what counts. Delivering means levelling, listening with every sense until the message has arrived, and leaving oneself out. A review should carry the few messages a person can absorb, not everything the reviewer has observed. In a review of a serious problem the subordinate moves from ignoring to denying to blaming others to accepting responsibility, and the step from blame to responsibility is emotional while the step to a solution is merely intellectual; any outcome that includes a commitment to act is acceptable, and agreement is not required. Managers write their best reviews for their stars and their most useful ones for their weak performers, which Grove thinks is backwards, since improving a star is the higher-leverage act.

The remaining chapters cover hiring, where the interviewer must for once judge potential and can only improve the odds; the resignation, which must be dealt with the moment it is announced; and compensation as feedback, with a performance bonus rising as a share of pay and the honest observation that merit pay requires someone to come last. Grove's answer to the Peter principle is that achievers alternate between meeting and exceeding requirements as they are promoted, and that a person promoted too far should be openly returned to the job he did well rather than pushed out. The book ends on training, which Grove insists is the manager's own job and one of the highest-leverage things a manager can do. The arithmetic is four lectures, three hours of preparation for each hour taught, twelve hours in all, for ten people who will work twenty thousand hours in the coming year; a one per cent improvement returns two hundred hours. Training must be tied to how things are actually done, must be a process and not an event, and must be given by someone the students see as a practising authority, which a hired trainer cannot be.

Since 1983

The book was reissued in 1995 with an introduction in which Grove described what had changed: the Japanese memory makers who forced Intel out of the business it was founded on, the arrival of e-mail, and a workplace that had become faster and less kind, in which every employee competes with anyone anywhere who can do the same job and owns their career as a sole proprietor. The three basic ideas, he wrote, had not changed. It was reissued again in 2015, when a generation of founders who had been handed it by their investors made it the most recommended management book in the technology industry, and its management by objectives, taught by Grove to John Doerr at Intel, became the objectives and key results system used across that industry. Its debts and its dated parts are both visible now. Task-relevant maturity is Hersey and Blanchard's situational leadership with the name changed; Maslow's hierarchy, adopted because it matched Grove's observation, has not held up as psychology, though the distinction between needs that get people to work and needs that make them perform survives without it. Forced comparative ranking has been abandoned by most of the companies that adopted it. The book assumes a shared building, a long career in one firm and a manager who is always he. What has lasted is the frame: output over activity, leverage as the measure of a manager's time, the pairing of indicators, monitoring as the line between delegating and abdicating, the meeting as a medium, and the rule that the style of supervision is set by the person's maturity in this task and not by the manager's preference.