Before we begin
This is a full-length summary of Richard Rumelt's book Good Strategy, Bad Strategy, published in twenty-eleven. Rumelt spent his career at the UCLA Anderson School of Management, and before that he was an engineer at NASA's Jet Propulsion Laboratory, working on spacecraft design. That engineering background matters, because it shapes the whole book. Rumelt does not think of strategy as vision, or ambition, or leadership. He thinks of it as design work: figuring out what is actually going on, deciding on an approach, and then building a set of actions that fit together and reinforce each other.
The book has three parts. Part one draws the line between good strategy and bad strategy, and gives you a structure for building the good kind. Part two is a tour of the sources of power a strategy can draw on. Part three is about how to think like a strategist, and how to keep your head when everyone around you is losing theirs.
Listen for one idea above all others: strategy is not what you want to achieve. Strategy is how you propose to overcome an obstacle. If there is no obstacle named, there is no strategy in the room.
Part One. Good and Bad Strategy
The opening image
Rumelt starts in eighteen-oh-five, off the coast of Spain. Napoleon has conquered most of Europe and wants to invade England, but to cross the Channel he needs control of the sea. At Trafalgar, the combined French and Spanish fleet of thirty-three ships meets a British fleet of twenty-seven. The accepted tactic of the day was for both fleets to hold a line and fire broadsides at each other. Admiral Nelson did something else. He broke his fleet into two columns and drove them straight at the enemy line, perpendicular to it. His lead ships took an enormous risk, but Nelson had judged that the less-trained French and Spanish gunners would not be able to compensate for the heavy swell that day. The French and Spanish lost twenty-two ships. The British lost none. Nelson was killed, and British naval dominance lasted a century and a half.
Rumelt draws out the pattern. Nelson's problem was that he was outnumbered. His strategy was to accept risk to his lead ships in order to break the coherence of the enemy fleet, because he believed that in the resulting melee his more experienced captains would win. That is what good strategy looks like. It is simple. It does not require a thick deck of slides. It does not fall out of a matrix or a framework. A leader identifies the one or two critical issues in a situation, the pivot points where effort gets multiplied, and then concentrates action and resources on them.
What passes for strategy, and why it is not
Rumelt then gives four examples of the opposite, and you will recognise all four.
A strategy retreat: two hundred senior managers in a hotel ballroom, a film of the company's products being used around the world, an address by the chief executive with dramatic music announcing goals of global leadership, growth and high shareholder return, breakout groups for buy-in, and a release of balloons. Everything except a strategy.
Lehman Brothers: by two thousand and six the strain in American housing was visible. Sales had peaked, price appreciation had stopped, a small rise in interest rates had already triggered foreclosures. The chief executive's response was a strategy of gaining market share by growing faster than the industry, which in Wall Street language meant increasing its risk appetite, taking the deals competitors were rejecting. The firm ran on about three per cent equity with much of its debt very short-term. A good strategy recognises the nature of the challenge and offers a way of surmounting it. Simply being ambitious is not a strategy. Lehman ended a hundred and fifty-eight years of business two years later.
Iraq after two thousand and three: the army-to-army fighting ended quickly, an insurgency gathered, and American units fell back on search-and-destroy missions out of secure bases, the approach that had failed in Vietnam. There were high-sounding goals — freedom, democracy, reconstruction, security — and no coherent approach to the insurgency. That changed in two thousand and seven when General Petraeus arrived with an actual idea: an insurgency can be defeated as long as the large majority of civilians support a legitimate government, so shift the military's focus from patrolling to protecting the population, because a population not living in fear of retaliation will supply the information needed to isolate the insurgents. Replacing amorphous goals with a problem-solving approach changed the results dramatically.
And a conference lunch, where the chief executive of a web services company put down his fork and said, "Strategy is never quitting until you win." Rumelt replied, "Winning is better than losing," and changed the subject.
That last one captures the disease. A word that can mean anything has lost its bite. Business speech has turned marketing into marketing strategy, data processing into I.T. strategy, and making acquisitions into a growth strategy. Cut some prices and someone will call it a low-price strategy.
Rumelt's definition is narrow on purpose. A strategy is a cohesive response to an important challenge. It is a coherent set of analyses, concepts, policies, arguments and actions that respond to a high-stakes challenge. Note what that excludes. It is not a decision made by a senior person; the word strategic should not simply mark the pay grade of the decision maker. It is not a goal. And, importantly, it is not divorced from action.
That last point deserves a moment. Many people assume a strategy is a big-picture direction, separate from any specific action, and that everything after is implementation. Rumelt argues this is the source of the most common complaint in business, which sounds like this: "We have a sophisticated strategy process, but there is a huge problem of execution. We almost always fall short of our goals." His diagnosis is that these executives have confused strategy with goal setting. When the strategy process is basically a game of setting performance targets, so much market share, so much profit, then there really is a yawning gap between the ambition and the action. Strategy is about how an organisation will move forward. A strategy that fails to define plausible, feasible immediate actions is missing a critical component. Those actions are not implementation details. They are the punch in the strategy.
Good strategy is unexpected
Rumelt's first observation about good strategy is almost comic: one of its great advantages is that other organisations do not have one, and do not expect you to have one either.
His example is Apple in nineteen ninety-seven. After Windows ninety-five, Apple went into a death spiral. Wired ran an article called "One hundred and one ways to save Apple," including "sell yourself to IBM or Motorola." By September, Apple was two months from bankruptcy, and Steve Jobs came back as interim chief executive.
What Jobs did was both obvious and unexpected. He shrank Apple to a scale appropriate to being a niche producer in a brutally competitive market. He talked Microsoft into investing a hundred and fifty million dollars, exploiting Bill Gates's anxieties about what a dead Apple would mean for Microsoft's antitrust position. He cut fifteen desktop models back to one, all portables back to one laptop, and printers and peripherals entirely. He cut development engineers, software development, distributors, and five of the company's six national retailers. He moved manufacturing to Taiwan and cut inventory by more than eighty per cent. A new web store sold direct.
Almost all of that is elementary business practice. Of course you simplify to your core in a financial nosedive. And yet nobody expected it. Jobs explained it to Rumelt in a few sentences: the product line was too complicated and the company was bleeding cash, a friend of the family had asked which Apple computer to buy and he could not give her clear guidance, and he was appalled there was no Apple consumer computer under two thousand dollars.
Then Rumelt asked the harder question. Apple had under four per cent of the market, and network effects made the Windows and Intel standard look unassailable. What was the long-term strategy? Jobs smiled and said, "I am going to wait for the next big thing."
That is not a general formula, and Rumelt says so. But it was a wise reading of the situation. Jobs was not setting a market share target or pretending levers could restore leadership. He was focused on the real sources of and barriers to success in his industry: recognising the next window of opportunity and having the quickness to pounce. Two years later came the iPod, then online music, then the iPhone.
Rumelt pairs this with Desert Storm. Commentators predicted trench warfare and twenty thousand American casualties. What they had not anticipated was that Schwarzkopf had a strategy: a massive secret left hook, moving two hundred and fifty thousand soldiers west of Kuwait and north into empty Iraqi desert, then turning east into the flank of the Republican Guard, while the marines made a slow feint north and eight thousand more waited offshore as a diversion. The ground war lasted a hundred hours.
Here is the strange part. Envelopment was not a secret. It was the army's published doctrine, in a field manual available to anyone for twenty-five dollars, with a diagram. So why the surprise? Rumelt's answer is that the real surprise was that such a pure and focused strategy was actually implemented. Most complex organisations spread resources rather than concentrate them, in order to placate interests. Schwarzkopf had to suppress the ambitions of the air force, the marines, individual army units, every coalition partner and the political leadership in Washington. The army's best light infantry was assigned to support French armour, and protested. Air force commanders who believed the war could be won from the air had to be forced to support the land offensive.
The lesson: having conflicting goals, dedicating resources to unconnected targets and accommodating incompatible interests are the luxuries of the rich and powerful, and they make for bad strategy. Good strategy requires leaders willing to say no. Strategy is at least as much about what an organisation does not do as what it does.
Discovering power
The second natural advantage of good strategy comes from insight: seeing a source of strength or weakness that others have not seen.
Rumelt uses David and Goliath as the archetype. If you list the two combatants' strengths and weaknesses in the conventional way, David looks doomed. It is only after the stone is slung that your viewpoint shifts. You realise the shepherd's sling is a real weapon and youthful quickness is a real asset. You realise David discarded the armour because it would only have slowed him, and that bronze would not have saved him from a blow at close range anyway. And you discover a critical weakness: Goliath's armour did not cover his forehead. David's weapon delivered force with precision at a distance, which neutralised size and strength. The story teaches that our preconceived ideas of strength and weakness may be unsound.
The business version is Walmart. Rumelt teaches a case set in nineteen eighty-six, and begins by writing the conventional wisdom of the era on the whiteboard: a full-line discount store needs a population base of at least a hundred thousand. The class lists the familiar reasons for Walmart's success. Sam Walton's leadership. Big stores in small towns. Everyday low prices. Computerised warehousing and trucking. Non-union. Low administrative expenses.
Then Rumelt asks the question that stops the room. If those are the reasons, and the case was published in nineteen eighty-six, why was Walmart able to run rampant over Kmart for the next decade? Was the formula not obvious? Where was the competition?
His point is that whenever an organisation succeeds greatly, there is at the same time blocked or failed competition. Kmart, once the leader in low-cost variety retailing, spent two decades expanding internationally and ignoring Walmart's logistics innovations. It filed for bankruptcy in two thousand and two.
Then a sharper question. Both companies installed bar-code scanners in the early nineteen eighties. Why did Walmart benefit more? Most retailers saw scanners as a way of avoiding the cost of changing price stickers. Walmart went further, building satellite-based information systems and using that data to run inbound logistics and to trade with suppliers in return for discounts. The pieces are complements: the bar codes, the integrated logistics, the frequent just-in-time deliveries, the large stores with low inventory. A competitor could copy one element and gain almost nothing. It would have to adopt the whole design.
But the deepest insight comes last. Rumelt turns to a participant and says: you claimed Walton broke the conventional wisdom, but that wisdom was based on the straightforward logic of fixed and variable cost. It takes a lot of customers to spread overhead. Exactly how did Walton break the iron logic of cost? Then he puts the man in a role: imagine you are a Walmart store manager in nineteen eighty-five, unhappy with head office, and your father offers to buy the store so you can run it yourselves. The answer comes back: no, we could not make a go of it alone. The store needs to be part of the network.
That is it. It is not the store. It is the network of a hundred and fifty stores, the data flows, the management flows, the distribution hub. The network replaced the store, and a regional network of a hundred and fifty stores serves a population of millions. Walton did not break the conventional wisdom. He broke the old definition of a store.
And in doing so he broke a deeper doctrine of his era: decentralisation. Kmart had long given each store manager authority to choose product lines, pick vendors and set prices. The forgotten cost of decentralisation is lost coordination. Stores that do not choose the same vendors cannot benefit from an integrated network, and stores that do not share information cannot learn from each other. A large organisation may balk at adopting a new technique, but that is manageable. Breaking with one's basic philosophy is rare absent a near-death experience.
The same move appears in national security. Andy Marshall, the long-serving director of net assessment at the Pentagon, explained that Cold War defence planning had become driven by the annual budget cycle: the Joint Chiefs estimated the Soviet threat, the Pentagon produced a shopping list of counters, Congress funded some fraction, and the cycle began again. That process conditioned American action on Soviet strengths expressed as threats, not on Soviet weaknesses and constraints. There was a war strategy, a catastrophic spasm, but no plan for competing over the long term. The paper Marshall and James Roche wrote in nineteen seventy-six reframed it: a true competitive strategy means taking actions that impose exorbitant costs on the other side. Invest in technologies that are expensive to counter and where the counters do not add to Soviet offensive capability — greater missile accuracy, quieter submarines, systems that make Soviet systems obsolete, and the selective advertising of dramatic new technologies. No charts, no formulas. Just an idea: use your relative advantages to impose out-of-proportion costs on your opponent and complicate his problem of competing with you.
Bad strategy has four hallmarks
Rumelt coined the term bad strategy at a Washington seminar in two thousand and seven on the decline of national security strategy. He argues bad strategy is not the absence of strategy, and it is not miscalculation. It has a life and logic of its own. Learn its four hallmarks and you will be able to spot it in minutes.
The first is fluff. Fluff is a superficial restatement of the obvious combined with a generous sprinkling of buzzwords, using inflated words and apparently esoteric concepts to create the illusion of high-level thinking. Rumelt's example is a retail bank's internal memo declaring that "our fundamental strategy is one of customer-centric intermediation." Intermediation means the bank takes deposits and lends them to others. In other words, it is a bank. Customer-centric could mean better terms or better service, but its policies showed no such distinction. Strip away the fluff and the statement reads: our bank's fundamental strategy is being a bank. A hallmark of true expertise is making a complex subject understandable. A hallmark of bad strategy is unnecessary complexity masking an absence of substance.
The second hallmark is failure to face the challenge. A strategy is a way through a difficulty. If the difficulty is not defined, you cannot assess the strategy, so you cannot reject a bad one or improve a good one. His case is International Harvester, once the fourth largest corporation in America. Its nineteen seventy-nine corporate strategic plan was thick with detail: five divisional plans, market share targets, a beautiful hockey-stick profit forecast. What it did not mention was the elephant in the elevator. Harvester had grossly inefficient work organisation. Work rules allowed employees with seniority to transfer jobs at will, each transfer setting off a cascade of other transfers. Its profit margin had been about half its competitors' for a long time, and it had the worst labour relations in American industry. The plan did not name any of this. The company lost more than three billion dollars between nineteen seventy-nine and nineteen eighty-five, closed thirty-five of forty-two plants, and shed eighty-five thousand workers.
Rumelt's line here is the one worth memorising. If you fail to identify and analyse the obstacles, you don't have a strategy. Instead, you have either a stretch goal, a budget, or a list of things you wish would happen.
The counterexample is DARPA, the American defence research agency, whose own statement of its challenge is blunt: matching military problems with technological opportunities is extremely difficult because some military problems have no obvious technical solution and some emerging technologies have unclear military consequences. So DARPA focuses on what it calls the DARPA-hard niche. And its guiding policy comes with specific operating rules: it retains program managers for only four to six years, to limit empire building and to ensure a successor willing to challenge a predecessor's work; and it keeps a very limited investment in overhead and physical facilities, so entrenched interests cannot thwart new directions. Compare that with vague aspirations like "retain the best talent" and "maintain a culture of innovation."
The third hallmark is mistaking goals for strategy. Rumelt tells the story of Chad Logan, chief executive of a graphic arts company, whose plan was called twenty-twenty: revenue growth of twenty per cent a year and margins of twenty per cent or better, against a history of about twelve per cent margins and flat share. The plan's first page listed "our key strategies": we will be the graphic arts firm of choice; we will delight our customers; we will grow revenue by at least twenty per cent each year; we will maintain a profit margin of at least twenty per cent; we will have a culture of commitment; we will foster an honest and open work environment; we will support the broader community. Three weeks of consultation had produced it, and there was good buy-in.
Rumelt asked what had to happen for twenty-twenty to be realised. He was looking for a point of leverage: a strength to build on, or a change in the industry opening up an opportunity. Logan pulled out a Jack Welch quotation about reaching for the impossible.
Rumelt's private association was with the First World War, where the philosophy of the age held that willpower, spirit, morale and aggressiveness were the keys to success, and generals flung motivated men at fortified machine-gun positions for three years. At Passchendaele, General Haig shelled German fortifications despite being warned it would break the dikes and flood the fields. It did, turning the ground to a quagmire that drowned tanks, horses and the wounded, and the doctrine of one last push continued for three months. Five miles of ground were gained, more than seventy thousand Allied soldiers died and a quarter of a million were wounded. Churchill called it a forlorn expenditure of valour and life without equal in futility.
The point is not that motivation is bad. It is that the troops at Passchendaele did not suffer from a lack of motivation. They suffered from a lack of competent strategic leadership. A leader may justly ask for one last push, but the leader's job is also to create the conditions that make that push effective, and to have a strategy worthy of the effort called upon.
Rumelt told Logan he had a lot of ambition but no strategy, and advised spending a month with a small team reviewing who his buyers were, who he competed with, what was changing, and where he might get a jump — ending with a very short list of the most important things to do. Logan thanked him and hired someone else, who ran a workshop called Visioning, in which the group stretched their aspirations from "bigger" to "very much bigger" and were then challenged to think twice as big as that.
The fourth hallmark is bad strategic objectives. Rumelt distinguishes goals, meaning overall values and desires, from objectives, meaning specific operational targets. A nation may have goals of freedom, justice, peace and security. It is strategy that transforms those into a coherent set of actionable objectives, and a leader's most important job is creating and constantly adjusting that bridge.
Bad objectives come in two flavours. The first he calls a dog's dinner: a scrambled mess of things to accomplish. The mayor of a small city in the Pacific Northwest had a plan containing forty-seven strategies and a hundred and seventy-eight action items, of which item one hundred and twenty-two was "create a strategic plan." The Los Angeles school district's plan for its weakest schools contained seven strategies, twenty-six tactics and two hundred and thirty-four action steps. Such lists grow out of planning meetings where every stakeholder makes a suggestion and the whole day's collection gets swept into the plan. Then, in recognition that it is a dog's dinner, the label "long-term" is added so that none of them need be done today.
The second flavour is blue sky: an objective that simply restates the desired state of affairs, skipping the fact that nobody has a clue how to get there. One of that school district's key strategies was to "build school and district leadership teams that share common beliefs, values, and high expectations." There is no diagnosis of why leadership is weak. And asking for transformational leadership is absurd when the plan itself explains that many administrators struggle with daily problems, when nobody knows how to create transformational leaders even in good conditions, and when those leaders remain embedded in a bureaucracy where they cannot change the colour of their paper without permission.
Rumelt adds a subtle warning here. When a leader characterises the challenge as underperformance, it sets the stage for bad strategy. Underperformance is a result. The true challenges are the reasons for the underperformance. Unless leadership offers a theory of why things have not worked in the past, or why the challenge is difficult, it is hard to generate good strategy.
Why is there so much of it?
Rumelt gives three pathways.
The first is the unwillingness or inability to choose. He sat in on a nineteen ninety-two strategy discussion at Digital Equipment Corporation where three senior executives argued for three directions. One said DEC was and always had been a computer company integrating hardware and software into usable systems, which the others named the Boxes strategy. The second said boxes had become a commodity and the only real resource was customer relationships: the Solutions strategy. The third said the heart of the industry was semiconductor technology: the Chips strategy.
Why not do all three? Two reasons. First, if you resolve conflict by adopting every option on the table, nobody has any incentive to sharpen their arguments in the first place. Only the prospect of a real choice calls forth people's best case for their own proposal and their best critique of the others. Second, both Chips and Solutions were dramatic transformations requiring wholly new skills, and it is not feasible to do two deep transformations of a company's core at once.
Their preferences created what is known as Condorcet's paradox: in paired votes, Boxes beat Chips, Solutions beat Boxes, and Chips beat Solutions, cycling forever with no stable majority. The chief executive had made the mistake of asking for consensus. What emerged was: "DEC is committed to providing high-quality products and services and being a leader in data processing." A political outcome reached by people who could not agree what to give up, so they set nothing aside, hurt no interest group, and crippled the whole. DEC was acquired by Compaq six years later.
The contrast is Intel. By nineteen eighty-four it was clear Intel could not match Japanese prices in memory chips. Andy Grove recalls: "We persevered because we could afford to." The turning point came when he asked chairman Gordon Moore what a new chief executive brought in from outside would do. Moore replied immediately: he would get us out of memories. Grove went numb, then said, "Why shouldn't you and I walk out the door, come back and do it ourselves?" Even after forming the conviction it took over a year, because memory had been the engine driving research, production, careers and pride at Intel.
Rumelt's conclusion is uncomfortable and true. Any coherent strategy pushes resources toward some ends and away from others. Unless collective ruin is imminent, a change of strategy makes some people worse off, so there will be powerful forces opposed to almost any change. When people evade the work of choosing, you get vague goals everyone can agree on — direct evidence of leadership's insufficient will or political power to make hard choices. Put differently: universal buy-in usually signals the absence of choice.
The second pathway is template-style strategy: the fill-in-the-blanks sequence of vision, mission, values, strategies. Rumelt traces its lineage from Max Weber's writing on charisma through the nineteen eighties books arguing that transformational leadership could be taught to anyone. He quotes Peter Drucker: "Effective leadership doesn't depend on charisma. Dwight Eisenhower, George Marshall, and Harry Truman were singularly effective leaders, yet none possessed any more charisma than a dead mackerel." The template frees consultants and leaders from the work of analysing real challenges, and because it is couched in positives, no feelings are hurt. His illustration of charisma without strategy is the Children's Crusade of twelve-twelve, in which two charismatic young men gathered thousands of followers on the strength of a vision. Those who reached Marseilles boarded seven ships. Two wrecked; the other five were captured and everyone aboard sold into slavery. Charisma is a potent way to overcome inertia, but thousands sacrificed to no end. To achieve great ends, charisma must be joined to careful attention to obstacles and action, as Gandhi did with orchestrated demonstrations, marches, publicity and prison terms that eroded the British rulers' self-image of fairness.
The third pathway Rumelt calls New Thought: the belief, running from Emerson through the modern positive-thinking industry, that all you need is the right mental attitude. If believing makes it so, then analysing obstacles is not merely unnecessary, it is harmful.
The kernel
Here is the structural heart of the book, and if you remember one thing, remember this.
A good strategy has three elements, which Rumelt calls the kernel.
The first is a diagnosis: a definition or explanation of the nature of the challenge. A good diagnosis simplifies the overwhelming complexity of reality by identifying certain aspects of the situation as critical.
The second is a guiding policy: an overall approach chosen to cope with or overcome the obstacles identified in the diagnosis.
The third is a set of coherent actions: steps that are coordinated with one another to carry out the guiding policy.
The medical analogy makes it concrete. A patient presents with signs, symptoms and a history. The doctor makes a clinical diagnosis, naming a disease. The therapeutic approach chosen is the guiding policy. The specific prescriptions for diet, therapy and medication are the coherent actions.
On diagnosis, Rumelt's colleague John Mamer sat in on ten of his classes and observed that there is really only one question being asked in each case: "What's going on here?" That is right. A great deal of strategy work is not deciding what to do; it is comprehending the situation.
Diagnosis matters because it is a judgement, not a fact. Consider Starbucks in two thousand and eight, with flat store traffic and return on assets fallen from fourteen per cent to about five and a half. One executive might diagnose that as a problem in managing expectations. Another as a search for new growth platforms. A third as an eroding competitive advantage. None of these is an action, but each suggests a range of things that might be done and rules others out. And none can be proven correct. Diagnosis is a judgement about the meanings of facts.
There is a further criterion. Whereas a social scientist seeks the diagnosis that best predicts outcomes, good strategy is based on the diagnosis that promises the most leverage over outcomes. Student performance is better explained by social class than by spending per pupil, but that knowledge yields few useful policies. Diagnosing school performance as a problem of organisation explains less of the variation, but unlike social class, organisation is something policy can act on.
The classic case of a diagnosis reshaping history is George Kennan's long telegram of nineteen forty-six, framing the Soviet Union as a regime whose antagonism to the West was structural rather than negotiable, but whose leaders were realists about power. The guiding policy that followed was containment. Note that containment is often called a strategy, but it lacked action; every president from Truman onward struggled to turn it into actionable objectives, which over time produced NATO, the Berlin Airlift, Korea, missiles in Europe and Vietnam.
The business version is Lou Gerstner at IBM in nineteen ninety-three. The prevailing view, inside the company and on Wall Street, was that IBM was too integrated for a fragmenting industry, and preparations were under way to break it into separately listed pieces. Gerstner changed the diagnosis: in an increasingly fragmented industry, IBM was the one company with expertise in all areas, and its problem was not that it was integrated but that it was failing to use the integrated skills it had. The guiding policy followed: offer customers tailored solutions to their information processing problems, leveraging the brand and the breadth, willing to use outside hardware and software as required. Its primary value-adding activity shifted from systems engineering to consulting, from hardware to software.
On guiding policy, Rumelt is precise. It is not a goal or a vision or an image of a desirable end state. It defines a method of grappling with the situation and rules out a vast array of possible actions. Like guardrails on a highway, it directs and constrains action without fully defining its content. Wells Fargo's vision was to satisfy all of its customers' financial needs and be the premier provider in every market — ambition with no information about how. Its actual guiding policy was to use the network effects of cross-selling: the more products the bank sells a customer, the more it knows about that customer and about its whole network of customers, and the more products it can then create and sell.
The homely example is Stephanie, who owns a corner grocery and faces hundreds of possible adjustments. Keep prices down or add organic produce? Stock Asian staples for the students nearby? Open longer? Add a checkout? Paint the ceiling green or white? An economist would say maximise profit, which is correct and useless. She diagnosed her challenge as competition with a cheaper supermarket open twenty-four hours, and observed that most of her customers walk past every day. That let her frame the real choice as one between two customer groups: price-conscious students, or time-poor professionals. Reducing thousands of individual choices to a choice among a few customer groups is a dramatic reduction in complexity. She chose the busy professional with little time to cook, and the actions organised themselves: a second checkout for the five o'clock rush, more parking in the alley, the student snack space converted to prepared take-home food, late hours dropped, staffing concentrated after work and at lunchtime.
On coherent action, Rumelt is emphatic that most people stop at the guiding policy and call it the strategy. His illustration is a European business group president, frustrated that a Pan-European product initiative had gone nowhere despite a quarterly executive committee, developer rotations, a new products group, and promotion criteria that included contribution to the initiative. Rumelt asked: suppose you absolutely had to have Pan-European products in market within eighteen months. What would you do? The executive immediately said he would close one of the development groups and possibly both, start fresh in the Netherlands with the best people from each, create a single group with full profit responsibility, and add budget overrides, promotions for people who helped and career problems for people who did not. Then he said: that would be a very painful road. Many noses would get out of joint. It took him another nine months to decide the initiative was in fact important.
The point is that as long as strategy remains at the level of intent and concept, the conflicts between competing values stay tolerable. It is the imperative of action that forces a decision about which issue is actually most important. There is no greater tool for sharpening strategic ideas than the necessity to act.
Coherence means the actions do not fight each other. Rumelt's example of incoherence is Ford under Jacques Nasser, who told him brand is the key to profits in the automobile industry and then acquired Volvo, Jaguar, Land Rover and Aston Martin — while the company's older guiding policy of economies of scale was alive and well, with a senior executive insisting you cannot compete unless you produce a million units per year on a platform. So the brand acquisitions were conjoined with actions to put Jaguar and Volvo on a common platform, diluting both. Volvo buyers do not want a safe Jaguar, and Jaguar buyers do not want a sporty Volvo.
He also warns against the opposite failure, a list of non-conflicting but uncoordinated actions. One client's initial strategy was to close a plant in Akron and open one in Mexico, spend more on advertising, and start a three-hundred-and-sixty-degree feedback program. All may be good ideas, but they do not complement one another. They are strategic only in the sense that each requires senior approval. That is sound operational management, not a strategy.
Finally, coherence is imposed. Strategy is visible as coordinated action imposed on a system, an exercise of centralised power used to overcome the system's natural workings. But coordination is costly, because it fights the gains from specialisation — to specialise is roughly to be left alone to do one thing without interruption, and as anyone on a coordinating committee knows, coordination interrupts and de-specialises people. So seek coordinated policies only when the gains are very large. The brilliance of good organisation is not making sure everything is connected to everything else. Down that road lies a frozen, maladaptive stasis.
Part Two. Sources of Power
Part two is a catalogue of the places strategic power actually comes from. Think of it as the set of moves available to you once you have a kernel.
Leverage
Leverage is anticipation, plus insight into what is pivotal, plus a concentrated application of effort.
Anticipation does not require psychic powers. Mostly it means considering the habits, preferences and policies of others, along with the inertias and constraints on their changing. While the sport utility vehicle boom was running, Toyota invested more than a billion dollars in hybrid technology on two anticipations: that fuel economy pressure would eventually make hybrids a major category, and that other manufacturers, once offered the chance to license Toyota's technology, would take it rather than develop their own. Both proved right.
The most striking anticipations Rumelt cites came from Pierre Wack and Ted Newland in Shell's planning group. Wack told him that certain aspects of future events are predetermined: if there is a storm in the Himalayas, you can confidently predict flooding in the Ganges plain in a day or two. The storm they had discerned in nineteen seventy was in the pattern of incomes and populations of key oil-producing countries. Iran, Iraq and Venezuela all had high reserves, large growing populations and ambitious development goals, so they would be strongly motivated to push for price increases. Those increases would in turn make Saudi Arabia and Kuwait realise that oil in the ground might appreciate faster than the dollars it bought once pumped.
Wack also gave Rumelt the sharpest warning about scenario planning ever written. If you do standard scenario forecasting, you end up with a graph with three lines labelled high, medium and low. Everyone looks at it and believes they have taken uncertainty seriously. Then, of course, they plan on medium. But they are missing the risk. The risk is not that the price of oil may be high or low. The risk is that it will go high, suckering you into a major investment, and then turn and dive to low, leaving you with useless assets. Which is exactly what happened in the nineteen eighties.
A pivot point is a natural or created imbalance where a relatively small adjustment can unleash much larger pent-up forces. Rumelt's business example is Seven and i Holdings in Japan, whose president explained that Japanese consumers are easily bored: there are more than two hundred soft drink brands with new ones every week, and a typical seven-eleven displays fifty varieties with seventy per cent turnover a year. The company built a method of collecting information from store staff about local tastes, formed quick-response merchandising teams, and partnered with second and third-tier food manufacturers to bring new own-brand offerings to market quickly and cheaply using those manufacturers' spare capacity. Its advantage in China was completely different: cleanliness and service, in a market where consumers were used to being supplicants at a retail counter.
Concentration is about threshold effects. Below a certain level of effort, nothing happens. Advertising works this way, which is why it can pay to pulse advertising into short bursts rather than spread it evenly, or to roll out a product region by region. Business strategists will often prefer to dominate a small segment over having the same number of customers spread thinly across a large one. Politicians prefer a plan that delivers a clear benefit to a recognisable group over larger benefits spread thinly. And there is a psychological version: it may have more impact on public opinion to completely turn around two schools than to make a two per cent improvement in two hundred.
The illustration is Harold Williams at the Getty Trust, who in nineteen eighty-three had to spend about sixty-five million dollars a year on art. The obvious move was to buy art. As he told Rumelt, all that would really accomplish would be to drive up the price of art and move some of it from New York and Paris to Los Angeles. Instead he decided the Getty would transform the subject: build a complete digital catalogue of all art including dance, song and textiles; educate art teachers; host advanced research on art and society; assemble the best conservation talent in the world. He chose an objective that was nicely scaled to the resources at his disposal, so that his money would make a large and visible difference.
Proximate objectives
A proximate objective is one close enough at hand to be feasible. It names a target the organisation can reasonably be expected to hit, even overwhelm.
Rumelt's demonstration case is the moon landing, which is usually cited as the great example of an audacious vision. He argues the opposite. Kennedy diagnosed the problem as world opinion: the Soviet Union was leveraging its poorer technological base by concentrating on space spectaculars. Being first to land people on the moon would be a dramatic affirmation of American leadership, and America had far greater resources to draw on. Crucially, the objective had been judged feasible. Wernher von Braun had written a memo to the vice president noting that the Soviets led in heavy-lift rockets and would win a race to a manned orbiting laboratory or an unmanned moon vehicle, but that a manned lunar landing required a performance jump by a factor of ten over anything either nation had, which favoured the country with the bigger resource base. Kennedy's speech came a month after that memo. And Kennedy did not merely point at the objective. He laid out the steps: unmanned exploration, larger boosters, parallel development of liquid and solid fuel rockets, and a landing vehicle.
Compare that with the War on Drugs, or with calls for energy independence, which are not proximate because nobody has a feasible path within the existing framework.
The most useful story in this chapter is about ambiguity. Rumelt worked at the Jet Propulsion Laboratory on Surveyor, the unmanned lander. The vexing problem was that nobody knew what the lunar surface was like. It might be soft powder from eons of meteoric bombardment, or a nest of needle-sharp crystals, or a jumble of boulders. Engineers could design a vehicle, but they could not defend any design against someone else's story about the horrors of the lunar surface.
His boss, Phyllis Buwalda, wrote a specification. The lunar surface, she declared, was hard and grainy, with slopes no more than about fifteen degrees, scattered small stones, and boulders no larger than about two feet across. Rumelt looked at it and said it looked a lot like the south-western desert. She smiled and said yes, doesn't it. He objected that she did not actually know what the moon was like. Her answer: the engineers can't work without a specification, and if it turns out to be a lot more difficult than this, we aren't going to be spending much time on the moon anyway.
Her specification was not the truth. The truth was that they did not know. It was a strategically chosen proximate objective, one the engineers knew how to tackle. Five Surveyors landed successfully in nineteen sixty-six and sixty-seven, and the model turned out to be remarkably accurate.
Rumelt generalises: an important duty of any leader is to absorb a large part of the complexity and ambiguity in a situation and pass on to the organisation a simpler problem, one that is solvable. Many leaders fail badly at this, announcing ambitious goals without resolving any of the ambiguity about the specific obstacles. To take responsibility is more than a willingness to accept blame. It is setting proximate objectives and handing the organisation a problem it can actually solve.
He also punctures a common belief. Many writers suggest that the more dynamic the situation, the further ahead a leader must look. This is illogical. The more dynamic the situation, the poorer your foresight will be, and therefore the more proximate a strategic objective must be. The logic becomes that of taking a strong position and creating options, not of looking far ahead. He quotes a description of chess masters, who spend much of a game making moves whose only immediate purpose is to improve their position: increasing the mobility and options of their own pieces, decreasing the opponent's freedom, and imposing stable patterns that produce enduring strength. When enough positional advantage accumulates, it can be cashed in.
The applied example is a business school whose draft plan listed research visibility, alumni giving, a global business program, entrepreneurship enhancement and a sustainability initiative. Rumelt asked them to imagine they were allowed only one objective, and that it had to be feasible. What single feasible objective, when accomplished, would make the biggest difference? By the end of the day they had settled on getting students into better jobs, made concrete as: select ten target firms that should be hiring their graduates but were not, and create faculty committees to study those firms' recruiting practices and build programs to meet their standards. Plus a new course of studies in media management, on the theory that it would attract media people to the school and raise its profile, with two of the ten target firms being media companies.
Finally, what is proximate for one organisation is distant for another, and the reason is layered competence. Rumelt learned this from a former helicopter pilot who explained that when your engine fails you have about one second to get the controls right, and you cannot concentrate on the crisis if flying is not automatic. First you learn to fly. Then at night. Then in formation. Then in combat. Master all that, make it automatic, and you can begin to think about landing on a rolling, pitching deck at sea. To concentrate on an objective necessarily assumes that many other important things are already taken care of. Asking a startup that has not yet mastered coordinating engineering, marketing and distribution to concentrate on opening European offices is pointless. It has not yet learned to fly the business.
Chain-link systems
A system has chain-link logic when its performance is limited by its weakest link. When there is a weak link, the chain is not made stronger by strengthening the other links.
The Challenger disaster is the stark case: an O-ring failed, and no amount of quality in the guidance, the communications or the crew training could compensate. In home renovation, the equivalent is a limiting factor. If a house is near a noisy highway, no amount of marble in the bathrooms will fix its value. As an investor you want limiting factors that can be fixed, like paint, rather than ones that cannot, like highway noise.
The trap in organisations is quality matching. If you run one link of the chain, there is no point investing in making your link better while the other link managers are not. Worse: improving one link alone requires better resources and more expensive inputs, so the system's overall profit actually declines. The incentive to improve is dulled everywhere at once, and the whole system gets stuck in a low-effectiveness state. General Motors between nineteen eighty and two thousand and eight had exactly this shape. Improving the transmission does little good if the knobs fall off the dashboard. Improving fit and finish helps little if the designs remain pedestrian. Improving the designs only raises costs unless design for manufacturability is mastered.
The flip side is the good news. Excellence is the mirror image of being stuck. If a system is chain-linked and you do manage to get every link excellent at once, competitors cannot copy you piecemeal either. They face the same stuck problem in reverse. That is a large part of why IKEA has been so hard to imitate: everything about it, from the design and the flat packing to the store layout and the catalogue, is matched to everything else.
Design
Rumelt's model for design-type strategy is Hannibal at Cannae. Eighty-five thousand Roman soldiers faced about fifty-five thousand Carthaginians. Hannibal arranged his troops in an arc bulging toward the Romans, with his weaker Gauls and Spaniards in the centre and his heavy Carthaginian infantry on the flanks of the bulge. When the Romans hit the centre it fell back, as ordered, and they pushed forward shouting victory. Hannibal's cavalry galloped in wide arcs around the sides, defeated the smaller Roman cavalry, and kept going. The centre bowed inward, then stopped and held. The heavy infantry on the flanks turned in. The cavalry closed the rear. And then the ruthless genius showed: as the Romans' superior numbers pressed into the bowed arc, their ranks were squeezed so tightly that many soldiers could not raise their weapons. Numerical superiority was nullified. At least fifty thousand Romans died that day, including a quarter of the Republic's elected leadership.
Three things show in bold relief. Premeditation: Cannae was designed in advance, not improvised — by definition, winging it is not a strategy. Anticipation: the Romans were not simply enveloped, they were enticed into being enveloped, their own mobility, courage and initiative turned against them. Hannibal did not assume a perfectly rational opponent; he saw an organisation with a history, doctrine and standardised training, led by men with identifiable biases, and he raided the consul's camp the night before to anger him into seeking immediate battle. And design of coordinated action: separated units acting in preplanned cohesion around a central design.
Rumelt uses this to attack the idea that strategy is a choice. Choice and decision evoke someone considering a list of alternatives and picking one. Hannibal was not briefed by a staff presenting four options on a slide. He faced a challenge and designed a novel response. Many effective strategies are more designs than decisions, more constructed than chosen. Doing strategy is more like designing a high-performance aircraft than deciding which forklift to buy.
To feel what design means, imagine tuning a BMW. You cannot make a driving machine out of high-quality off-the-shelf parts; the chassis, steering, suspension, engine and controls all have to be tuned to one another. Vary forty or fifty parameters and you eventually find a sweet spot. Then add price, and you are optimising smile per dollar. Then brand, dealers, reliability, resale value. Then other drivers with other tastes. Then competitors — and the question becomes where you can put more smile per dollar on a driver's face than she can get elsewhere, which forces you onto a much smaller subset of models and customers.
That yields the central trade-off of the book, which Rumelt learned designing the Voyager spacecraft. With three thousand pounds of launch weight, the design was easy: bolt together well-understood subsystems, and the divisions barely had to coordinate. With only twelve hundred pounds, everything interacted. Cutting the power unit's weight meant less power for the radio, which meant a more focused dish, which meant more accurate pointing, which meant better sensors and more attitude-control fuel.
The principle: a design-type strategy is an adroit configuration of resources and actions that yields advantage in a challenging situation. Given fixed resources, the greater the challenge, the greater the need for clever, tight integration. Given a fixed challenge, higher-quality resources reduce the need for it. Resources and tight coordination are partial substitutes.
And tight integration has a cost. A tightly integrated design is harder to create, narrower in focus, more fragile in use and less flexible in the face of change. A Formula One car is faster around a track than a Subaru Forester, but the Forester is useful for far more purposes.
This leads to the arc of enterprise, one of the most useful ideas in the book. A very powerful resource position produces profit without great effort, and the easy life breeds laxity. Xerox's patents on plain-paper copying meant it could sell for three thousand dollars a machine that cost seven hundred to make; it did not need a design-type strategy, and its strategic plans were merely financial projections. It is also human nature to associate current profit with recent actions, even though current plenty is the harvest of planting seasons long past — which is why books get written recommending everyone adopt the successful firm's dress code and its method of allocating parking spaces. It is the disconnect between current results and current action that makes analysing success so hard.
The practical implication: to see effective design-type strategy, look away from the long-successful incumbent and toward the company invading its market space. Canon working around Xerox's patents with reliable desktop copying. The young Microsoft besting IBM. The young Walmart besting Kmart. FedEx pushing aside traditional air freight. Enterprise Rent-A-Car against Hertz and Avis. Nvidia taking graphics chips from Intel. Google taking search from Microsoft and Yahoo. Study Microsoft today and you see a mature giant, reaping past victories and as tied to its installed base and its mix of conflicting initiatives as IBM was in nineteen eighty-five.
The book's best worked example is Paccar, which makes Kenworth and Peterbilt heavy trucks. It is the number two producer in a mature, low-growth, brutally competitive industry, and has averaged sixteen per cent return on equity over twenty years against twelve for its competitors, with remarkably stable profits, and has not lost money since nineteen thirty-nine.
Selling a truck at a premium should be hard. Three obstacles stand in the way of quality leadership: nobody believes you make the longest-lasting trucks until they have already lasted a long time; designing very high quality machinery is not a textbook problem, since designers learn from other designers over time and a company accumulates that wisdom only by being a stable place for talented engineers; and buyers resist paying an up-front premium for future savings even when the numbers are clear.
Paccar's design handles all three. The subtle move is to see quality through the eyes of the owner-driver rather than purely as operating cost. Owner-drivers care about efficiency, but the truck is also their home, office and lounge, and they prize a Harley-Davidson kind of aura even as interiors move toward a Lexus feel. They buy from experienced dealers, choosing from hundreds of customising options, and each truck is built to order. Fleet operators do not care about aura, but they care about driver turnover and idle time; running two drivers halves idle time, which means one is resting in the sleeper and comfort matters again. And when drivers meet at a truck stop, the owner-drivers have the highest status and their opinions carry the most weight, so fleet buyers get pushed the same way by their own drivers.
Every element is specialised to the others. Paccar does not make small trucks, and within large trucks it does not make cheap ones. Flexible manufacturing raises variable costs but gives designers stability. Higher margins create a more loyal dealer network. And it works partly because the industry is not high-growth, so it does not attract huge outside investment: to attack Paccar directly a rival would need new brands, new designs and probably new dealers, and the high-end segment is not big enough to justify that. Rumelt's test is to imagine a Frankenstein truck company assembled from generic parts of the others — medium-priced designs aimed at fleets, dealers aimed at picky owner-drivers, engineers trained to cut costs to the bone. It would not last long.
He also notes what threatens such a design: not specific new products or competitive moves, but changes that undermine the logic of the design itself. If trade rules push shippers toward foreign trucks instead of American owner-drivers, Paccar's position is at risk. If sophisticated computer selling in dealerships undermines the importance of the dealer's own knowledge, that erodes it too.
Focus, and the growth trap
Focus, in Rumelt's technical sense, is a particular combination of policy and market segmentation. He teaches it through Crown Cork and Seal, a can maker that returned nineteen per cent per year to shareholders over thirty-five years in an industry where the big three earned four to five per cent return on assets.
The conventional description is that Crown specialised in containers for hard-to-hold products such as aerosols and carbonated drinks. True but not useful. The industry structure was terrible: beverage companies typically kept two suppliers, can plants were often built to serve a single customer, the products were indistinguishable, and there was a constant threat that the buyer would simply install a can line and do the job itself. Crown's actual pattern was to concentrate on the segments the majors served badly — short runs, rush orders, the customers whose needs did not suit long production runs — and to build its whole set of policies around serving them. The lesson for the reader is method as much as content: to identify a company's strategy, work backwards from the details of its policies, rather than accepting the label the company puts on itself.
The growth chapter is the sharpest warning in the book for anyone running a product organisation. Crown, having built that beautifully focused position, was nearly sunk later by an all-out pursuit of size. And Rumelt describes an adviser at Telecom Italia proposing growth for its own sake. His argument is that healthy growth is not engineered. It is the outcome of increased demand for special capabilities, or of an expanded or upgraded capability. It is the reward for successful innovation, cleverness, efficiency and creativity. Growth achieved by buying revenue is not that. Buying growth to please a stock market is the reverse of strategy: it dissipates the very focus and coherence that produced the value in the first place.
Advantage
Advantage is rooted in asymmetry. Two identical armies on a featureless plain: neither has the advantage. It is the leader's job to identify which of the countless asymmetries are critical and can be turned into advantages.
The memorable image is a venture capitalist named Susan talking to a start-up team that had invented a clever microporous fabric and wanted to build a clothing company. She told them: you have won an Olympic gold medal in the fifteen hundred metres. You have a good chance at the ten thousand metres, and I might back you at that. But you want to switch from running to wrestling gorillas. That's not a good idea and I can't back you at it.
The definition is standard: you have a competitive advantage if you can produce at lower cost or deliver more perceived value than competitors, or some mix. Subtlety comes from realising that costs vary with product and application, and buyers differ in location, knowledge and taste, so most advantages extend only so far. Whole Foods has an advantage over an ordinary supermarket only for certain products and only among shoppers with good incomes who value organic food.
Sustainability requires an isolating mechanism: something that stops rivals duplicating the underlying resources. A patent is the cleanest form. More complex forms are reputations, commercial and social relationships, network effects, dramatic economies of scale, and tacit knowledge gained through experience. And Rumelt notes drily that any claim that a particular I.T. system or training program will provide competitive advantage is a misuse of the term, since an advantage on sale to all comers is a contradiction.
Now the part that most readers find genuinely new. Rumelt argues that having a competitive advantage does not, by itself, make you richer.
His thought experiment is a silver machine left behind by a passing UFO. It produces ten million dollars of silver a year at no cost: no energy, no materials, no labour. The finder sells it to a new owner for a hundred million. Interest rates are ten per cent. Does the new owner have a competitive advantage in silver? Obviously yes; zero cost is about as low as costs go. But the new owner is not one dollar wealthier. Ten million a year is an ordinary ten per cent return on what he paid.
The resolution is to distinguish competitive advantage from financial gain. And then comes the more useful idea, which Rumelt credits to the entrepreneur Stewart Resnick. The silver machine's advantage is real, but it is not interesting, because there is no way for an owner to engineer an increase in its value. The machine cannot be made more efficient. Silver cannot be differentiated. One small producer cannot lift global demand for silver. Owning it is no more interesting than owning a bond.
Resnick's own definition: a business is interesting when I can see ways to increase its value. The typical nut farmer cannot control his destiny; he accepts what the trees yield and what the market pays, and he cannot afford investment in market development, yield research or efficient processing. Roll had a holding large enough to earn back research costs and, crucially, large enough that stimulating national demand for almonds and pistachios would pay. Every Californian grower benefits from higher demand, but Roll was the only one big enough to make the investment worthwhile — and because newly planted trees take seven to ten years to mature, it had a long window to build the branding and the large-scale processing capacity that keeps smaller farmers out. Resnick added a note on ownership: being private is what makes a ten-year manoeuvre possible. He bought that land from oil companies looking ahead one quarter, wanting the assets off their books to improve their ratios.
So wealth increases when advantage increases, which means a value-creating strategy must make progress on at least one of four fronts.
Deepening the advantage means widening the gap between buyer value and cost. The warning here is against believing improvement is natural or can be produced by pressure and incentives alone. In nineteen-oh-nine Frank Gilbreth more than doubled bricklaying productivity without increasing anyone's workload, after millennia of essentially no improvement: he moved the supply pallets to chest height, used a movable scaffold so masons did not carry bricks up ladders, and got the mortar consistency right so a brick could be set with one press of the hand instead of several taps of a trowel. Improvements come from re-examining the details of how work is done, not from cost controls or incentives.
Broadening the extent of the advantage means extending it into new fields, which requires looking away from products, buyers and competitors and looking instead at the skills and resources underneath. DuPont began in explosives; its chemistry led to cellulose, synthetic rubber and paints, that work led to polymer chemistry, which led to Teflon, nylon, Mylar, Dacron and Lycra. But there are two kinds of extension. Those based on proprietary know-how benefit from the fact that knowledge is not used up when applied and may even be enhanced. Those based on customer beliefs — brands, relationships, reputation — can be diluted by careless extension, and a failure in the new arena rebounds onto the core. Disney did it carefully: realising it had defined itself too narrowly, it settled on three guidelines — no bad language, no uncomfortable sexual situations, no gratuitous violence — which let it release Pirates of the Caribbean and National Treasure under the Disney name.
Creating higher demand means engineering an increase in demand for the scarce resources underlying the advantage. Rumelt thinks this is the most basic of business stratagems, ignored by theorists fixated on merely having a sustainable advantage. The Resnicks' pomegranate business is the illustration: they funded research into the fruit's health properties, bought acreage until they controlled a sixfold increase in United States capacity, and then rejected the industry standard of diluting an expensive juice with blander apple or grape juice. Lynda Resnick's concept was the opposite — a hundred per cent pure juice delivering a hundred per cent of the health benefits, sold not as a soft drink but as a new category, a fresh refrigerated antioxidant juice distributed beside fresh produce.
And strengthening isolating mechanisms means making imitation harder. The caution is that if the isolating mechanisms around your methods are weak, rivals will sensibly wait and free-ride on your improvements — which is itself a reason firms stop improving.
Dynamics
Riding a wave of change is Rumelt's phrase for the strategy of capturing the high ground when an industry's structure shifts. His long example is Cisco Systems, which rode three interlinked waves at once: the rise of data networking, the deregulation and digitisation of telecommunications, and the emergence of software as the locus of competitive advantage.
Because forecasting industry change is so hard, he offers five guideposts rather than a method.
The first is rising fixed costs. When the fixed cost of competing rises — because of a step change in required investment, or a new technology that only scale can fund — the industry consolidates, and firms that cannot reach the new scale are squeezed out.
The second is deregulation. Over the preceding thirty years the American government dramatically changed the rules for aviation, finance, banking, cable television, trucking and telecommunications, and in each case the competitive terrain shifted dramatically. Much of what looks like superior performance under regulation turns out to have been a subsidy.
The third is predictable bias in forecasting. People rarely predict that a trend will peak and then decline. If sales are growing rapidly, the forecast is continued growth with the rate gradually easing to normal. That may hold for a frequently purchased product, but for a durable good — flat-screen televisions, fax machines, power mowers — there is an initial rapid expansion, then everyone interested has one, and sales can drop sharply, after which they track population growth and replacement demand. A related bias is copying the visible surface of a winner's strategy: as aviation was deregulated, consultants advised airlines to copy Delta's Atlanta hub-and-spoke system, but Delta's profits had come from subsidised prices on the short-haul rural routes it fed from Atlanta, and those subsidies were disappearing.
The fourth is incumbent response. You must think through the structure of how incumbents will respond, and in particular whether their existing profit streams make responding irrational for them.
The fifth, and the most useful, is the attractor state. An attractor state describes how an industry should work in the light of technological forces and the structure of demand — should, in the sense of an evolution toward efficiency, meeting buyers' needs as efficiently as possible. Cisco's vision of "I.P. everywhere" was an attractor state: all data moving as internet protocol packets whatever the medium, and all information coded into those packets whether voice, text, pictures, files or video, with the intelligence supplied by devices at the endpoints rather than by the network. Competitors were envisioning intelligent networks and value-added services, meaning carriers supplying proprietary protocols and hardware for each service.
The critical distinction between an attractor state and a corporate vision is that the attractor state is based on overall efficiency rather than on one company's desire to capture most of the pie. It has a gravity-like pull, though no guarantee of arrival. And it comes with two useful companions: accelerants that speed movement toward it, and impediments that slow it. For the electric power industry, Rumelt argued the obvious attractor state was nuclear, and the major impediment was a convoluted, highly uncertain licensing process involving local, state and federal authorities and the courts. Applied to news media, he saw the attractor state as specialists along each of three dimensions — territory, frequency and depth — rather than generalists, because with electronic access there is no longer a good reason to bundle local, national and world news together with weather, sports, comics, puzzles, opinion and personal advice. Which is a fair description of what happened.
Inertia and entropy
The other side of riding a wave is understanding why organisations cannot turn. Rumelt names three kinds of inertia.
Inertia of routine. Standard routines are revealed as obsolete by sudden outside shocks. Two years after American airline deregulation, Rumelt was asked to help Continental Airlines think through a four-hundred-million-dollar aircraft purchase. His analysis was that deregulated fares would move to parallel costs: prices and margins would rise on short-haul routes and fall on long-haul ones, so profits would come only from a low-cost operating structure or from cleverness in grabbing short-haul routes too thin to support vigorous competition. The industry consensus was that business travellers would remain price-insensitive. Rumelt's contrary reasoning was simple: business travellers like comfort, but their employers pay, and employers care about cost. Despite the chief executive's speeches about a new competitive spirit, the company's planning, pricing and marketing routines were unchanged from the era of regulation. The new competitive spirit was pure aggressiveness, unalloyed by craft.
Cultural inertia. At A T and T the problem was not the competence of individuals but the work norms and mindsets. Bell Labs did fundamental research, not product development; a request for demonstration code was received as if Boeing engineers had been asked to design toy aeroplanes. Through decades as a regulated monopoly, this culture had flourished, and the breakthroughs of a tiny number of very talented people had been used to justify a contemplative life for thousands of others. Now, facing competition and the opportunities of mass-market computing, it was a huge impediment. Worse, most senior managers did not understand technology well enough to see the problem, and the few who did had almost no chance of changing the character of a laboratory that had produced Nobel laureates.
Inertia by proxy. This is the subtlest and the most commercially dangerous. Sometimes a firm does not respond to an attack because responding would undermine still-valuable profit streams, and those streams persist because its customers are slow to switch. The telephone companies looked asleep while new competitive carriers grew at fantastic rates in the late nineteen nineties. That growth was taken as evidence of competitive superiority, and it unleashed a frenzy of investment and stock appreciation. But the apparent inertia was inertia by proxy, induced by customers who were slow to switch even in the face of dramatic price differences. When the telephone companies finally began to respond in two thousand, the bubble popped, and not a single one of the new carriers survived. Rumelt's warning: inertia by proxy disappears suddenly, when the incumbent decides that adapting matters more than protecting old profits. And the effect is magnified, because the customers who switched away are by self-selection the ones most sensitive to a better offer.
Then there is entropy, which is different from inertia. Inertia is resistance to change. Entropy is the tendency of a well-ordered organisation to become disordered on its own, without any outside shock at all, unless a manager keeps working at it. Rumelt's diagnostic tool for entropy is the hump chart. Whenever you can assign profit to individual products, outlets, areas or segments, sort them from most to least profitable and plot the cumulative total. If there are no cross-subsidies, the bars rise smoothly to a maximum. If some products or locations are being subsidised by others, you get a hump: the bars rise to a peak and then sag downward as the loss-making operations pull the whole down. He describes a retailer whose managers were shocked to see the chart, because it implied dropping something like half their locations. Entropy is why maintaining coherence is continuous work rather than a one-off act.
Part Three. Thinking Like a Strategist
Strategy as hypothesis
Rumelt's argument in part three is that a good strategy is, in essence, a hypothesis about what will work, and a good strategist behaves like a scientist.
The illustration is Howard Schultz. He did not have a theory of American coffee retail; he had an anomaly. On a trip to Milan he noticed espresso bars everywhere, and he noticed that Americans drank weak coffee and did not have such places. His first attempt to reproduce the Italian bar in Seattle was not right, and the concept was adjusted repeatedly. Deduction is only enough if you already know everything worth knowing. In the real world, strategy proceeds by noticing an anomaly, forming a hypothesis about what is going on, and then testing it, with the willingness to be wrong.
This matters for how you should hold your own strategy. A strategy is not a promise. It is your best current explanation of the situation plus your best current bet about what to do. It deserves to be defended vigorously and abandoned when the evidence goes against it.
Using your head
The central psychological problem is what Rumelt calls quick closure. When we come up with an idea, we spend most of our effort justifying it rather than questioning it. Letting go of a judgement is painful and disorienting: to search for a new insight you would have to give up the comfort of being oriented and cast around again, with the risk of coming up empty-handed. Our minds dodge that work, and we are not conscious of the dodge.
But you do not have to be a captive to it. You can choose how you approach a problem. Rumelt says this personal skill — the ability to think about your own thinking, to make judgements about your own judgements — is more important than any strategy concept, tool, matrix or framework.
He offers four mental tools.
The first is the kernel itself, used as a checklist. When insight strikes, it never arrives as a full-blown strategy; it strikes in one of the three elements. It may be an insight into an action, or into an overall direction, or into a diagnosis. There is nothing wrong with a localised insight, but the kernel reminds you that a strategy is an internally consistent argument leading from facts on the ground to a diagnosis, then to an overall directive, then to action. Rumelt's own habit is problem-first: at the start of most consulting engagements the client wants an appraisal of a course of action, and he almost always backs up and tries to build a better diagnosis first.
The second is problem-solution, which is working backwards. People think of strategy as what an organisation does. But strategy also embodies an approach to overcoming a difficulty. So shift your attention from what is being done to why it is being done, from the directions chosen to the problems those choices address. Identifying the obstacles gives you a much clearer picture of the range of possible strategies, and shows you how a change in some factor might radically alter which strategies work.
The third is create-destroy. Overcoming quick closure is simple in principle: look for additional insights. But when people are asked to generate alternatives, they typically add one or two shallow ones — a straw man, or generic options like "walk away" or "more study." A real new alternative flows from a reconsideration of the facts and addresses the weaknesses of the existing options. Creating higher-quality alternatives requires that you try hard to destroy the ones you already have, exposing their fault lines and internal contradictions.
The fourth is the panel of experts, which is how Rumelt does the destroying. He carries a mental panel of people whose judgement he values, and runs an internal dialogue with them before putting an idea in front of anyone real. This works because humans are extraordinarily good at recognising and modelling whole personalities. Thinking through how a particular remembered expert would respond is a richer source of criticism than an abstract framework, because what you get from a person is not a theory but a viewpoint integrated into a personality. Steve Jobs earned his place on Rumelt's panel at a nineteen ninety-seven meeting where a UCLA colleague offered the school's line about being the entrepreneurial school. Jobs asked which Silicon Valley entrepreneurs had come out of the program that he would recognise. The colleague answered truthfully that there were none. "Well, then, you've failed," Jobs said.
Rumelt also insists on practising judgement, and gives a concrete method: commit your judgements to writing in advance. He describes the student who prepares a case thoroughly, sits through the discussion, and thinks "I thought of that" at every turn. What she has missed is the chance to pre-commit and thereby evaluate her own judgement. Committing to a diagnosis in advance increases the probability you will disagree with others, and therefore the chance of learning something. Do the same before any meeting: what issues do you expect, and who will take which position?
Keeping your head
The last chapter is about independent thinking, and it is the least practical and most sobering.
The case study is Global Crossing, which built a transatlantic fibre cable in the late nineteen nineties. The economics looked irresistible: build capacity for one and a half units of cost and sell it for eight. But that is precisely the signal that others will build too. Undersea cable is the worst industry structure imaginable — enormous fixed cost, essentially zero marginal cost, an undifferentiated product, and no way to withdraw capacity. Once capacity exceeds demand, price collapses toward zero, and it stays there. The point is not that the people involved were stupid. It is that a spread that large between cost and price is not an opportunity; it is a warning.
Rumelt then turns to the two thousand and eight financial crisis and argues it was almost certain to occur, and that the mechanism was social herding plus what psychologists call the inside view. The inside view is the tendency to evaluate a situation by its specific details rather than by the base rate of similar situations. It is why everyone believed housing was different this time. Social herding is the tendency to believe that if everyone around you is doing something, they must know something. He draws the parallel to the Johnstown flood, the Hindenburg, the aftermath of Hurricane Katrina and the Gulf oil spill: in each case, warnings existed, base rates existed, and the inside view prevailed.
The question he leaves you with is whether one can be independent without being eccentric, and doubting without being a curmudgeon. His answer is that independent judgement in strategy is not contrarianism for its own sake. It is the discipline of insisting on a diagnosis, checking it against base rates rather than stories, and being willing to hold a position that the people around you find uncomfortable — while remaining genuinely open to being wrong.
What this means if you run a product organisation
A few translations, offered as one reader's reading rather than as Rumelt's own words.
Your roadmap is almost certainly not a strategy. A roadmap is a list of things you intend to do, usually with dates. Rumelt's test is brutal and useful: what is the challenge this responds to, and what is your explanation of why it has not been overcome already? If your quarterly planning document has no diagnosis section, you have a budget with feature names attached.
Watch for the four hallmarks in your own artefacts. Fluff appears as "customer-obsessed platform strategy." Failure to face the challenge appears as a document that names no obstacle, only opportunities. Mistaking goals for strategy appears as an objectives-and-key-results tree with no theory of why the numbers will move. Dog's dinner objectives appear as a roadmap with fourteen themes, none of which can be cut.
The kernel is a document format. Diagnosis: what is going on, in one paragraph, naming the one or two things that are critical. Guiding policy: the approach, which should rule out large classes of action. Coherent action: a small number of moves that reinforce each other, with the things you are stopping named explicitly.
Choose a proximate objective for each planning cycle. Not an aspiration. Something the team can hit, even overwhelm, given what it can already do. And absorb the ambiguity yourself, the way Phyllis Buwalda did: write the specification, own the fact that it is a judgement, and give the team a problem they can solve.
Check whether your product is a chain-link system. If quality depends on every link — onboarding, performance, support, pricing, documentation — then investing heavily in one link while the others stay weak will not move the metric, and worse, it will cost you money and teach the organisation that investment does not pay. Chain-linked problems have to be attacked across all links at once, usually by one leader with the authority to do so.
Ask whether your advantage is interesting. You may well have a real advantage: an install base, a data asset, a distribution relationship. The question that matters for the next three years is whether you can see specific ways to increase its value — deepen it, extend it, grow demand for it, or protect it better. If you cannot name a move on at least one of those four fronts, you are holding a silver machine.
And expect entropy. Coherence is not a state you reach; it is maintenance work. Build the hump chart for your own portfolio at least once a year, and see which parts of the product are quietly subsidised by the rest.
Rapid recap
Here are the things worth keeping.
Strategy is a cohesive response to an important challenge, not a goal, not a vision, and not the pay grade of the decision maker.
The kernel has three parts: a diagnosis of the challenge, a guiding policy for dealing with it, and a set of coherent actions that carry out that policy.
Bad strategy has four hallmarks: fluff, failure to face the challenge, mistaking goals for strategy, and bad strategic objectives, which come as either a dog's dinner or blue sky.
Bad strategy comes from three places: unwillingness to choose, template-style planning, and the belief that attitude is enough.
If you fail to identify and analyse the obstacles, you don't have a strategy; you have a stretch goal, a budget, or a wish list.
Universal buy-in usually signals the absence of choice.
Underperformance is a result, not a challenge. The challenge is the reason for it.
Good strategy is unexpected, because most organisations spread resources to placate interests rather than concentrating them.
Leverage is anticipation, plus a pivot point, plus concentration.
A proximate objective is close enough to be feasible. The more uncertain the situation, the more proximate your objective must be. A leader's job is to absorb ambiguity and hand the organisation a solvable problem.
In a chain-link system, strengthening one link does nothing. Excellence is the flip side of being stuck, and it is why the excellent are hard to copy.
Strategy is more often designed than chosen. Given fixed resources, the greater the challenge, the tighter the integration required.
Resources and tight coordination are partial substitutes. Strong resources breed laxity, which is why upstarts teach design better than incumbents do.
Having a competitive advantage does not make you richer. Wealth increases when advantage increases: deepen it, broaden it, raise demand for it, or protect it better.
An advantage is interesting only when you can see ways to increase its value.
Healthy growth is a reward for successful innovation, not something to be bought.
Watch for attractor states: how an industry should work on efficiency grounds, as distinct from what any one company wishes would happen.
Inertia comes in three forms: routine, culture, and proxy. Inertia by proxy ends suddenly, and it has bankrupted many attackers who mistook a sleeping incumbent for a beaten one.
Entropy is continuous. Coherence decays on its own.
Fight quick closure with four tools: the kernel as a checklist, problem-solution to work backwards to a diagnosis, create-destroy to force real alternatives, and a panel of experts in your head to do the destroying.
Write your judgements down before the meeting. It is the only way to find out whether they were any good.
And finally, the sentence to carry out of the book: a good strategy honestly acknowledges the challenges being faced and provides an approach to overcoming them. Everything else is exhortation.