Name the Obstacle First

The Core Insight

In September 1997 Apple sat two months from bankruptcy. Steve Jobs cut fifteen desktop models to one and killed the printers and peripherals. He dropped five of six national retailers, moved manufacturing to Taiwan, and cut inventory by more than 80 percent. Apple held under 4 percent of the personal computer market, and Jobs described the long-term plan as waiting for the next big thing.

Richard Rumelt wrote Good Strategy Bad Strategy in 2011 to separate strategy from its impostors. A strategy is a cohesive response to a challenge someone named and analyzed first. Its anatomy holds three parts, which Rumelt calls the kernel: a diagnosis, a guiding policy, and coherent action. The Apple cuts are the third part, and they hold together because the first two were done.

Most leaders believe strategy is ambition made concrete: a vision, a growth target, a list of goals with owners and dates. Rumelt believes strategy is a judgment about one obstacle plus the coordinated actions that deal with it. Without a named obstacle you hold a stretch goal, a budget, or a wish list.

Bad strategy is the avoidance of that judgment. Choosing one path means giving up the others, and a leader who cannot pay that price writes goals instead. Underperformance is a result, and the challenge is the reason behind it.

The Framework

The kernel has three parts, and a document missing one of them is something else.

• Diagnosis names the nature of the challenge and marks a few features of the situation as the critical ones.
• The guiding policy is the overall approach for dealing with that challenge, a signpost rather than a destination.
• Coherent actions are coordinated steps that support each other and carry the guiding policy into the world.

A diagnosis is a judgment about what the facts mean, and for messy problems it stays an educated guess. Rumelt prefers the one that points at something a leader can act on.

Nvidia shows all three parts at once. The diagnosis: demand for graphics has no natural ceiling, and the Silicon Graphics pipeline can migrate onto one chip. The guiding policy was a substantial upgrade every six months against an industry cycle of eighteen. Three development teams overlapped on eighteen-month schedules, simulation cut re-spin delays, and one unified driver served every chip. That cycle held Nvidia in the leading position about 83 percent of the time. The GeForce 256 put the whole pipeline on a 100 dollar chip. The 1992 Reality Engine it outran cost more than 100,000 dollars.

Bad strategy carries four signatures, all visible on the page.

• Fluff restates the obvious in heavy words, as when a bank described its business with a phrase meaning that it is a bank.
• Failure to face the challenge leaves nothing to judge, because a problem nobody defined has no better or worse answer.
• Goals stand in for strategy, as in the plan Rumelt was handed demanding 20 percent growth and 20 percent margins.
• Objectives scatter into a to-do list, like the city plan whose 178 action items included creating a strategic plan.

Bad strategy is the active avoidance of choosing. Digital Equipment argued between Boxes, Solutions, and Chips, and the managers' rankings cycled with no stable majority. Robert Palmer picked Chips in 1992, five years late, and Compaq bought the company in 1998. Universal agreement usually marks the absence of a choice.

Key Ideas

Focus Is Coordinated Policies Aimed at One Target

John Connelly rebuilt Crown Cork and Seal in the early 1960s around the work the majors avoided: short production runs. The majors chased long runs for captive customers and earned 4 to 5 percent on assets. Crown ran smaller plants, kept at least two customers in each, and held spare lines idle for rush orders. Its price per can ran roughly forty to fifty percent above the majors, and Crown was fifty to sixty percent more profitable. Shareholders took 19 percent a year for thirty-five years.

Focus has two halves: policies coordinate until their effects overlap, and the combined power gets aimed at one segment. Rumelt reaches that target by method. Look at how rivals earn their living, list every policy that departs from the norm, then infer what the departures share.

Coordination is what makes a strategy hard to copy. Wal-Mart and Kmart both adopted bar-code scanners in the early 1980s, and only Wal-Mart profited. The scanner paid off inside a network of 150 stores fed by one distribution center.

Give the Organization a Problem It Can Solve

A proximate objective sits close enough that the organization can expect to hit it. Scientists argued whether the lunar surface was deep powder, glassy needles, or boulders, and the argument blocked every design decision at JPL. Phyllis Buwalda wrote a specification anyway: hard and grainy, slopes no steeper than about fifteen degrees, boulders no larger than about two feet. Engineers cannot work without a specification. Five Surveyors landed in 1966 and 1967, and Apollo 12 set down two hundred yards from Surveyor 3.

Kennedy's moon shot was proximate for the same reason. Von Braun wrote that a landing needed a jump in rocket performance by a factor of 10, which erased the Soviet lead.

The rule inverts the usual planning instinct: the more uncertain the situation, the closer the objective must sit. A leader absorbs the ambiguity and hands down a problem the team can solve.

The Weakest Link Prices Every Other Link

A chain-link system performs no better than its weakest part. An O-ring failed on January 28, 1986, and the Challenger broke apart above the ocean. A stronger engine changes nothing about that day.

Improving one link alone pays nothing, and better inputs cost more, so the improvement lowers profit. That is how organizations get stuck, and Rumelt reads decades of General Motors as the same pattern.

Getting unstuck takes sequence. Marco Tinelli found three weak links in a Lombardy machine company in 1997: build quality, the technical competence of the sales force, and cost. He ran twelve months on quality alone, then a campaign on sales, then nine months on cost. He suspended local measurement and rewards and absorbed the losses centrally.

The same linkage makes excellence durable. IKEA stayed uncopied for more than fifty years, because each activity is efficient, the activities depend on each other, and the grouping is odd.

Design Substitutes for the Resources You Do Not Have

Rumelt states the trade-off plainly: given a fixed bundle of resources, a bigger challenge demands tighter integration of resources and actions. Better resources reduce the need for clever design, and tight designs are narrower and more fragile.

Paccar returned 16 percent on equity over twenty years against 12 percent for its competitors. The whole design points at the owner-driver, whose truck is home, office, and status. Experienced dealers build to order, and premium prices fund their loyalty. A 110,000 dollar Kenworth run 125,000 miles a year costs another 115,000 dollars a year to operate before wages.

A design also decays without maintenance. In 1921 Ford held 62 percent of the market with the 495 dollar Model T, and six GM divisions crowded into one price band. Sloan gave each brand its own price range, and GM led the industry by 1931. By 2008 GM sold nine sedans and coupes at 25,500 dollars against Toyota's two, and filed for bankruptcy in June 2009.

Advantage Pays Only While It Grows

Advantage sits in an asymmetry a rival cannot copy, held there by a patent, a brand, or tacit knowledge. An advantage on sale to every buyer is a contradiction.

Rumelt separates advantage from money with a machine that makes 10 million dollars of silver a year at zero cost. At 10 percent interest it sells for 100 million dollars, and the buyer earns an ordinary return. eBay ran a 26 percent after-tax margin and a 13 percent return on assets in the year to December 2009. Its market value sat flat for more than seven years.

Value rises on four fronts. Deepening widens the surplus between buyer value and cost, and broadening extends the skills into new products. The third creates demand for what the resource makes, and the fourth strengthens the isolating mechanisms. POM Wonderful took the third route, growing pomegranate planting to six thousand acres and creating the pure juice category it dominated by 2004.

Growth itself is an outcome. Crown's revenue grew 3.1 percent a year through the 1980s while shareholders earned 18.5 percent against 8.6 percent for the S&P. Under Avery the company bought twenty others and became the world's largest container maker. Revenue growth fell to 2.4 percent a year, and return on capital fell from 15.3 percent to below 5. The stock fell from 55 dollars to 5 dollars.

The Crowd Is Not a Diagnosis

Rumelt gives five guideposts for reading a wave of change. They are rising fixed costs, deregulation, predictable forecast errors, incumbent inertia, and the state an industry moves toward on efficiency grounds. Inertia is the one a founder can plan against. Big-city telephone companies withheld DSL, one-third the speed of a T1 line at one-thirtieth the price, to protect T1 revenue.

Global Crossing built the AC-1 cable for 750 million dollars, about 1.5 million dollars for each STM-1 circuit. It sold circuits at 8 million dollars against a consortium price of 18 to 20 million dollars. By the end of 1998 the company sold 35 percent of capacity for 950 million dollars. The market valued it at 19 billion dollars, then 38 billion six months later.

The structure said the opposite of the price: a commodity with open entry, non-proprietary technology, collapsing capacity cost, and sunk capital. The next cable promised 2,048 circuits for the same 750 million dollars, and Atlantic plans for 2001 reached 16,384 circuits. The price per circuit fell from 6.5 million dollars to 2 million dollars in late 1999, and to 325,000 dollars in early 2002. Global Crossing filed for bankruptcy in December 2001.

The rebuttal at the time was the share price itself. Rumelt calls that a closed circle, where no price settles the overcapacity question.

Practical Applications

Open the current strategy document and delete every sentence that states a goal. Read what remains. An empty page means the diagnosis was never written, and writing it is this week's work.

Write the challenge on one page, and name the obstacle rather than the shortfall. Flat revenue is the result you already see. The page carries the reason behind it.

Pick one proximate objective for the quarter, close enough that the team expects to hit it. The test is whether an engineer can write a specification from it.

Find the weakest link before funding the strongest one. Money spent on a link that does not bind returns nothing, so run one campaign at a time. Rumelt measures each unit by the gain it adds to operating profit. In one chain of 28 garden outlets, the best store added 1.05 million dollars and the worst subtracted 0.97 million dollars. Closing the five worst and copying what the best store did doubled profit in two years.

Who This Is For

Founders writing a first real strategy document get the most from this book, along with operators handed a plan that is a list of targets. The kernel works as a diagnostic on someone else's plan in under an hour.

Skip it if you want a planning template or a forecasting method. Rumelt offers no procedure that generates a diagnosis, and he says so.

Read the cases as retrospective. Rumelt chose examples whose outcomes he already knew, and several of the sharpest ones are anonymized client work no reader can check. The kernel also does not settle which diagnosis is right. He gives three rival readings of Starbucks in 2008, after return on assets fell from 14 percent to about 5.5 percent. None of them is provable. The book rejects a bad strategy faster than it produces a good one, and the 2011 vantage leaves its closing bets unresolved.

The Decision

Run the kernel against your own strategy document this week, line by line. Mark every sentence as diagnosis, guiding policy, or action. Sentences that fit none of the three are goals, and the goals were never the missing part.

The document passes three tests. A stranger reads your diagnosis and names the same obstacle. The guiding policy rules some options out. The actions support each other rather than sit in a list.

A failed pass costs an afternoon of writing and one hard choice. Intel took more than a year to leave memories after Moore and Grove agreed on the exit. The 386 then made it the largest semiconductor company in the world by 1992.

Cut the document to one page: the obstacle, the approach, and the actions that hold together. If it does not fit on the page, the thinking is not finished.