Good Strategy, Bad Strategy | Richard Rumelt

ELI5/TLDR

Most things companies call a “strategy” are not one. They are a wish list of goals, some inspiring words, and a list of everything everyone wants. Richard Rumelt, who has spent his life on this, says a real strategy is much smaller and more concrete: figure out the single hardest problem you actually face, then decide on a few connected things you will do about it. He suggests dropping the word “strategy” entirely and calling it an action agenda — what are we going to do about this problem, starting now.

The Full Story

A strategy is a response to a challenge

Rumelt strips the word back to its root. “Strategy” comes from the Greek strategos — Athens elected leaders to deal with the issue of the day. Persians invading, plague in town, no money for the temple. So at its core, a strategy is always a way of dealing with a specific challenge. Not a vision. Not a set of values. A response to a problem.

“Strategy is always about dealing with an issue, a challenge, a problem.”

The challenge can be a threat (a new technology is killing your market) or an opportunity (you stumbled onto something and don’t know what to do with it). Either way, there is a hard thing in front of you, and strategy is what you do about it.

The kernel — three parts, all required

This is Rumelt’s framework, and it is deliberately small. A real strategy has three pieces, and if any one is missing, what you have is something else.

1. Diagnosis. Figure out what is actually going on. You can’t understand all of reality, so a diagnosis is partly a decision about what to pay attention to — and a hypothesis about how things connect. Crucially, diagnosis is not just “understand the world,” it is “understand what makes this hard.” Rumelt’s favorite question to a client is exactly that: what makes it hard? A CEO says “we want to open in Australia.” He replies: then just tell someone to do it — why are you and I talking about it? Push, and the real obstacle comes out: we don’t know anyone there, they’ll kick us out.

2. Guiding policy. The overall approach — how you are going to deal with the situation. This is where most lists die. Rumelt had a client with seventeen “priorities.” That is not a policy; it is a wish list.

“You wouldn’t want to be in a commercial airplane and hear the tower say to the pilot, I’m giving priority to the following three planes on runway five.”

Priority means first. It does not mean the grab-bag of everything that might matter.

3. Coherent action. The part people leave out because they like strategy to feel lofty and conceptual. You have to actually do something, and the things you do must not fight each other. Don’t say you’ll burn less oil while importing more. Don’t promise to grow faster and raise return on equity, when one usually costs the other. (His phrase for that contradiction: “this is baby talk.” Yet CEO after CEO stands up and says it.)

His concrete example: Microsoft adapting to AI. Diagnosis — see the threat and opportunity. Guiding policy — invest in a leading AI player and fold it into search. Coherent actions — actually do those specific things. Not rocket science. The mystery to him, after a lifetime of this, is why so many organizations don’t do it.

Bad strategy has tells

The chapter people emailed him about most. Bad strategy is not just the absence of good strategy — it has recognizable signatures:

Goals dressed up as strategy. A set of profit or performance targets. “Grow faster” is not a strategy, it is a statement of values, like “I’d like to be taller or have more hair.” The strategy is what’s holding you back, and what you’ll do about it.
No diagnosis. He cites a US plan for getting 17 intelligence agencies to “work together more effectively.” You don’t need to be a spy to translate that: these agencies aren’t cooperating, and nobody has asked why it’s hard after 34 years. The barrier is never named, so nothing changes.
Fluff. Fancy, abstract words — word salad — used to make a situation sound more strategic than the thinking behind it.
Incoherent actions. Doing A and B when A and B undermine each other.

Power — the reason it can work

A strategy needs a reason it should succeed. Rumelt calls that reason power (he avoids “advantage” because power shows up in more forms). It always rests on an asymmetry — something that makes the contest not a coin flip. Two equally matched fighters: 50/50. To win reliably you need to be different in some exploitable way.

Power can be many things: being first to recognize something, a reputation, relationships, an invention, a customer base, the ability to attract the best people for a while. Today the dominant form in tech is network effects — the product gets more useful as more people use it (the idea traces back to telephones; a phone connected to no one is useless). Social networks, Amazon’s convenience moat, and now AI, where the size of your training data may compound into a lead competitors struggle to close. He notes Google’s search keeps improving on the sheer volume of searches it sees, which is why Bing struggles to catch up.

But inherited power can curdle. IBM’s strength under Gerstner was that big corporations trusted and relied on it — so he leaned into serving them. Then computing moved to the cloud, and IBM’s biggest customers were the most reluctant to move (their IT departments didn’t want layoffs). The very asymmetry that saved IBM later weighed it down.

For a small company, the move is to redefine your space small enough that you can actually see your asymmetry — own a niche the market leader doesn’t.

Insight, and where it comes from

The Crux — his later book, named after the hardest single move on a climb — pushes one idea: focus on the hardest part of the problem, and the way through often reveals itself. He runs through designers doing exactly this. I.M. Pei needed an entrance to the Louvre that wouldn’t obscure the palace, so he made it transparent — a glass pyramid (and made it a pyramid because a flat glass roof would collect dirt). The Voyager team couldn’t pinpoint Jupiter precisely enough to navigate, so they photographed it from halfway out against the distant stars to triangulate.

“Insight is not magical. It comes from immersing yourself in the nature of the problem.”

We want to innovate but we’re scared of insight, he says, because it isn’t guaranteed to arrive on schedule. So people reach for a strategy off a list of common strategies instead of doing the hard study that produces a real one.

His method for getting better at diagnosis: read history and biographies. There is no science of strategy the way there is physics — no equation tells you how thick the beam must be. Strategic thinking runs on analogy to past human experience, so the more situations you’ve studied, the more planks you have to grab when the water gets confusing. (His image: you’re thrown overboard, you grab the first floating plank — the first explanation that comes to mind — and the most important mental tool is to think again and ask if there’s a bigger plank over there.)

Organizations are where strategy goes to die

The hardest part of strategy often isn’t the thinking — it’s the politics. Donald Rumsfeld told him the Defense Department had an expert on everything, but they all disagreed and each carried a private agenda: a contract to win, a company to protect, an idea to push. Every piece of information arrives with an agenda attached. Inside any organization, action means changing what people do and who holds power — someone who was the alpha may not be anymore. That is why people freeze.

His blunt prescription: you need a hierarchy and a decider. At some point someone has to say “we’re doing this, Bob’s in charge of that part.” He’s pointed about a cultural drift — the bookstore management section is now all leadership, and most of that is about perfecting yourself in the hope people will magnetically follow, so you never have to do the embarrassing thing of telling someone what to do. Strategy has been displaced by mission statements; management has been displaced by leadership.

Nokia is his cautionary tale: it replaced engineers with lawyers and accountants who lacked feel for the hardware-software problem, then installed a matrix structure that so diffused power that no one was actually in charge. The CEO could pound the desk about Apple’s touchscreen, but there was no one with the authority to go build one.

Founders are making a bet

For a startup, strategy is a bet under heavy uncertainty — like a wildcatter drilling for oil. Be clear about the nature of the bet, because reality reveals itself only in pieces. Research on Silicon Valley startups: nearly all begin aiming at one product-market target, and the ones that survive are the ones that switch — wrong customer, different customer, different need — pivoting over a year or two until something clicks. So a founder has to hold two minds at once: convinced you’ll win, yet willing to ship and move when it isn’t working.

His hunting tool for opportunities — the “value deny”: what should you be able to buy but can’t? (He couldn’t buy reliable baggage insurance; in Hong Kong you can drop luggage downtown and it’s checked at the airport, but not in the US.) And the “perfect X” exercise: design the perfect window — lets in light but not when you want dark, lets in air but not noise or bugs — and the gaps become business ideas. Salesforce started from “how should this be — not a computer, a web page, like Amazon for software.”

Key Takeaways

A strategy is a response to a specific challenge, not a vision, mission, or set of values.
The kernel has three required parts: diagnosis (what’s going on, and what makes it hard), guiding policy (the overall approach), and coherent action (a few connected things you’ll actually do). Missing any one means it’s not a strategy.
Rename it: call it an action agenda, not a strategy — “what are we going to do about this problem,” starting now (months to a few years out, not a ten-year vision).
Diagnosis means understanding what makes the problem hard, not just understanding the world. Ask “what makes it hard?”
Bad-strategy tells: goals masquerading as strategy, no diagnosis, fluff/word-salad, and self-contradicting actions.
“Priority” means first — a list of seventeen priorities is a wish list, not a policy. Focus is the source of power; each extra yes risks turning a good strategy into a bad one.
Power rests on an asymmetry — something that makes the contest not 50/50. Forms include being first, reputation, relationships, inventions, customer base, talent, and network effects.
Inherited advantages can become liabilities (IBM’s loyal big customers slowed its cloud shift).
Small players win by redefining the space small enough to find their own asymmetry.
The Crux: focus on the single hardest part of the problem; insight tends to appear when you immerse in the difficulty, not from picking a strategy off a list.
Strategic skill comes largely from analogy to history — read histories and biographies, not just theory.
Organizations stall on strategy because of diverse interests and fear of action; you need a hierarchy and a decider.
Founders are making a bet under uncertainty; survivors pivot until something clicks while holding conviction and adaptability at once.
Idea-finding tools: the “value deny” (what should you be able to buy but can’t?) and “design the perfect X.”

Claude’s Take

This is one of the cleaner hours of business thinking you’ll find, and the framework holds up because it’s small enough to actually use. The kernel — diagnosis, guiding policy, coherent action — and the bad-strategy tells are genuinely portable; you can hold any document up against them and watch most “strategies” fail the test. The single best reframe is “action agenda.” It quietly defuses the whole mission-vision-values industry by asking the only question that matters: what are we doing about the hard thing, starting now.

What keeps this from a 9 or 10 is that the back third leans on Rumelt’s anecdotes more than his framework — the design stories (Pei, Voyager, SpaceX) are charming but do a lot of motivational lifting, and “immerse yourself and insight will come” is true but not very actionable. There’s also a faint everything-is-obvious-once-stated quality: he keeps saying strategy isn’t mysterious, which is comforting but slightly undersells how hard the diagnosis step actually is in practice. The transcript is also auto-generated and garbled in places, so a few quotes here are cleaned-up reconstructions of clearly-intended meaning.

Score: 8. The kernel and the bad-strategy chapter alone are worth the listen, and the “action agenda” rename is the kind of thing that changes how you read every strategy doc afterward.