Good Strategy Bad Strategy: Rumelt's Kernel as a Fluff Detector

This is one of RoadmapOne ’s articles on Strategy for Product Leaders .

Good Strategy Bad Strategy is Richard Rumelt’s 2011 book and framework for distinguishing real strategy from aspirational fluff. At its core is the kernel: three required elements — a clear-eyed diagnosis of the specific challenge, a guiding policy for addressing it, and a set of coherent actions that execute the policy. Any strategy missing one of the three is what Rumelt calls bad strategy: goals, vision, or values dressed up as strategy.

Most product strategy documents fail Richard Rumelt’s test. They are lists of ambitions (“drive category leadership”, “become customer-obsessed”, “achieve operational excellence”) dressed up in slide furniture and presented with the authority of strategy. They are not strategy. They are what Rumelt calls fluff: aspirational word salad that describes no specific challenge, commits to no specific response, and produces no specific action. You could put most product strategy documents into a different company’s deck without changing a word, and nobody would notice.

Good Strategy Bad Strategy, Richard Rumelt’s 2011 book, is the best diagnostic I know for telling the difference. Rumelt, a UCLA Anderson strategy professor who spent decades watching real strategy fail in real boardrooms, argues that every good strategy contains a three-part structure he calls the kernel: a clear-eyed diagnosis of the challenge, a guiding policy for addressing it, and a coherent set of actions to execute the policy. Anything else—goals, vision statements, values, high-level themes, lists of initiatives—is either fluff or input, not strategy.

The framework is uncomfortable because it is ruthless about what strategy is not. It is not a goal. It is not a list of objectives. It is not an OKR tree. It is not a set of values. It is not a vision. Those are useful things—but they are not strategy, and confusing them with strategy is the single most common failure mode Rumelt sees. I’ve seen it on every board I’ve sat on. Product leaders who internalise Rumelt’s kernel stop writing strategy documents and start writing strategy.

Sign up (free) to translate strategy into squad allocation

The Kernel of Good Strategy

Rumelt’s kernel has three parts, and all three are required:

1. Diagnosis. A clear-eyed statement of the specific challenge the organisation faces. Not a vague environmental scan—a named, specific, bounded problem. “Our enterprise segment is churning because our security posture does not meet the requirements of IT buyers in regulated industries” is a diagnosis. “We need to do better in enterprise” is not.
2. Guiding policy. The overall approach chosen for addressing the diagnosed challenge. It narrows the space of possible actions by committing to a direction. “We will prioritise achieving SOC 2 Type II and ISO 27001 within twelve months, and we will refuse enterprise deals that require additional certifications until these are in place” is a guiding policy. “We will invest in compliance” is not.
3. Coherent action. A set of concrete, mutually-reinforcing steps that execute the guiding policy. Actions are coherent when they pull in the same direction—when doing action A makes action B work better rather than fighting it. A coherent action set for the compliance example includes hiring a CISO, pausing any non-compliance-aligned enterprise work, restructuring the sales pipeline to defer deals that require additional certifications, and aligning the roadmap to prioritise security features first.

That’s it. Three parts. If any one is missing, vague, or incoherent with the other two, what you have is not strategy—it is aspiration, planning, or activity.

The elegance of the kernel is that it forces the hard conversation forward. You cannot write a guiding policy without a diagnosis, because the policy has nothing to be a policy about. You cannot design coherent actions without a guiding policy, because actions have no direction. Skipping the diagnosis—which is where most organisations skip—produces activity without strategy. Doing lots of things. Having lots of initiatives. Looking busy. Making no material progress on any specific problem, because no specific problem was ever named.

What Rumelt Calls Bad Strategy

Rumelt identifies four hallmarks of bad strategy, and they are worth memorising because every one of them hides inside a document that looks professional:

Fluff. Strategy written in superficial abstractions—“leveraging synergies”, “holistic transformation”, “customer-centric innovation”—that say nothing specific. The test: could you take the fluff out and have anyone notice? If not, it’s decoration.

Failure to face the challenge. The document describes a desired state but never names the specific obstacle preventing that state from being achieved. The “challenge” section is either missing or pitched so broadly (“increasing competition”, “changing customer expectations”) that it produces no actionable insight. A strategy that doesn’t face a challenge is a wishlist.

Mistaking goals for strategy. The document presents a list of ambitious goals—“double revenue”, “enter three new geographies”, “achieve best-in-class NPS”—as if the statement of goals were itself a strategy. It is not. Goals are what you want. Strategy is how you will actually get it, given the specific obstacles in the way. This failure mode is endemic in OKR-shaped organisations: the OKR framework produces excellent goals, but OKRs are not a substitute for OKR-driven strategy that names how those goals will be achieved against specific competitive and market realities.

Bad strategic objectives. The document contains specific objectives, but the objectives are either impossible to achieve given current resources, incoherent with each other, or disconnected from the diagnosis. Fifteen top priorities, seven “horizons”, objectives that contradict each other when executed—these are the signatures.

If any of these four apply to your current product strategy document, Rumelt’s position is that you do not have a strategy. You have an artefact that looks like strategy. The distinction matters.

Why Most Product Strategy Skips the Diagnosis

The diagnosis is the part of the kernel organisations most often skip, and it’s worth understanding why. Writing a diagnosis is uncomfortable. A proper diagnosis names a problem you cannot immediately solve, surfaces choices the organisation has been avoiding, and often implicates past decisions the current leadership team made themselves. The diagnosis makes the gap between aspiration and reality undeniable, in public, in writing, with an audit trail.

Compare that with writing goals. Goals feel generative—you’re declaring the future you want. Goals are celebrated, applauded, rallied around. Nobody gets in trouble for writing ambitious goals. Diagnoses, by contrast, are uncomfortable to deliver and uncomfortable to receive. A diagnosis that says “we have over-invested in features for customers who do not pay us enough to justify the capacity” is politically expensive. A goal that says “we will grow enterprise revenue by 40%” is politically free.

The result is predictable. Most organisations produce strategy documents that are all goals and no diagnosis. They define the mountain they want to climb without naming the crevasse between them and the mountain. The coherent-action section then becomes a list of initiatives pointed vaguely uphill, none of which engages with the crevasse, because the crevasse was never named.

This is the specific failure mode the kernel is designed to prevent. The discipline of diagnosing first forces the organisation to confront the obstacle before celebrating the destination. Everything downstream—the guiding policy, the coherent actions, the OKRs, the business case/PID , the roadmap allocation —improves in quality when the diagnosis is sharp.

Sign up (free) to apply the kernel to your product roadmap

The AI-Era Reframe: Diagnosis Is Now the Scarce Work

Rumelt’s kernel was written when building a competing product was expensive. In that world, the guiding policy and coherent action were at least partially about resource allocation—how to spend scarce build capacity against the diagnosed challenge.

AI has changed which half of the kernel is scarce. Build cost has collapsed to near zero. Two engineers with current AI tooling can ship a credible product in weeks that would have taken a team of twelve a quarter three years ago. The coherent action side of the kernel is now relatively cheap to execute.

What has become brutally more expensive is the diagnosis. When anyone can build anything, the competitive advantage collapses into which problem you chose to solve for whom. The diagnosis—the specific, bounded, named challenge—is now the scarce input. A sharp diagnosis produces a coherent product that wins its segment. A soft diagnosis produces a generic product that loses to anyone with a sharper diagnosis of the same market, because they can match your feature set in weeks.

For product leaders in 2026, this has a concrete implication: the single most valuable hour you spend on strategy is the hour you spend sharpening the diagnosis. An extra hour of feature planning produces linear returns. An extra hour of diagnosis produces exponential returns, because a sharper diagnosis makes every subsequent coherent action more effective. This inverts the time allocation most product teams default to: they spend 5% of strategy time on diagnosis and 95% on planning. It should be the opposite.

The Fluff Detector for Product Strategy

Here is a concrete test for any product strategy document. Read it, then try to answer three questions:

1. What is the specific challenge? State it in one sentence, naming the customer segment, the competitive reality, and the bounded obstacle. If you have to reach for abstractions (“in a rapidly changing market”, “to remain competitive”), the strategy has no diagnosis.
2. What is our chosen approach? State it in one sentence, committing to a direction that rules out other directions. If the “approach” could be claimed by any of your competitors, it is not a guiding policy—it is a platitude.
3. What are we actually doing, and why do those actions reinforce each other? List the top three actions. Explain why action A makes action B work better. If the actions are independent initiatives that happen to be running in parallel, you have a portfolio—not a coherent strategy.

If you cannot answer all three questions, the document is not strategy. It is an aspirational artefact that looks like strategy but will not survive contact with the market. This test is especially useful during board review: a CPO presenting a strategy document that does not survive the three questions is, charitably, still forming their strategy. Uncharitably, they are covering for not having one. The board’s job is to force the conversation, not to nod through the artefact.

This test also reveals why so many OKR programmes stall. An OKR tree is, typically, the “coherent action” half of the kernel—without a diagnosis or guiding policy above it. The OKRs point at goals, the goals point at vision, the vision points at values, and nowhere in the stack is there a specific diagnosis naming the challenge the OKRs are supposed to overcome. The tree grows; the organisation does not move. Fixing this does not require abandoning OKRs—it requires inserting the two missing parts of the kernel above them. OKRs work well as coherent action. They fail as strategy substitutes.

Sources of Power for Product Leaders

Rumelt’s book spends a long chapter on what he calls sources of power—the structural leverages that turn coherent action into disproportionate results. Most of them apply directly to product leadership:

Leverage. Strategy produces leverage when a small input triggers a large output—when the action compounds or catalyses something beyond itself. For product leaders this usually lives in the capability layer: investing in a platform that a dozen squads can build on top of produces more leverage than building twelve separate features.

Proximate objectives. Rumelt’s term for an objective that is close enough to achieve, specific enough to galvanise action, and large enough to matter. Distant objectives (“triple revenue in five years”) produce energy but no focus. Proximate objectives (“sign three SOC 2 enterprise customers in the next six months and use them as the template for the next twenty”) produce both. Set proximate objectives in the next twelve to eighteen months; let the distant vision live higher up the stack.

Chain-link systems. A system whose performance is determined by its weakest link. Product organisations are often chain-link systems in disguise: you can have brilliant engineering, but if the sales team can’t sell to enterprise, the product won’t reach enterprise. Strategy for chain-link systems means identifying the weakest link and fixing it specifically, rather than improving across the board. The improvement that matters is in the binding constraint. Everything else is noise.

Design. Rumelt uses “design” to mean the deliberate shaping of a set of mutually-reinforcing choices—segments served, products offered, channels used, capabilities built—so that the whole is more coherent than the sum of its parts. For product leaders this is where the kernel meets the objective tagging discipline: the coherent portfolio is one where every Objective, every squad allocation, every WIP is pulling in the same direction as the guiding policy, and the tags make the coherence (or incoherence) visible.

Focus. The deliberate narrowing of activity to a small number of high-leverage problems. This is one of the hardest moves in product, because every stakeholder has a legitimate ask and narrowing means explicit refusal. Rumelt: focus is always at the expense of something. If nothing is being refused, you have not focused—you have declared an intention.

Rumelt and Playing to Win: Companions, Not Competitors

Good Strategy Bad Strategy and Playing to Win are the two strategy books I recommend to product leaders, and they work best read together.

Rumelt’s kernel is the diagnostic—it tells you whether the strategy in front of you is real strategy or dressed-up aspiration. Lafley and Martin’s cascade is the choice scaffold—it gives you five coherent questions to answer once you’ve diagnosed the challenge. Used together: Rumelt forces you to name the challenge; Playing to Win forces you to make the coherent choices that address it.

The mapping is clean. Rumelt’s diagnosis corresponds to everything upstream of the Playing to Win cascade—the honest assessment of where you are and what obstacle is blocking you. Playing to Win’s Winning Aspiration and Where to Play answers are the guiding policy. How to Win, Capabilities, and Management Systems are the coherent action. The two books together produce a complete strategic artefact: a diagnosis, a policy, and a set of mutually-reinforcing choices ready to be turned into a business case/PID .

A product leader who uses only one of these will have a gap. Use only Rumelt and you risk good diagnosis paired with incoherent choices. Use only Playing to Win and you risk beautifully coherent choices attacking the wrong problem. Use both and you get strategy.

How the Kernel Meets the Roadmap

The diagnosis sits above the OKRs . The guiding policy sits above the objective tagging framework you’ve chosen. The coherent action is what the squad-sprint grid in RoadmapOne actually executes. When every squad-sprint WIP is tagged against the kernel’s elements, a simple analytic question becomes answerable: what percentage of our capacity is engaging the diagnosed challenge?

When the answer is 80%, the kernel is doing its job—the organisation has focused its capacity on the problem it chose to solve. When the answer is 15%, the kernel has diagnosed something the organisation is not actually allocating capacity to. That gap is the honest conversation that should drive the next planning cycle. The tagging frameworks in the objective tagging directory are the operational bridge between the kernel’s abstractions and the grid’s concretes. Once you have a diagnosis and a guiding policy, pick the lens that surfaces the right kind of coherence: RGT if the challenge is maintenance crowding out progress; Three Horizons if the challenge is short-term pressure starving long-term bets; BRICE-style prioritisation if the challenge is the backlog failing to reflect the guiding policy.

FAQ

What makes a bad strategy?

A bad strategy fails one or more of Rumelt’s four tests. It is built on fluff (superficial abstractions and buzzwords that say nothing specific); it exhibits failure to face the challenge (no clear statement of the specific obstacle blocking progress); it mistakes goals for strategy (presenting ambitions as if they were a plan); and it contains bad strategic objectives (objectives that are impossible, incoherent with each other, or disconnected from the diagnosis). Most corporate strategy documents exhibit at least two of these.

What is the kernel of good strategy according to Rumelt?

The kernel of good strategy has three parts: (1) diagnosis — a clear-eyed statement of the specific challenge the organisation faces; (2) guiding policy — the overall approach chosen for addressing that challenge; and (3) coherent action — a set of concrete, mutually-reinforcing steps that execute the policy. All three parts are required. Any strategy missing the diagnosis is almost certainly what Rumelt calls “bad strategy”—goals dressed up as strategy.

What are the four hallmarks of bad strategy?

Rumelt identifies four hallmarks: (1) fluff — superficial abstractions and buzzwords; (2) failure to face the challenge — no clear statement of the specific obstacle; (3) mistaking goals for strategy — presenting a list of ambitions as if it were a plan; (4) bad strategic objectives — objectives that are impossible, incoherent, or disconnected from the diagnosis. Most corporate strategy documents exhibit at least two of these.

How is Good Strategy Bad Strategy different from Playing to Win?

Rumelt’s kernel is a diagnostic framework: it tells you whether a strategy is real or is goals-masquerading-as-strategy. Playing to Win is a choice-making cascade: it gives you five coherent questions to answer. Used together, Rumelt forces you to name the specific challenge; Playing to Win forces you to make the coherent choices that address it. Most product leaders benefit from reading both.

Does the framework still apply in the AI era?

Yes, and the relative weight of the three parts has shifted. Build cost has collapsed, which means the “coherent action” half of the kernel is now relatively cheap to execute. What has become scarce is diagnosis — the specific, named, bounded challenge. When anyone can build anything, competitive advantage collapses into which problem you chose to solve for whom. An extra hour of diagnosis now produces exponentially more value than an extra hour of planning.

What’s an example of bad strategy in a product organisation?

The most common example is a strategy document structured as: vision → values → five-year goals → OKR tree → initiative list. No diagnosis. No specific challenge named. No guiding policy that commits to one approach over others. The document reads like strategy but is, by Rumelt’s test, an aspirational artefact. A product leader presenting such a document to a board should expect the question: what is the specific challenge this strategy engages? If the answer is a generality, the strategy is not real yet.

How do I apply Rumelt’s kernel to my current product strategy?

Start by trying to write the diagnosis in one paragraph, naming (a) the specific customer segment, (b) the specific competitive reality, and (c) the specific obstacle blocking progress. If you cannot, the diagnosis isn’t sharp enough yet—work on it before anything else. Once the diagnosis is sharp, write one paragraph for the guiding policy committing to an approach. Then list the top three coherent actions and explain how action A makes action B work better. Review against your current roadmap allocation — is your capacity actually serving the policy?

Conclusion

Rumelt’s Good Strategy Bad Strategy is the sharpest fluff detector I know. The kernel—diagnosis, guiding policy, coherent action—is three parts that every real strategy contains and that fluff-strategy systematically omits. Product leaders who internalise the kernel stop confusing goals with strategy, stop producing beautifully formatted artefacts that don’t survive market contact, and start having the uncomfortable conversations that produce actual direction.

In 2026, one part of the kernel has become dramatically more valuable than the other two. The collapse of build cost has turned diagnosis into the scarce input. A sharp diagnosis now matters more than it has ever mattered, because every subsequent coherent action rests on it and the market now clones execution in weeks. Spend your strategy hours where the leverage is: name the challenge, sharpen it, and make the rest of the organisation face it honestly. The business case that follows, the cascade of strategic choices that structures it, and the squad-sprint allocation that executes it will all be better for it.

Or keep writing vision statements. That’s a strategy too. Just not a good one.

Sign up (free) to make your diagnosis visible in your roadmap