Rumelt's strategy kernel

Richard Rumelt's Good Strategy / Bad Strategy gives the cleanest test. A good strategy has a kernel of three parts: a diagnosis of what is hard, a guiding policy for how you will deal with it and a set of actions that reinforce each other. Ambition alone is not strategy.

The big idea

1. Diagnosis — what is the critical challenge? Compress reality into the few factors that matter.
2. Guiding policy — the overall approach you will take to overcome that challenge (not a list of goals).
3. Coherent actions — mutually reinforcing steps that implement the policy.

Bad strategy looks like fluff ("best-in-class experiences"), goals without a path ("grow ARR 40%") or a laundry list of unrelated initiatives that try to please every stakeholder.

The guiding policy creates focus. Focus is the scarce resource. Every coherent action should make the others more valuable.

Diagnosis is the step teams skip. They jump to "we need AI" or "we need mobile" without naming the obstacle. A useful diagnosis is specific enough that two smart people could disagree with it — and therefore discuss evidence.

Walking through it

Bad: "Become the leading collaboration platform. Delight users. Expand internationally. Invest in AI. Improve NPS."

Better kernel:

• Diagnosis: SMBs churn because setup takes 3 weeks and they never reach the aha moment; enterprise sales wants features that would slow that onboarding further.
• Guiding policy: Optimize for time-to-value for teams of 5–50; refuse custom enterprise workflows until activation for that segment is above 40%.
• Actions: Rewrite onboarding around one template; kill three low-usage enterprise integrations; hire one solutions engineer for mid-market only; instrument activation events and review weekly.

Better onboarding data makes the solutions engineer more effective; refusing custom work frees engineering for activation. The actions reinforce the policy.

Sources of diagnosis: churn and expansion interviews; activation funnels; win/loss ("what alternative won?"); support and sales-call themes; technical constraints that block a segment.

Write the diagnosis as a short paragraph, not a SWOT dump. "We lose mid-market deals because security review takes eight weeks and we have no SOC2 narrative" is actionable. "Competitive landscape is intensifying" is not.

Rumelt's hallmarks of bad strategy: fluff (buzzwords restating the obvious); failure to face the challenge; goals mistaken for strategy; a long list of unrelated "priorities" that cannot all be true at once. If a leadership offsite produced twelve priorities and zero refusals, you have a portfolio wish list.

Common mistakes

• Skipping diagnosis. "We need AI" is a tactic posing as a challenge.
• Actions that do not reinforce each other. Five initiatives that please five stakeholders are a portfolio, not a kernel.
• Refusing nothing. Strategy requires saying no to good ideas that do not serve the policy.

Check yourself

1. What is wrong with a strategy whose only line is "grow ARR 40%"?

It is a goal with no diagnosis and no policy for how. Rumelt would call it goals mistaken for strategy.

2. Why does "refuse custom enterprise workflows until activation is above 40%" belong in a kernel?

It is a guiding policy: a choice that concentrates force and makes later actions coherent.

Key takeaways

• Kernel: diagnosis, guiding policy, coherent actions.
• Diagnosis must be specific enough to disagree with.
• Strategy requires refusals; twelve priorities is a wish list.

Further learning

Rumelt's own talk is worth an hour if you have ever sat through a "strategy offsite" that produced slogans:

Next on the map: product strategy.