Good strategy has three parts, and most strategy documents have one

By Meet Patel · 2026-10-03 · 6 min read

Here is a sentence from a strategy document I have invented for this post: “Our strategy is to become the leading platform for small businesses in the Gulf by 2028, delight our customers, invest in AI and expand into three new markets.” Every clause is reasonable and every clause would fit on the slide of a thousand companies. Richard Rumelt has a name for what the sentence is: a list of goals. His argument in Good Strategy/Bad Strategy (2011) is that a list of goals and a strategy are different objects, and that a real strategy has a three-part structure he calls the kernel.

This post sets out the kernel, adds Michael Porter's test for whether the middle part rules anything out, and ends with five questions you can run against your own strategy document in an afternoon.

The kernel: diagnosis, guiding policy, coherent actions

In Sachin Rekhi's summary of the book, the three parts read as follows. The diagnosis explains the nature of the most pressing challenge the organization is facing. The guiding policy specifies the approach to dealing with the obstacles called out in the diagnosis. The coherent actions are feasible, coordinated policies, resource commitments and actions designed to carry out the guiding policy.

The order matters. Actions have no direction without a policy, and a policy has nothing to answer without a diagnosis. Rekhi's summary gives two cases from the book. At Digital Equipment Corporation, leadership produced a vague consensus statement rather than making hard choices, and failed to address its crisis. At 7-Eleven Japan, the company diagnosed that the global model was not working locally, set a guiding policy around product variety and newness, and carried it out through local insight collection and partnerships with manufacturers. The second case has all three parts and each part follows from the one before.

Rumelt also lists four hallmarks of bad strategy, which Leading Blog summarizes: fluff, failure to face the challenge, mistaking goals for strategy, and bad strategic objectives. The invented sentence above exhibits at least three of them.

Diagnosis: name the obstacle

A diagnosis states what is in the way. “We want to lead the market” describes a destination, and a destination does not explain why you have not arrived. A symptom makes a poor diagnosis as well: “revenue is below plan” reports an outcome and names no cause, so every proposed fix looks equally plausible.

A usable diagnosis contains a cause and can be wrong: “customers with fewer than 25 employees leave within a year because setup takes six weeks and they have nobody to run it” can be checked against churn data and exit interviews. If the diagnosis cannot be checked, the rest of the document rests on nothing.

Guiding policy: the part that rules things out

The guiding policy is where most documents fail, and Porter explains why in “What Is Strategy?” (Harvard Business Review, November-December 1996). He writes that “strategy is making trade-offs in competing” and that “the essence of strategy is choosing what not to do.” A trade-off, in his definition, means that more of one thing necessitates less of another.

His illustration is Southwest Airlines. According to the article, Southwest offered short-haul, low-cost, point-to-point service between midsize cities and secondary airports, with gate turnarounds of only 15 minutes. It eliminated meals, assigned seats and interline baggage checking, and it flew a standardized Boeing 737 fleet. Each of those choices excludes something customers of other airlines expect, which is why full-service carriers could not copy Southwest without abandoning their own positioning.

This gives you a practical test for any guiding policy: would a sensible competitor choose the opposite? “Deliver an excellent customer experience” fails, because nobody picks the opposite. “Sell only to companies of 5 to 25 employees and decline custom integrations” passes, because a rival could reasonably serve large accounts and build bespoke work.

Coherent actions: fit among activities

Porter defines strategy as “creating fit among a company's activities.” Applied to the kernel, the third part asks whether the actions reinforce each other and the policy. Four initiatives that each have a good business case but would work equally well under a different policy are a project list. Four initiatives that each depend on the policy, and make each other cheaper or more effective, are coherent actions. The test is to remove one action and see whether the others get weaker.

A reasonable reading of the Southwest list is that the quick turnaround and the single aircraft type support each other: one set of crew procedures, one set of spare parts. Porter argues that fit across many activities makes an advantage harder to copy than any single activity would, because a competitor finds it harder to match an array of interlocked activities than to replicate one.

Five questions to run on your strategy document

1. Find the diagnosis. Can you underline one sentence that names an obstacle and a cause? If the document opens with a vision and never mentions what is in the way, it fails here.
2. Check that it can be wrong. Name the data that would show the diagnosis is mistaken. If none exists, the diagnosis is an opinion.
3. Try the opposite. State the reverse of the guiding policy. If a reasonable company might choose it, the policy rules something out. If nobody would, it is a value statement.
4. List what you will stop. Porter's trade-off means a policy has costs. Write the specific requests, segments or features you will decline because of it.
5. Delete each action in turn. Does the policy weaken, and do other actions lose their purpose? An action that changes nothing when removed was never part of the strategy.

Run the test with two people who did not write the document and give each ten minutes. If they disagree about which sentence is the diagnosis, that disagreement is the finding.

A worked example, with invented figures

Take a hypothetical 40-person company that sells payroll software. Its first draft says: “Become the leading payroll platform for small businesses by 2028; delight customers; invest in AI; enter three markets.” It fails questions one, three and four.

A rewritten version, run through the kernel:

• Diagnosis: customers with fewer than 25 employees leave in their first year because implementation takes six weeks and they have no administrator to run it.
• Guiding policy: sell only to companies with 5 to 25 employees in one country, with a product that is live in a week, and decline custom integrations.
• Coherent actions: templated onboarding that completes payroll setup in five working days; one flat per-employee price with no implementation fee; support staffed in the customer's time zone; one bank integration built well, with no others until the first is stable.

Run the tests. The policy has an opposite that a rival might choose (serve enterprises with custom work). It names what will be declined. Remove the flat price and the onboarding template loses its commercial logic, because a fee would restore the friction the diagnosis identified. The actions lean on each other. Whether the diagnosis is correct remains an empirical question, which is the reason question two exists.

Where this fits with other strategy work

I have written elsewhere that constraint is strategy, and the kernel gives that idea a structure: the guiding policy is the constraint, written down. Pricing is often one of the coherent actions, and your pricing is a product decision that should be checked against the policy in the same way as everything else.

A strategy document earns its name when someone can use it to refuse a request. Read yours with that in mind. If a senior colleague proposes a new initiative next week, the document should tell you in a minute whether it fits, and the answer should sometimes be no.

Written by Meet Patel — startup operator and growth strategist in Dubai.

Read on themeetpatel.com