Strategy review checklist

A ten-pass procedure for reviewing any strategy document. Allow about ninety minutes.

Before you criticise anything

1. Steelman. Write the document's argument in one paragraph as if you believed it. If you cannot, you are about to review your own preferences.
2. Name the decision. What is being decided, by whom, by when? A review with no decision in it produces commentary.

Structure

3. Decompose. For each element, which question does it answer - pricing mechanic, unit economics, sequencing, or channel? Elements that answer none are decoration; two elements answering the same one are a duplicated bet.
4. Wedge sentence. For [people] who [urgent reason], we solve [narrow problem], which accumulates [bridge asset], which makes [expansion] materially easier than selling it cold. Attack the bridge asset specifically.
5. Cap specification. Metered unit, shape, soft threshold, hard threshold, and the bet about what stays free. Then the number nobody wrote down: the conversion rate the plan requires, and the signups that implies.
6. Moat split. Divide accumulated value into re-derivable and captured-only. State the time to lock on the second half alone.

Evidence

7. Label every number: primary / primary-reported / trade aggregation / unsourced. Flag precise unattributed percentages, sets where every figure supports the same conclusion, and sector substitution.
8. Classify every constraint: legal / capability / commercial / imagined. For each legal one, name the regulator and instrument, ask whether it regulates the activity or the artifact, read the regulator's own page, check the date, write the finding with the URL.
9. Stage every competitor on age, funding with date, customers and how stated, distribution, feature recency. Then: validated or taken? Taken needs distribution lock or switching cost at scale.

Write it

10. Five sections, no more:
    • What it gets right. Specific, and first. A review that opens with objections gets read as an attack and argued with rather than used.
    • What does not survive checking. Claim, finding, source, and whether it changed a conclusion in the document.
    • The decision. Yours, stated plainly, with the changes you would make to the proposal.
    • Falsifiers. Checkable conditions, agreed in advance, that would reverse the decision.
    • The next N weeks. Ordered, with the highest-information action first.

Three habits that do most of the work

• Ask what number decides this, then spend your research there. Most research is done on numbers that could not have changed anything.
• Distinguish a wrong conclusion from a wrong method. An author can reach the right answer through an unchecked claim; say so, because the method will be used again on a different question.
• Separate what is unknown from what is unknowable. "We do not know if agents will accept recording" is a week of interviews. "We do not know if the market will exist in 2029" is not a research task. Only the first belongs in a plan.