slug: governing-thought view: skill audience: llm layer: block agent: architect companion: corpus/storymakers/frameworks/block/governing-thought.md

Governing Thought — operational reference

One sentence. The single declarative sentence at the apex of a Minto pyramid that the entire document exists to defend. One verb that commits, one because, one defensible claim — or it isn't a governing thought.

Decision tree

Q1. Does the document make a recommendation, decision, or argue a position?
    NO  → exploratory deck (workshop, hypothesis tree, regulatory enumeration).
          Don't force a governing thought. Use topic titles.
    YES → Q2

Q2. Can you state the answer in one declarative sentence with a working verb?
    Test: *[VERB] [OBJECT] because [REASON]*. "Acquire Target B for €340M because…"
    NO  → analysis isn't done. Finish the work. Don't fake an apex.
    YES → Q3

Q3. Can you state the question that sentence answers in one line?
    Test: SCQA's Q. "Which of the three targets should we pursue, and on what terms?"
    NO  → write the SCQA opening first; the Q forces the A.
    YES → Q4

Q4. Do your 3–5 supports, taken together, force the apex sentence to be true?
    Test: read only the supports out loud. Do they sound like *because*-clauses
    for the apex? Or could the same supports defend a different sentence?
    NO  → either the apex overclaims (rewrite it) or the supports underdeliver
          (rebuild them).
    YES → ship it.

Recipe (45-minute apex authoring)

Write the question the deck answers. One sentence. Example: "Should we acquire Target B, and on what terms?" (5 min)
Write the apex draft. Form: [VERB] [OBJECT] because [REASON]. Don't polish yet — just commit. (10 min)
Strip qualifiers. Remove "we believe", "may", "could", "is likely to", "subject to", "potentially", "appears to". Each one is a hedge. (5 min)
Force a verb. If the sentence is a noun phrase ("Recommendation: acquisition of B"), rewrite as a verb sentence ("Acquire Target B"). (5 min)
Bound the claim. Add the what, when, how much, why-now — whatever the audience needs to act. "Acquire Target B" → "Acquire Target B for €340M in Q3 before Schaeffler closes". (10 min)
Test against supports. Write each support as one sentence. Read them in order. Do they force the apex? If not, fix one side. (10 min)

Anti-patterns to refuse

Topic titles — "Considerations for European expansion", "Strategic options for growth", "Path forward on pricing". Reject. Force a verb.
Multi-sentence apex — "We evaluated three targets. Target B emerged as the strongest. We recommend further diligence." Two-plus sentences = couldn't decide which was the answer. Collapse to one.
Qualifier-laden apex — "We believe Target B may emerge as the most likely candidate, subject to further analysis". Every hedge marks a place the analysis isn't done.
Noun-phrase apex — "Acquisition of Target B." That's a label, not a claim. Use a verb that does work.
Orphaned apex — answer floating with no SCQA. "Acquire Target B for €340M" with no setup leaves the reader asking why is this the question?
Fake apex — "Acquire Target B for €340M to capture the European service-network gap" — but the supports only cover financial returns, never strategic fit. Confidence the supports don't earn.
Process apex — "We followed a three-phase methodology to evaluate the targets." Methodology is not an answer. Skip to what the methodology found.
Verb-of-being apex — "Target B is the right acquisition." Technically a sentence; structurally a topic. "Is" is the verb of analysts who haven't decided what verb to use.

When NOT to use it

Workshop discovery decks — the deck is a discussion artefact; forcing an apex pre-empts the conversation the workshop exists to have.
Exploratory hypothesis trees — the document's job is to enumerate possibilities, not commit to one.
Regulatory disclosure — 10-K risk factors, prospectus sections; the writer must list without recommending.
Discovery / problem-definition decks — no answer yet; an apex would be invented.
Status updates with no recommendation — "here's what changed this week" doesn't have a because.
Educational content where revealing the answer too early kills the lesson — case-method teaching, narrative reveal.

Output checklist

A governing thought is shippable when:

One sentence. Not two. Not a paragraph.
Has a working verb (not "to be" pretending to mean something).
Names the what, the because, and (where decision-relevant) the when and how much.
No qualifiers ("we believe", "may", "could", "subject to") unless the analysis genuinely warrants them.
The audience's question (the Q in SCQA) is articulable in one line.
The 3–5 supports beneath it, read alone, force the apex sentence.
A senior reader reading only the apex would know what the deck is asking them to do.

Canonical signals (for matchers)

Keywords (lowercase, verified non-zero corpus presence):

key takeaway, key message, key insight, bottom line, main point
we recommend, we believe, our view, the answer is
in summary, in conclusion, so what

Component kinds: title (action-title slide where the title is the apex), callout (apex pulled out as a coloured box), subtitle (apex sometimes lives one tier below the deck title).

Slide-level shape: A title slide where the title is a complete declarative sentence with a working verb, often followed by 3–5 supporting sub-bullets or a callout block restating the apex. Look for a single sentence (not a noun phrase) at the top of the page that the rest of the page proves.

References

Minto, The Pyramid Principle, 1981 (first ed.) / 1985 (rev.) / 1996 (expanded) — the canonical text; Chapter 2 introduces the governing thought.
Minto — McKinsey alumni interview ("MECE: I invented it…") — primary first-person account.
McKinsey internal style guide — operationalises the apex-first rule.
Zelazny, Say It with Charts, 1985 — every chart defends one governing thought.