The Minto Pyramid Principle
Lead with the answer, support with grouped reasoning, communicate in pyramids not narratives
What is the Minto Pyramid Principle?
The Minto Pyramid Principle is a structure for writing and presentations, developed by Barbara Minto at McKinsey, in which ideas are arranged in three tiers: (1) a single governing answer at the top, (2) a small set of grouped, mutually exclusive arguments beneath it, and (3) the evidence supporting each of those. The practical rule that follows from the shape is to lead with the answer rather than build up to it, so the reader receives the conclusion first and reads the rest as support for a point they already hold. In practice a memo might open with the answer — "we should close the Denver warehouse this quarter" — then give three grouped reasons beneath it (the lease renews in March, volume has moved to Kansas City, the same staff can run both), with the shipping and cost figures for each sitting one level below that. Minto paired the shape with SCQA — Situation, Complication, Question, Answer — as the way to open. The approach is widely adopted in consulting, law, and business writing. Direct evidence for its superiority over other structures is limited — it is established professional practice rather than an experimentally validated method.
Most people structure communication the way they thought through the problem: chronologically, building context before landing the point. Minto inverts this. The Pyramid Principle says your audience needs the answer first, not last — then the grouped logic that justifies it. The name is literal. Barbara Minto developed the method at McKinsey, where she worked on the firm’s writing, and set it out in her book "The Pyramid Principle": every document is a pyramid with one governing thought at the apex, resting on a handful of grouped supporting arguments, each resting on its own evidence. Two rules keep the shape honest — ideas at any level must summarise the ideas beneath them, and ideas in a group must be mutually exclusive and collectively exhaustive (the MECE test that came out of the same tradition). Read top-down it delivers the answer immediately; read bottom-up it is the audit trail for how the answer was reached. It is widely used in management consulting, business writing, and structured problem-solving. Below are the core practices, each with the cognitive mechanism behind it.
Practices
- Lead with the answer, not the build-up
State your conclusion or recommendation first, then provide the supporting evidence.
- Frame problems with SCQA (Situation–Complication–Question–Answer)
Use the SCQA framework to establish context cleanly before stating your recommendation.
- Group supporting ideas into mutually exclusive, collectively exhaustive buckets
Organize your three to five supporting points so they don’t overlap and together cover the argument completely (MECE).
- Use parallel structure across every level of the pyramid
Ideas at the same level of the hierarchy should be grammatically and logically parallel.
- Apply the "so what?" test to every supporting point
Each supporting point should directly and uniquely answer the question above it in the pyramid.
- Write to the question your reader has in mind, not the question you found interesting
Start from the reader’s situation and their pressing question — not from your analysis and its conclusions.
How these practices connect
All six practices manage the same limited resource: the reader's working memory. Front-loading the answer, sequencing context through SCQA, keeping supporting points MECE, making parallel items grammatically parallel, applying the "so what?" test, and starting from the reader's question are each a way of reducing how much the reader has to hold in suspension before the structure resolves. When any one of them is skipped, the reader is forced to carry unresolved context, ungrouped points, or a mismatched question while reading, which is where comprehension breaks down.
Do I need to use SCQA if I'm already leading with the answer?
Yes, they solve different problems and work together rather than compete. Leading with the answer (lead-with-the-answer) determines what comes first in the document; SCQA (scqa-situation-complication) determines how you build the shared context that comes after it, so the Answer at the end of SCQA is the same answer you already stated up front. Skipping SCQA after leading with the answer often means the supporting Situation and Complication get delivered as an unstructured dump rather than a sequence that makes the complication feel relevant.
What's the difference between MECE and parallel structure, aren't they the same thing?
No — MECE (group-ideas-into-mutually-exclusive-categories) is about content: whether your supporting points overlap or leave gaps. Parallel structure (use-parallel-structure-across-branches) is about form: whether items at the same level are phrased the same grammatical way. You can have three MECE points written in three different grammatical forms, which still forces the reader to work out that they're peers; and you can have three grammatically parallel points that actually aren't logically equivalent, which creates false symmetry. Both are needed because MECE without parallel form hides the structure, and parallel form without MECE fakes a structure that isn't there.
How do I know if a supporting point should be cut?
Apply the "so what?" test (apply-the-so-what-test): ask whether the point above it in the pyramid would actually fail to hold without it. A point that is true and interesting but not load-bearing should be cut even though nothing in it is wrong. This test also interacts with MECE grouping (group-ideas-into-mutually-exclusive-categories) — a point that duplicates another bucket's coverage will usually fail the "so what?" test too, because the parent claim doesn't actually depend on it once the other bucket already covers that ground.
Where does SCQA actually start, with my analysis or with the reader?
It starts with the reader. Writing for the reader's question in mind (write-for-the-reader-s-question-in-mind) determines what the Situation and Complication in SCQA (scqa-situation-complication) need to be: if the Situation you write doesn't match the reader's actual mental model, the Complication built on top of it lands on the wrong foundation and the structure collapses before reaching the Answer. In practice this means you identify the reader's pressing question before drafting the Situation sentence, not after.
Is leading with the answer just the same idea as writing for the reader's question?
They're closely related but not identical. Writing for the reader's question (write-for-the-reader-s-question-in-mind) is about choosing which question you're answering in the first place — the reader's, not the one your analysis happened to produce. Leading with the answer (lead-with-the-answer) is about where that answer goes once you've chosen the right question. You can lead with an answer to the wrong question and still fail the reader, which is why the reader's-question practice has to come first even though it's less visible on the page.
Do all these practices have the same amount of evidence behind them?
No. Leading with the answer, MECE grouping, parallel structure, and writing for the reader's question are described as mechanistic — grounded in how working memory and structural cues function during reading. SCQA and the "so what?" test are described as anecdotal, meaning their support comes from practitioner experience rather than controlled study. That doesn't make them less useful, but it means the confidence you place in "this will work" should be lower for SCQA and "so what?" than for the other four.
How do you choose between these?
Which practice to reach for depends on where the specific piece of writing is breaking down, not on who is writing it. If the reader would have to wait too long to find out your point, the fix is leading with the answer. If the reader doesn't yet share your context and would reject the point without it, SCQA gives you a sequence to earn that shared ground first. If you already have several supporting points and they feel repetitive or incomplete, that's a MECE problem. If the points are sound but hard to scan as a set, that's a parallel-structure problem. If a point feels like padding, run the "so what?" test on it. And if the whole piece feels technically correct but somehow beside the point, back up further than any of these — you may be answering a question the reader never had, which only writing for the reader's question in mind fixes.
What these approaches do not settle
SCQA assumes the Situation is genuinely shared knowledge; when the audience disputes that framing, the structure loses them before the Answer arrives, and no amount of tightening the Complication fixes that. The "so what?" test is described as harder to apply to inductive arguments — where points together suggest a conclusion rather than necessitate it — which covers most real business writing, so the test is a discipline to apply with judgment rather than a mechanical pass/fail. And lead-with-the-answer, SCQA, and write-for-the-reader's-question all explicitly carry the same caveat: they are defaults for analytical, informational communication, not universal rules — persuasive, narrative, or hostile-audience writing can legitimately withhold the answer and build context first.