The Pyramid Principle is a rule for structuring writing: state the conclusion first, then the two or three reasons that support it, then the evidence underneath each reason. It comes from Barbara Minto's work as a consultant at McKinsey & Company, and it governs how consultants write recommendations, build slide decks, and structure client reports. The supporting reasons have to be MECE, mutually exclusive and collectively exhaustive: they don't overlap, and together they cover the whole argument. Before any of that gets written, a short framing exercise called SCQ (Situation, Complication, Question) tests whether there's actually an argument to make. And before slides get built, consultants storyboard the whole sequence with action titles and no data, a discipline called ghost-deck storyboarding, so the story reads correctly from the titles alone.
The Pyramid Principle is Barbara Minto's rule for structuring writing and presentations: the conclusion is stated first, followed by the two or three arguments that support it, followed by the evidence beneath each argument. It inverts the order in which analysis is performed, where facts are gathered, the analysis is worked through, and the conclusion arrives at the end. A client, a board member, or an examiner reading a deliverable retraces the author's reasoning only by choice; the answer-first order supplies the conclusion, then the structure supporting it, then the evidence for a reader who goes looking.
Minto developed the method while working as a consultant at McKinsey & Company and later published it as a standalone framework; it has since become close to a default in management consulting. This guide covers the method as a consultant applies it to a client deliverable: framing the question before writing (SCQ), building the answer-first structure (the pyramid and MECE), writing titles that carry the argument on their own (action titles), and storyboarding the sequence before any chart is built (ghost-deck storyboarding).
The structure of the Pyramid Principle
The rule states that the conclusion comes first, then the structure that supports it, then the evidence underneath each piece of that structure. Minto's term for the top line is the governing thought, the single sentence that answers the reader's question.
The pyramid shape describes the relationship, not a slide layout. The governing thought sits at the top. Underneath it sit a small number of supporting arguments, usually three, each one answering "why" or "how" of the governing thought. Underneath each of those sits the evidence that proves it. If a supporting argument needs more support than a single layer of evidence provides, it gets broken into its own smaller pyramid one level down.
The method governs order rather than content. A report built data-first, analysis-second, conclusion-last contains the same facts as a report built answer-first, and the two present differently to the reader. The first asks the reader to follow a chain of reasoning before learning where it ends. The second states the endpoint, then allows the reader to verify the reasoning at whatever depth they choose.
SCQ framing: Situation, Complication, Question
Before the pyramid is built, a three-step framing exercise tests whether the material supports an argument rather than only a topic.
- Situation. The stable context the reader already agrees with. No argument yet, just the facts everyone accepts.
- Complication. The change or tension that makes the stable picture unstable, and the element that establishes why the deliverable exists.
- Question. The question the complication forces. If deriving the question from the complication takes real effort, the complication is not sharp enough yet.
Worked example, from a mid-size bank client's BSA/AML gap analysis:
| SCQ element | Worked statement |
|---|---|
| Situation | The bank's BSA/AML program passed its last two examinations with no material findings. |
| Complication | A newly expanded product line moved a meaningful share of customers into higher-risk categories the existing transaction-monitoring rules were never tuned for, and the next examination is five months out. |
| Question | Which of the assessment's findings need to be fixed before the next exam, and in what order? |
The governing thought answers that question. Where the situation and complication do not produce a question, the deliverable is not ready to write and remains a collection of data without a claim to support. The Terms of Reference stage of an engagement is usually where the question first gets named; see the guide to writing Terms of Reference for how to pin it down before work begins.
The pyramid: governing thought and MECE buckets
The governing thought answers the SCQ question directly, in one sentence, stated as the conclusion rather than a preview of it.
Governing thought: "Fix transaction-monitoring tuning and the CDD refresh backlog before the exam; the assessment's other findings can follow the standard remediation cycle."
Under that governing thought sit the supporting arguments, and this is where MECE, Mutually Exclusive and Collectively Exhaustive, does its work. Each bucket has to avoid overlapping with the others, and together the buckets have to cover the whole argument. A bucket that overlaps another double-counts evidence and muddies which reason actually carries the recommendation. A gap between buckets is where the reader's obvious follow-up question goes unanswered.
Worked MECE check against the governing thought above:
- A. Exposure. Transaction-monitoring tuning and the CDD backlog account for the highest dollar exposure among the assessment's fourteen findings.
- B. Timing. The exam window closes before a standard remediation cycle would clear the CDD backlog on its own schedule.
- C. Containment. The remaining lower-priority findings already have compensating controls in place, which is why they can wait.
The buckets are checked against both MECE requirements. Mutually exclusive: exposure, timing, and containment are three different dimensions of the same finding set, and none restates another. Collectively exhaustive: together they answer the three questions a board needs answered to accept a prioritization decision, namely how large the risk is, why the timing requires action now, and why deferring the remainder is safe. A bucket restating "these findings are important" would fail the exclusivity test by duplicating bucket A. Omitting a "why the rest can wait" bucket would leave a predictable board question unanswered.
Action titles
Once the pyramid is built, every slide or section title states that layer's piece of the argument in a full sentence rather than labeling the topic underneath it. Minto's rule, adopted throughout consulting practice, is that a title carries the conclusion rather than the subject.
| Version | Title |
|---|---|
| Topic label (weak) | Transaction Monitoring Findings |
| Action title (does the job) | Transaction-monitoring tuning carries the largest dollar exposure of any open finding, and the shortest fix window |
Each title carries one message. A title requiring "and" to hold two claims that do not support each other indicates two slides combined into one.
The title-only read test catches storyline problems before build time is spent: every title in the deck or report is read top to bottom, with all body content ignored. If the titles alone tell a coherent, logically supported story, the structure passes. If they do not, the defect sits in the pyramid rather than in the individual slides, and the pyramid is where it is corrected.
Ghost-deck storyboarding
A recurring failure in consulting deliverables is proceeding directly to the data and allowing what the data shows to dictate the story. Ghost-deck storyboarding is the discipline that prevents it: the full slide or section sequence is built, with real action titles and a named exhibit type for each, before any chart is built or any number is pulled.
The sequence runs: outline the pyramid (governing thought and MECE buckets), write the action title for every slide or section in order, tag each with the type of exhibit it requires without building it, run the title-only read test, and only then pull data and build the real exhibits.
Worked ghost storyboard for the bank gap-analysis example:
| # | Action title | Exhibit needed | Supports |
|---|---|---|---|
| 1 | Fix transaction-monitoring tuning and the CDD backlog before the exam; the rest follows the standard cycle | Title / summary | Governing thought |
| 2 | The assessment's fourteen findings sort into three groups: highest exposure, timing-constrained, and already contained | Comparison grouping | Frame |
| 3 | Transaction-monitoring tuning carries the largest dollar exposure of any open finding | Bar chart | Bucket A |
| 4 | The exam window closes before a standard remediation cycle would clear the CDD backlog | Timeline | Bucket B |
| 5 | The remaining findings already have compensating controls in place | Comparison table | Bucket C |
| 6 | Recommended 90-day fix sequence, by owner | Process sequence | Recommendation |
Naming the exhibit type at this stage, without building it, is what makes the storyboard cheap to fix. Swapping a bar chart for a comparison table costs one line in an outline. Finding the same problem after the chart is built costs the build. For the fuller decision table on which exhibit type carries which kind of message, a matrix versus a process flow versus a true hierarchy, see the companion guide on choosing consulting diagram types.
Where this fails a client, a board, or an examiner
- No governing thought. The deliverable opens with background instead of the answer, and the reader has to build the conclusion themselves.
- Buckets that overlap. Two "reasons" that are really the same reason stated twice, which makes the recommendation look better-supported than it is.
- Buckets with a gap. The reader's obvious next question, why is it safe to wait on the rest, isn't covered anywhere in the structure.
- Topic-label titles. A deck of "Findings," "Analysis," "Recommendations" tells the reader nothing until they read the body of every slide.
- Data-first building. Charts get built before the story is locked, and the deliverable ends up organized around whatever the data made easy to show rather than what the reader needs to decide.
These failure modes tend to surface at the moment a consultant is fitting an engagement's findings into a Statement of Work's committed deliverables; see the guide to writing a compliance consulting Statement of Work for how deliverable structure gets committed before the pyramid is even built, and the worked SOW examples for what a real deliverables section looks like in practice.
The integrity guardrail: selection for clarity, never concealment
Every technique in this guide is a selection discipline. A governing thought selects one sentence from everything the author could state. MECE buckets select three reasons from everything that is true. Ghost-deck storyboarding selects which exhibits are built. Selection is unavoidable, since a deliverable presenting every fact at equal weight would be unreadable.
The distinction that matters is what the selection serves. Ethan Rasiel's description of this discipline, from his account of consulting practice, states that the task is to build a story from some, but not all, of the facts gathered, for the audience at hand. That describes competent writing. It becomes a different matter where "some, not all" operates to withhold the fact that would change the reader's decision.
In practice, the gathered material sorts into three groups: what is load-bearing for the recommendation, what is supportive but could be cut without changing the conclusion, and what is disconfirming or complicating, meaning the facts that cut against the recommendation. The first group forms the deliverable. The second fills gaps or moves to an appendix. The third is addressed in the body or the appendix rather than dropped, since its absence carries its own risk to the deliverable's credibility.
The test applied before a deliverable ships is whether the selection would still read as fair, competent writing if the client independently discovered the fact left out. A selection that passes that test is selection. One that does not is concealment, and it converts a structured recommendation into a liability.
Primary sources
- Barbara Minto, The Pyramid Principle: Logic in Writing and Thinking (FT Prentice Hall / Pearson, revised edition). Developed while Minto was a consultant at McKinsey & Company; the source of the governing-thought, MECE, and answer-first structure this guide is built on.
- Ethan M. Rasiel, The McKinsey Way (McGraw-Hill, 1999). Source of the MECE issue-tree method used in the worked bucket example above.
- Ethan M. Rasiel, The McKinsey Mind (McGraw-Hill, 2001). Source of the "story from some, not all, of the facts" selective-narrative discipline behind the integrity guardrail above.
- Paul N. Friga, The McKinsey Engagement (McGraw-Hill, 2008). Source of the ghost-deck storyboarding sequence.