The pyramid principle: how to structure a message so the answer comes first

The pyramid principle, also called the Minto pyramid, is Barbara Minto's method for structuring a message: state the answer first, support it with a few grouped arguments and back each argument with facts. Every line summarizes the lines below it, and the ideas in each group are of one kind and in a logical order.

Is the pyramid principle McKinsey's method?#

Barbara Minto, a former McKinsey consultant, developed it and set it out in her book The Pyramid Principle. Her website says major consulting firms use her textbook, so it is not one firm's private method. It is behind what people call consulting-style decks: the answer first, titles that make a point, evidence underneath.

On her own website she describes the later textbook, The Minto Pyramid Principle, as superseding the book.

A pyramid has three levels: one answer on top, about three supporting points below it, and the facts that prove each point at the bottom. Every box is a full sentence or a number, never a topic label.

Here is a fictional case. The COO of a software company asks whether customer support needs more agents. The head of support has a ticket export and 30 minutes. Her pyramid looks like this:

• Answer: Fix onboarding before hiring more agents.
  • Point 1: Tickets grew by 2,000 a month in two quarters, a rise of 31%.
    • 6,400 a month in Q1, 8,400 in Q3.
  • Point 2: New customers cause three quarters of the growth.
    • Customers in their first 60 days: from 900 to 2,400 tickets a month (+1,500).
    • Established customers: from 5,500 to 6,000 (+500).
  • Point 3: Five setup questions make up 60% of the new-customer tickets.
    • 1,440 of the 2,400 tickets.

Read it from the top and each point answers a question the answer raises: is there a real problem (point 1), where does it come from (point 2), can we fix it (point 3)? Read it from the bottom and each group adds up to the line above it: 1,500 plus 500 is the 2,000 behind point 1, and 1,440 is 60% of 2,400.

1. Every line summarizes the lines below it. A summary says something; a label does not. “Ticket analysis” is a label. “New customers cause three quarters of the growth” is a summary. If you cannot write that sentence, you do not yet know what your facts mean.

2. The ideas in a group are of the same kind. Three reasons, three steps or three findings, not a mix of a problem, a number and a to-do. A mixed group is why a list feels as if it does not hang together.

3. The ideas in a group come in a logical order. The usual orders are time (first, then, finally), structure (by region, product or team) and degree (most important first). Pick one and keep it for the whole group. In the case above the order is diagnostic: how big, where, why.

No. The rule of three that people attribute to McKinsey is a habit, not a law: a group of three is easy to remember and easy to say aloud. A group can hold two points or five. If you have seven or more, you have usually missed a way to group them.

Research on working memory suggests that people hold about four chunks at a time (Cowan, 2001), which is why short groups are easier to follow.

Groups at the same level should also not overlap and should not leave a gap. That test has its own name, MECE.

You can build it from the top or from the bottom. Real work mixes the two: you form a hypothesis, test it against the facts, and change the top when the facts disagree.

Top-down: when you already have an answer#

Start with the question and write your answer as one sentence. Then ask “why?” or “how do we know?” until every point has its support, and check that each level answers the question the level above raises.

Bottom-up: when you start with a pile of facts#

Collect the facts first and group like with like. Write one sentence for each group that says what its items add up to. Those sentences are your supporting points, and what they add up to is the answer on top.

The six steps#

SCQA is the short introduction that sets up the pyramid. It reminds the reader of the situation they already agree on, names the complication that changed it, states the question this raises, and then gives the answer, which is the top of the pyramid. Minto's website calls the first three parts the SCQ framework, a way to find the question in the reader's mind. Keep the introduction to a few sentences, and see SCQA for variants.

Put the answer on slide 2, turn each supporting point into a section, and write every slide title as a full sentence that states its point. The pyramid becomes the deck like this:

The answer slide is your executive summary, and it has to work if it is the only slide someone reads. The titles do the rest. Write them as action titles and read them in order: they should sound like the pyramid read from top to bottom. If you draft only the titles first, you have a ghost deck, and that is the cheapest moment to fix the logic.

Try your own pyramid: pick the occasion, the audience and your slot, and get a first outline of action titles.

The example board presentation works the same way. Slide 2 holds the answer and the request, slides 3 to 7 prove it, and slide 8 returns to the decisions.

• Labels instead of statements. “Pipeline” tells the reader nothing. “Pipeline covers next quarter's target 2.4 times, but 60% sits in two deals” tells them what to think.
• Too many points. Seven or nine items in one group usually mean a grouping you have not found yet.
• Overlap and gaps. If “high prices” and “weak sales” both appear as reasons, they overlap, because high prices may be why sales are weak. If a region is missing, the group is incomplete.
• A long introduction. If the answer arrives after two minutes of background, the pyramid is upside down again. Three or four sentences of SCQA are enough.
• Evidence that does not prove the point. If a chart sits under a title it does not support, either the title or the chart is wrong. Change one of them.

Lead with the answer when the reader wants a decision or a recommendation, which covers most business presentations. Hold it back in three cases:

• The audience will resist the answer. State the question early, then walk through the situation and the complication so the reader reaches the answer with you. The pyramid still organizes your thinking; you only change the order of delivery.
• You do not have an answer yet. A working session needs options and open questions, not a conclusion.
• The format is a story. In a training session or a talk where the journey is the point, a chronological order can serve better. How to structure a presentation compares the options.

AI is good at a first draft of the structure when you give it the question, the audience and the facts. It cannot decide what your answer should be, and a tidy structure around a wrong answer is still wrong.

The plan is the place to fix the pyramid: check that the first page after the title holds the answer, that each card's title is a point rather than a topic, and delete cards that support nothing. Check every figure against your source before you present, as with any draft.