The Minto Pyramid Principle: Logic in Writing, Thinking and Problem Solving

The bottom line

Clear writing puts ideas in a pyramid: the answer to the reader's question first, then grouped support beneath it. Readers take in one idea at a time and hold few at once, so fixing the order of ideas helps more than polishing sentences.

Key takeaways

• Give the summarizing idea before the ideas it summarizes. Readers group what they read anyway, and short-term memory holds only about seven items. Ch. 1 · Why a Pyramid Structure?
  • A husband handed nine grocery items one by one usually returns with only the paper and the grapes. Sorted into three categories (dairy, fruit, vegetables), the same nine items are easy to recall.
  • A speaker lists 15 bearded people seen in 15 minutes in Zurich, then New York, then London. The listener guesses London leads the trend, which is wrong. The real point was that facial hair is now accepted in business.
• Find the reader's question by defining the gap between the result now and the result wanted. The last thing the reader already knows becomes the Complication that raises the Question. Ch. 8 · Defining the Problem
  • A real estate firm sold the same way for 30 years and grew sales about 10% a year. Final quarter projections showed sales down 10%. The question became how to ensure continued growth.
  • Supermarket chains raised a test-market fee to $20,000 a week, then refused testing when the industry would not pay. That refusal was the Complication, and the question was how the industry should respond.
• Structure the analysis as yes-no questions before gathering any data. The questions show in advance which data matter and when the research is done. Ch. 9 · Structuring the Analysis of the Problem
  • Early consulting firms began every job with a full company and industry study. One major firm estimated that 60% of its fact-finding effort was wasted.
  • A Texas distributor held $27 million of central inventory. A tree showed high stock comes from ordering too much or keeping it too long, which cut five wordy "issues" to two yes-no questions.
• State the point each grouping implies instead of a blank line like "five problems". Writing the real summary completes the thinking and exposes gaps. Ch. 7 · Summarizing Grouped Ideas
  • A writer's "two organization problems" turned out to be four areas needing more delegation. The real problem was the inability to delegate authority.
  • A ten-step telecommunications program, sorted into three steps, revealed a missing need for a central manager.
• Answer the reader's question directly and put the reasons on the Key Line. A direct inductive answer is easier to absorb than a chain of therefores. Ch. 3 · How to Build a Pyramid Structure
  • Big Chief offered to cut a five-week billing cycle to one day. The reply memo described the new system but never said whether to accept. Rewritten, it said yes, for three reasons: needed information, better cash flow, less work.
  • A consultant's rambling cost memo became one answer: yes, composing costs can be cut by up to 10 percent, by removing unneeded steps and paying competitive wages.
• Lay out the page so the reader sees the whole argument in 30 seconds. The eye sees the logic of a page before the mind understands it. Ch. 10 · Reflecting the Pyramid on the Page
  • A memo's four slide requests, rewritten as a parallel numbered list, showed the writer never said what slides he wanted for point 4.
  • A contents page reading Introduction, Background, Findings, Conclusions, Recommendations tells the reader nothing about the thinking.

Controlling the sequence in which you present your ideas is the single most important act necessary to clear writing.

Barbara Minto, Why a Pyramid Structure?

Readers hold only a few ideas at once, so give them the summary first

Most unclear writing is a problem of structure, and the cure is to presort ideas into groups and state each group's point before its members.

• Unclear writing usually comes from the order of ideas, not from style, and order is far easier to fix. Ch. 1 · Why a Pyramid Structure?
  • Even a short document of about 2 single-spaced pages holds roughly 100 sentences. The reader must take each one in, relate it to the others and hold them together.
  • Changing style after formal education takes repetitive practice. Few working writers have time for that.
  • Long sentences and abstract or technical language are real problems. They are simply the harder ones to change, much like learning to type.
  • Sorting ideas is also how a writer finds out what they think: We cannot know what we think precisely until we have symbolized it in some way.
• Readers group whatever appears together, whether or not the writer helps them. Ch. 1 · Why a Pyramid Structure?
  • The Greeks looked at the night sky and saw figures, not separate points of light.
  • Six dots with some gaps smaller than others are seen by everyone as two groups of three. Closeness alone drives the grouping.
  • Past four or five items, the mind starts sorting them into categories so it can keep them.
  • The dots come from a series in Gestalt Psychology by Wolfgang Kohler (1970).
  • Shown pairs of unrelated nouns such as lake and sugar or girl and kangaroo, people link each pair with an imagined scene, like sugar dissolving in a lake. Most then recall every partner without hesitation.
  • George A. Miller put the limit at seven items, plus or minus two. Three is convenient and one is easiest.
• Grouping helps only when the logic of each group is stated before its items. Ch. 1 · Why a Pyramid Structure?
  • Nine items split into groups of four, two and three are still nine items. Naming the category lifts the reader one level up.
  • Ideas arrive one at a time, so the efficient order is the category first and then its members.
  • Left to guess, readers guess wrong or not at all. Even a right guess costs them effort.
  • Named categories work because each one suggests the items below it.
  • An article on equal pay for women opened with five points: a wider earnings gap, two meanings of equal pay, forcing employers to act in their own interest, and ending male workers' restrictive practices. The writer thought he had started at the top. Readers could find no link and gave up.
  • A mixed group like apples and chairs can only be summed up as "things", which says nothing.
• Ideas are worked out from the bottom up but delivered from the top down. Ch. 1 · Why a Pyramid Structure?
  • Sentences join a paragraph only if all are needed to defend one idea. Paragraphs form sections, and sections form a document, by the same rule.
  • Repeating this grouping and summarizing always ends in one thought at the top.
  • Each summary sentence turns its paragraph into a single point for the next level.
  • The level directly under the governing thought is called the Key Line.
• Three rules tested before drafting expose unfinished thinking. Ch. 1 · Why a Pyramid Structure?
  • Each idea summarizes the ideas grouped below it.
  • Each group holds one kind of idea, named by one plural noun such as reasons, steps or recommendations.
  • Each group sits in a logical order, so there is a reason the second idea comes second.
  • The order should match how the group was formed: deductive, chronological for cause and effect, structural for a divided whole (Boston, New York, Washington), or comparative for categories.
  • A broken rule points to flawed thinking, undeveloped ideas or unclear relationships. Fixing it before drafting avoids heavy rewriting later.

Every point raises a question that the level below must answer

A pyramid works as a dialogue: each idea provokes a question, and the ideas beneath it answer that question and nothing else.

• Each idea raises a question, and the line below must answer exactly that question. Ch. 2 · The Substructures Within the Pyramid
  • An idea is a statement that tells the reader something new. That newness makes him ask Why?, How? or Why do you say that?
  • The answer sits on the line directly below. It also tells him something new, so it raises the next question.
  • Go down a leg until the reader has no logical question left. Then climb back to the Key Line and take the next point.
  • G. K. Chesterton argued that pigs should be kept as pets for two reasons. They are extremely beautiful, and they could be bred to fascinating variations. Why beautiful? They are marvelously fat and typically English. Fat gives lovely curves to the onlooker and modesty to the owner. Englishness ties pigs to the land, a sign that power need not be unkind.
  • A reader can reject Chesterton's case and still see exactly why he makes it. Each grouping answers only the question raised just above it.
  • The stopping rule is about following, not agreeing. A reader who tracks the reasoning to the end may still disagree. Clear following is the most a writer can hope for.
  • A 20-page memo urging the purchase of a British Leyland franchise gave three reasons: faster growth than the industry, positive financial impact, and easy absorption. Each reason had its own answer beneath it, so the reader could pinpoint where he disagreed and ask logical questions.
• Answer questions only when the reader actually has them. Ch. 2 · The Substructures Within the Pyramid
  • Never raise a question before you are ready to answer it.
  • Never answer a question the reader has not yet asked. A section like that will need repeating or rereading.
  • The first statement must answer a question already in the reader's mind, or one he would have if he thought about his circumstances.
  • A reader looks for what he doesn't know only when he needs to. No need means no question, and the dialogue never gets started.
  • Held to this rule, The pyramid structure almost magically forces you to present information only as the reader needs it.
  • The dialogue also helps the writer. Most writers start with a hazy sense of their ideas, and the first version of any idea is less precise than it can become. Asking what question each point raises pulls the next idea into view.
• Build the top first: Subject, Question and Answer, checked against the story the reader knows. Ch. 3 · How to Build a Pyramid Structure
  • The top is one sentence with a subject and a predicate. Its subject is the document's subject, and it answers a question the reader already holds.
  • Check the Question against the Situation and Complication. Then name the New Question the Answer raises.
  • If the Key Line answers inductively, name the plural noun that covers its points, such as reasons for a Why? question.
  • Exact steps: draw a box and write in the Subject. State the question about it you want answered in the reader's mind by the end. Write the Answer, or note that you will find it. Write the Situation, the first statement he will accept as true. Imagine him saying "so what?" and write the Complication. If it doesn't raise your Question, change the Question or rethink the Complication.
  • Don't begin by drafting the opening paragraph. Use the parts of the introduction to draw the right points out of your head one at a time.
  • The method surfaces only what bears on the reader's question. It also pushes the writer to answer all of that question, not just part of it.
  • John J. Jackson, head of Accounting, wrote to Robert Salmon about a hamburger chain's offer to record its own delivery tickets on a computer disk. Reworked top-down, his memo became clear with almost no change in wording. The order of the ideas did the work, not the style.
• When the top won't come, build up from the details. Ch. 3 · How to Build a Pyramid Structure
  • If the Subject, the Question or what the reader knows stays unclear, drop to the Key Line. If that fails too, start from the bottom.
  • List every point. Judge the recommendations first, since action ideas are easier to assess. Then map cause and effect among the problems.
  • Draw a conclusion. Test it by running through Subject, Question, Answer, Situation, Complication, New Question and Key Line.
  • A young consultant, 2 weeks into his first assignment at a printing company in England, sent his manager a rambling memo. Composing costs were 40 percent of hardback costs and 50-55 percent of paperback costs. TTW ranked 20-50 percent below PAR, paid less than local printers, had lost two compositors, and ran overtime more than 50 percent over budget.
  • Listing his points showed that his two remedies said the same thing. Mapping cause and effect showed two separate lines of reasoning hidden in the facts.
  • Don't dictate everything to sort out later. Ideas once written look carved in gold, and the writer clings to disjointed thinking.
  • Used on someone else's document, bottom-up mapping treats the text as a closed universe. It clarifies what is said but does not judge whether it is true.

Open with a story the reader already knows: Situation, Complication, Question

An introduction works by reminding the reader of a story they accept, so the one Question they hold leads straight to the Answer.

• Start with a Situation the reader already accepts, anchored in a time and place. Ch. 4 · Fine Points of Introductions
  • An unfinished story holds attention until it resolves.
  • The opening statement must be self-sufficient and noncontroversial.
  • Points the reader agrees with first make them more open to points they may dispute.
  • Two Irishmen meeting on a bridge at midnight in a strange city pin the listener to a time and place. The teller then steers attention until the punch line.
  • Real openings anchor the same way. Energoinvest is weighing alumina exports from its Mostar plant to Ziar in Czechoslovakia. Health services, the Irish Health Service included, are under pressure. Stone tools fill the first 2.5 million years of the archeological record.
  • A named reader needs only a reminder of the question. A wide audience needs it planted through accepted facts. One test asks whether the material could have appeared in Business Week or Fortune.
  • The reader's natural response to a Situation is "so what?". That response opens the way for the Complication.
• The Complication is simply what happened next, and it raises exactly one Question. Ch. 2 · The Substructures Within the Pyramid
  • A Complication creates tension. It need not be a problem.
  • Most take one of four forms: a task is blocked, a known solution needs doing, a proposed solution needs testing, or an action failed.
  • Two questions must be related, so fold one into the other.
  • If the Question won't come, ask what question the body's points answer.
  • A board memo opened with a purpose line and five topics. Rewritten as a story, its Situation was a new organization, installed in October, that gave division managers full day-to-day authority. Its Complication was a Board long fixed on short-term problems. Its Answer was to hand daily matters to a broadened Executive Committee.
  • Levitt's Marketing Myopia: industries stop growing, people assume saturated markets, is that right? No, the cause is a failure of management.
  • Hayes and Abernathy contrast the U.S. with Germany. Germany imports 95% of its oil against 50%, yet its productivity growth ran more than four times the U.S. rate.
• Keep the introduction short, free of proof, and complete in every part. Ch. 4 · Fine Points of Introductions
  • Exhibits in an introduction signal overstating the obvious.
  • Situation and Complication run three or four paragraphs at most. A close reader may need one sentence.
  • Length follows the reader's needs, not the document's length.
  • Reordering the parts sets the tone, but all must be present.
  • Introductions are meant to remind rather than to inform.
  • A London Office memo proposed a Firm Development Project. Diversification work had grown 40 percent in 5 years. Still, no one could show a client deal that the firm made happen, and a promised magnum of champagne went unclaimed.
  • That memo was rewritten four ways: standard (S-C-A), direct (A-S-C), concerned (C-S-A) and aggressive (Q-S-C). The content stayed the same. The tone shifted each time.
  • History, background and definitions may appear, but only woven into the story.
• State Key Line points as ideas, and give each its own mini S-C-Q. Ch. 4 · Fine Points of Introductions
  • A list of topics tells the reader nothing and hides a jumbled argument.
  • The first point's lead-in recalls why its subject matters. Later lead-ins link to the point just covered.
  • A project-team memo listed six sections, from Background to Prerequisites for success. None carried a message. Benefits likely belonged under principles, and prerequisites under organization.
  • A TQM paper asked why leaders keep gaining while adopters don't. Its answer was Benchmarking and Activity-Based Management. One point's lead-in noted that cutting loan processing from 2 days to 2 hours seems enough, until you compare with competitors.
• Recurring reader questions produce standard introduction patterns. Ch. 4 · Fine Points of Introductions
  • Funding request: a problem, a costed solution, and the implied question of approval.
  • Progress review: what we said, what you asked us to check, and what we found.
  • A department's work grew 20% a year for 4 years while head count stayed at 14. The backlog reached 22 weeks. An IBM installation would cut backlog and overtime, so the memo urged approval.
  • A funding case rests on urgency, fit and cost offset. Extra benefits come only if the facts support them, and they are never the reason to act.
  • One client hoped a capacity problem would pass because a competitor was failing. The next review reported the problem remained, only worse.

Lead with the action and push deduction low in the pyramid

Deduction and induction carry the same reasoning, but induction puts the action first and is easier to read, so deduction belongs low in the pyramid.

• Ideas in a grouping can relate in only two ways: as a chain ending in a therefore, or as a class of like things that supports an inference. Ch. 5 · Deduction and Induction: The Difference
  • A deductive group states a situation. It then states a second situation that comments on the first one's subject or predicate, and ends with what follows from both.
  • An inductive group names its ideas with one plural noun. Its points do not derive from each other.
  • The test is simple. If the second point comments on the first, the group is deductive. If both fit the same plural noun, it is inductive. If neither, the points don't belong together.
  • Birds fly, I am a bird, therefore I fly. The summary leans on the last point: I fly because I am a bird.
  • French, German and Russian tanks at the Polish border suggest Poland is about to be invaded. Define the same movements as Poland's allies preparing to attack the rest of Europe, and the inference changes completely. The inference rests on how the group is defined.
  • A logic book offered two "deductive fallacies." One was that all rabbits are fast, some horses are fast, so some horses are rabbits. The second point never comments on the first. It only adds a member to the class, which makes it induction, so no deductive conclusion could follow.
  • A chain may skip steps the reader can supply. The claim that selling used newspaper to Asian countries could worsen Southern California's newsprint shortage needs six steps if spelled out in full. Writing every one would be slow and pedantic.
• Choosing between them is a layout decision, not a choice of logic. Ch. 5 · Deduction and Induction: The Difference
  • Sort the results of any analysis into a Recommendation Worksheet. It has three columns: what is going wrong, what causes it, and what to do.
  • Read it column by column and you get deduction. Turn it 90 degrees and the recommendations land on the Key Line, each with its own findings and causes beneath it.
  • The deductive layout makes the reader hold every problem, match each one to a cause, then match both to actions, all before learning what to do.
  • Deduction is not the stronger form. It is all the same reasoning; we are only discussing how to lay it out on the page.
  • Findings and conclusions differ only by an arbitrary level of abstraction. A summary of a group of findings is always a conclusion.
  • The inductive layout fences subjects apart. Everything about one subject sits in one place.
• Put deduction near the top only when the reader can't accept or follow the action without the reasoning first. Ch. 5 · Deduction and Induction: The Difference
  • There are two cases: the answer is alien to what the reader expects, or the action can't be understood without explanation. Few business readers fall into either.
  • Deduction stays clear when its points sit close together. It loses clarity when pages separate them, so the paragraph level suits it best.
  • Keep a deductive group to four points and no more than two therefores. Past that it gets too heavy to summarize.
  • A client asks how to cut costs. "Do A, B and C" fits an inductive pyramid. "Forget cutting costs and sell the business" triggers why, how, are you sure. So it needs a chain: a growing threat from abroad, a structure that can't respond, and other owners who could.
  • In David Hertz's article on risk analysis, readers needed the reasoning behind the approach before they could follow its steps. That justified a deductive Key Line.
  • A case for buying the Pacific Avenue warehouse stated the criteria, said the building met them, then said buy it. The last step raised no question, so the argument was overstructured. The inductive version recommends the purchase and lists how the building meets each criterion, such as its price and corner location.
  • If there is only one piece of evidence for a point, it has to be handled deductively.
• An inductive inference must speak only about the ideas under it, and facts with nothing in common are news. Ch. 5 · Deduction and Induction: The Difference
  • Name the group with one plural noun and drop any idea that doesn't fit.
  • Hold either the subject or the predicate constant across the group, and vary the other.
  • Check from the bottom up. If the top point could equally cover other things, lower its level of abstraction.
  • Maupertuis is called ingenious but impractical. His schemes back this: a city where only Latin is spoken, studies of the mind using opium, gravity to explain the embryo, and a hole dug to the earth's center. One noun, schemes, covers them all, so the inference holds.
  • One list of managers who ignore reality, reject criticism, keep losing activities, neglect details and never question policy concluded that they mismanage on purpose. The ideas are only examples of mismanagement. The inference reaches past them.
  • Low productivity, high overtime and uncompetitive prices were grouped as signs of a profit-improvement opportunity. That label is too broad for these three. The group is really a disguised chain, with each problem causing the next.
  • Japanese drives into China, Indonesia and Australia support a move into Southeast Asia. Drives into China, Iceland and Peru share nothing, so they support no point. Being true is not reason enough to include a fact.

A grouping proves itself by its order and by the point it implies

A list of ideas is only valid when it follows time, structural or degree order, and only finished when the point it implies is stated outright.

• Every inductive grouping must follow one of three orders, and a list with no order is broken. Ch. 6 · Imposing Logical Order
  • The mind forms groups through only three analyses. Finding causes gives time order. Dividing a whole gives structural order. Classifying gives degree order.
  • Deductive groupings need no chosen order, because the argument sets it.
  • Groups larger than five almost always hide closer links. Subgroup them and order the subgroups.
  • Buyers sent eight complaints about sales and inventory reports. Labeled by type, they fell into bad timing (1, 3), poor data (2, 4, 6) and unhelpful format (5, 7, 8). The order of the three shifts with the question. For why buyers hate it: late, can't find anything, then it's wrong. For how to fix it: decide data and layout, ensure reliability, send it on time.
  • A consultant's eight Phase 1 productivity steps were in sequence but at mixed levels. Regrouped, they became three steps: determine critical functions (1, 2), identify weaknesses in performing them (4, 5, 6), recommend practical ways to change (8). The regrouping showed where steps were thin.
  • Splitting the Ten Commandments into sins against God and sins against man adds insight a flat list of ten lacks.
• Time order exposes steps that mix causes with their effects. Ch. 6 · Imposing Logical Order
  • Picture yourself doing each action. Name what you hold when it ends.
  • Ask whether an action comes before the next one or is done in order to achieve it.
  • Trace a set of conclusions back to its underlying process, then check every step is covered.
  • A strategic planning cycle listed seven items: three company actions and four results tracing a product life cycle. Imagining the steps exposed a break after implementation. Item 6, high cash generation, was dropped. The fixed list: perceive need, develop strategy, implement it, assess market reaction, change strategy to match.
  • A subordinate planned to call investment evaluation technically unsound, simplistic and misleading. The third point was the effect of the other two, so it went on top. The process beneath was concept, technique, application. The draft never addressed application, so a reviewer would know what to ask.
  • A transportation manager's five objectives mapped onto design, construct, operate, maintain. The real objective was likely whether the Department is organized to do those four.
• Structural order needs MECE parts in the sequence of the whole, and degree order puts the strongest first. Ch. 6 · Imposing Logical Order
  • Parts follow the principle of division, how the thing functions, or how the eye reads a map.
  • Equally important, simultaneous changes go in the order you would draw them on blank paper.
  • Leading with the weakest point for drama is a style choice, not a logical one.
  • A radar set runs by function: modulator, oscillator, antenna, receiver, indicator. Burton Bernstein's Sinai piece in The New Yorker, June 4, 1979, moves clockwise across the map from the upper left.
  • A city had 25 departments under 23 committees. The plan cut to about 6 under 6 committees plus an administrative arm. Its four changes were ordered as drawn: committees, program administrations, internal services, then a Chief Executive.
  • Eight causes of New York's decline sorted into high costs, unsuitability of the area and alternative choice. Ranked by importance: high costs, difficult working conditions, attractive alternatives.
  • Eleven risks of making plastic bottles, placed on ROI and EPS trees, showed a possible ban or diluted profits. You cannot tell that nonsense is being written unless you first impose a structure on it. The company went ahead with immense success. The trees had missed the effect on sales.
• Summarize a group of actions by the effect they produce, worded as an end product. Ch. 7 · Summarizing Grouped Ideas
  • Actions relate only through their joint effect. Until it is stated, you cannot see which steps belong or are missing.
  • "Improve profits" fails. A 2 percent gain and a 10 percent gain need different steps.
  • Separate what is done before from what is done so that. Keep each group to five steps or fewer.
  • Labeling actions as Tasks, Objectives and Benefits only adds repetition.
  • Vague phrases became end products. "Strengthen regional effectiveness" became "Assign planning responsibility to the regions." Picturing regions writing annual plans prompted new steps, such as a planning review group.
  • An accounts receivable list hid a plan: Accounting bills at 1 month and sends a note at 2, salesmen call at 3, Directors at 4, then a collection agency. "Reduce accounts receivable" was still wrong, since one paid bill does that. The end product was a follow-up system, which exposed a gap: stop calling on chronic nonpayers.
  • A training proposal listed six tasks, five objectives and three benefits, with "transfer know-how" repeated. It became one promise, rapid transfer of planning know-how, with three supporting actions.
• Summarize a group of situations by the insight their shared feature implies. Ch. 7 · Summarizing Grouped Ideas
  • Strip elegant wording to bare subjects and predicates. Smooth language soothes and blocks critical thought.
  • Look for the same kind of subject, predicate or implied judgment. Identical subjects mean grouping by predicates, and the reverse.
  • No clear link means the grouping is wrong.
  • On the radio, John Wain claimed kinship with Johnson through background, Oxford and taste. A listener mocked Staffordshire, and the talk moved on. Saying they were the same kind of people would have given a real point to answer.
  • Five information complaints split into missing (2, 3, 5) and inadequate (1, 4). The summary: useless for planning. The order hinted at a third defect, adequate data badly presented.
  • Aftermarket conclusions became two: only some parts are attractive, and those are hard to enter. The reasoning never reached a therefore such as forget it or buy your way in.

Let the eye see the pyramid on the page, on the screen and in the sentence

Every surface the reader meets, from a memo's layout to a slide to a single sentence, should show the pyramid's shape so the reader can rebuild it.

• Show the Key Line with the lightest visual device that suits the document's length. Ch. 10 · Reflecting the Pyramid on the Page
  • With about a paragraph per Key Line point, underlining the points is enough to make them stand out.
  • When each point needs more than a paragraph or two, state the points first. Then give each one its own heading.
  • Hierarchical headings move subordinate ideas further right and give ideas at the same level the same look.
  • The Superfight Sweepstakes memo numbered and underlined three concerns as questions. How will viewers learn the rules? Will they do the forecast? Will the commercial be clear? Viewers could enter via TV, but the official rules ran only in the newspaper, a possible lottery problem.
  • The August 25 field sales memo listed three actions with deadlines. Select a chain by July 11, collect data by August 10, and organize and return it by August 15. Each action then got its own heading, starting with SELECTING THE CHAIN.
  • Keep underlined points under about a dozen words, with one subject and predicate. Use no more than four in a chained deduction or five in an induction. The full format looks rather ugly, so bold the major points and underline the lower ones.
  • Decimal numbering interrupts concentration and forces renumbering after edits. Pair it with headings and cite by number plus topic, such as Section 4.1 on manufacturing profits. Antony Jay's Section 5.1 splits into 5.1.1 Visuals, 5.1.2 Time and 5.1.3 The best way, so the numbers match the ideas.
• Headings serve the eye, so they follow strict rules and never carry the message alone. Ch. 10 · Reflecting the Pyramid on the Page
  • Never use a single heading, number or dash point at any level. A heading signals one member of a group.
  • Word headings in a group in parallel form, such as all verbs. Parallelism applies within a group, not across groups.
  • The text must read smoothly with the headings removed, because readers skim them.
  • Before any group of headings, state the major point and the ideas to come. Never put a section heading right after the title, or a subsection heading right after a section heading.
  • Don't add headings just to break up the page, as newspapers and magazines do. Often nothing below the major section headings is needed.
  • Numbered paragraphs are the one exception to reading smoothly without headings. Three centered stars, called a sunset, signal that a concluding comment is coming.
• Link sections by what they say, and end only when the ending adds something. Ch. 10 · Reflecting the Pyramid on the Page
  • Open a section with a short Situation-Complication-Question story the reader accepts, or pick up a key word from the part just finished.
  • A sound pyramid needs no conclusion. If you add one, show the message's significance and prompt action, never a lame restatement.
  • Add Next Steps only when the reader will act soon, and list only actions no one will question.
  • A report told the Ritz-Ryan hotel chain it was not exploiting common ownership of its hotel, restaurant and catering operations. Its transitions each picked up a key idea, such as the full-time Chief Executive. The closing summary named the Board and Chairman, a Group Managing Director, and three key executives each running a major business.
  • Lincoln ended his second inaugural with malice toward none and charity for all. That ending gave both an insight and a call to action. The pitch must fit the reader. An airline president would resent emotion over a planning system but welcome it on reregulation.
  • A European literature retrieval report closed by saying success would create a common market for information. It called the prospect exciting and offered to launch the pilot.
  • After 30 pages urging a purchase, the Next Steps were three. Ask the owner to lunch, confirm the money with the bank, and reconvene the Acquisitions Committee. Any step that might raise questions belongs in the body.
• A presentation is a show for a distractible audience, so the speaker carries the message and the slides only stress it. Ch. 11 · Reflecting the Pyramid on a Screen
  • Aim for about 90% exhibits and 10% text slides.
  • A text slide holds one idea in about 6 lines or 30 words. Write it as a statement, such as "Sales outlook is favorable", not a caption.
  • Pick the question each exhibit answers, then put the answer in its title.
  • Build from a storyboard. Write the introduction in full, set the Key Line and the level below atop the slides, script around them, and rehearse.
  • A Guiding Principles slide listed seven ungrouped principles for a Professional Health Care Sector supply chain vision. Presenters with 50 or 60 such slides read each word aloud or reworded them. Gene Zelazny calls this a visual recitation.
  • The Jackson Foods script explained high out-of-stock levels and their three causes. The matching slide, Current Reality, kept only a heading and three bullets. The most interesting thing in the room is always you, not the slides.
  • The Jackson Foods storyboard opened with a supply chain costing $12m a year, 14% of NPS. Its main point was to turn that chain into a competitive service. The steps were to stabilise service, cut obvious costs and build longer-term skills.
  • To size type, divide the farthest viewer's distance in feet by 32 to get letter height in inches. At 16 feet you need 0.5 inch. Zelazny rejects illegible slides outright and claims he can make them legible 98% of the time.
• Clear sentences come from picturing how the nouns relate, then copying that picture into words. Ch. 12 · Reflecting the Pyramid in Prose
  • Pick out the concrete nouns. Sketch their relationship with a few shapes and arrows, then write one plain sentence stating it.
  • Bad prose calls up no image. By the end of a long noun chain, the start has faded.
  • If a phrase can't be pictured, flag it and ask the writer.
  • One sentence buried the sales force, the store and the changed trade environment in abstractions. Pictured, the core link was salesman to store. The rewrite said the sales force must be redeployed to match the new trading environment.
  • A cash flow sentence resolved into one image. Feed the right information in and the system yields accurate analyses. Without the writer, nobody could say what that information was.
  • Reading of Thoreau cutting white pines by Walden Pond in March 1845, you build a picture bit by bit. George Miller calls it a memory image. You later read the content back off it.
  • Professor William Minto likened writing to filing a battalion through a narrow gap one man at a time. The reader must re-form the troops, so order is a duty owed to readers.

Find the reader's question in the gap between the result now and the result wanted

A problem document's introduction falls out of a problem definition: the gap between the result now and the result wanted, plus what the reader has already done about it.

• Define the problem by sketching four elements on one page before writing. Ch. 8 · Defining the Problem
  • The Opening Scene is a structure or process, such as an org chart, plant locations or a sales routine.
  • The Disturbing Event is what upset that stable scene. It may be external, internal or only recently noticed.
  • R1 is the undesired result. R2 is the desired result, stated with a number or a definite end state.
  • The real estate firm's Opening Scene was a three-part process: list prospects, write a script, deliver it. The likely causes sat inside it. The list was out of date, the script lacked punch, or the delivery failed. Fixes usually come from changing the structure or process.
  • Pitch the Opening Scene at a typical Fortune or Business Week reader, or picture telling the story to a friend. Keep it short.
  • If you lack facts for a Disturbing Event, as often happens in proposals, skip straight to R1. Do not invent one.
  • If R2 cannot be stated precisely, write the general state wanted. Finding the specific R2 then becomes the first step of the work. Without it, solutions cannot be compared.
• Where the reader stands on a solution decides which question the document answers. Ch. 8 · Defining the Problem
  • Many histories are layered: a solution was proposed, tried or rejected. The reader's most recent action becomes the Complication.
  • A common error is never settling, in your own mind, whether the reader has already acted.
  • Seven reader positions each map to a standard S-C-Q pattern.
  • The supermarket history had three layers. The maker wanted week-long shelf tests. Supermarkets balked, then took a modest fee. As chains formed, the fee climbed, and an industry committee could agree only on refusing to pay. That history fit in a few sentences of a speech.
  • Failed solution pattern: S holds the situation, R1, R2 and the solution. C says it did not work. Q asks what to do now.
  • Two positions are rare. In one the reader knows R1 but not R2, so the question is what the objectives should be. In the other, as in benchmarking, the reader knows R2 but is unsure whether R1 exists.
• Most introductions answer one of four standard questions. Appendix B · Examples of Introductory Structures
  • The four: what should we do, should we do what we plan, how do we do something, and why did it happen.
  • Consultants usually answer only the first three.
  • What to do: Simmons & Smith expected its NC membrane market to double in 3 years. It disliked the 30% dealer commission and saw dealers starting to compete. The answer was a separate distribution approach for each market.
  • Should we change: Diffraction Physics led European scanner sales until NCR/ICL offered cheap unbundled scanners. The answer was a major unbundled effort now.
  • How to implement: San Sebastiano could not create jobs after Department of Defense cuts, so it should start with actions local efforts could begin. Do we have a problem: Anielski Airlines learned deregulation was a catalyst, not a problem.
• A request for alternatives usually means R2 is vague. Appendix B · Examples of Introductory Structures
  • A recommendation either gets the reader from R1 to R2 or it does not.
  • Weighing options by pros and cons is a backdoor attempt to define R2.
  • Raise alternatives only if the reader already knows them, and put them in the Complication. Never raise them just to reject them.
  • Strictly speaking there is no such thing as an alternative solution to a problem.
  • The household distributor served 438 stores from warehouses built for 490, with 198 new stores planned. Its R2 was lowest capital and operating cost at today's speeds. The answer was incremental capacity and a new center in Georgia or the Carolinas for start-up in 2002.
  • The motor maker lost its largest customer to a competitor's 7 3/4 HP model after a ruling favored a 5-105 HP motor. It had three known responses, so "which makes most sense" was the right question.
  • If a reader insists on options despite a clear answer, put them in an appendix chart: options down the side, criteria across the top, check marks in the cells. Colefax's SABRE question, mainframe or branch, ended branch-based.
• Proposals follow the steps or the reasons to hire, and process fixes follow the changes. Appendix B · Examples of Introductory Structures
  • Use steps for repeat clients and reasons to hire in competitive bids. Drop the Benefits section, since solving the problem is the benefit.
  • Keep fees, staffing, timing and qualifications outside the logic.
  • For process fixes, draw the before and after, and put the changes on the Key Line. Explain only the steps where problems occur.
  • The telephone-software proposal's Key Line had four steps: find segments and buying behaviour, project spending by application, judge supplier positioning, and pick markets and approaches.
  • The Period Graph Books memo said only that the process needed streamlining. Drawing it showed data keyed in three times by three departments. Errors came at entry, at calculation and from unannounced presenter changes. Three fixes became the Key Line.
  • The DDT report compared mailed copies arriving in 7 to 10 days with screen delivery within an hour. The real question was feasibility. The answer: feasible but not yet practical, so start with a demonstration project.
  • The market-forecasting fix added steps the old process lacked: inventory policies, a trend forecast and expert adjustment before the monthly meeting.

Structure the analysis before gathering data, and invent a structure when none exists

Problem solving runs fastest when you picture the structure that produced the result, turn it into yes-or-no questions, and gather only the data that answer them.

• Thinking before collecting data saves work and yields a report that shows its insight. Ch. 9 · Structuring the Analysis of the Problem
  • The data-first habit began in early consulting, when firms lacked industry knowledge and studied everything on every job.
  • It produced loosely relevant facts, missing support for major recommendations, and long, dull reports.
  • Better firms work like the scientific method. They form hypotheses, then design crucial experiments that rule each one in or out.
  • The approach to avoid had four steps. Find the industry's key success factors, assess the client's strengths and weaknesses, compare the two, then recommend. Its reports ended up grouped under Operations or Marketing, or under Findings, Conclusions and Recommendations.
  • Barrows' new Information Systems Division (ISD) grew faster than expected. It still fell behind on orders despite new production planning and control systems. The consultant's proposal listed eight broad data areas to cover by interview, with no way to judge relevance.
  • The better plan was to draw a flow diagram of current operations first. He would then ask six yes-or-no questions on lead times, purchased items, stock availability, capacity, system costs and management reports. That let him name each data source, assign responsibility, set a schedule and estimate cost before starting.
  • A report that takes huge effort and still reads poorly is a warning sign. It means the thinking necessary to write clearly did not take place early enough in the process.
• A diagnostic framework pictures the problem area and turns each possible cause into a yes-or-no question. Ch. 9 · Structuring the Analysis of the Problem
  • The framework comes from the Opening Scene. It can be built only by dividing, tracing cause and effect, or classifying, with MECE top branches.
  • Each yes-or-no question defines the data needed and shows in advance when the research is done.
  • A headache splits into physical causes (external, such as a bump, allergies or weather, or internal) and mental ones (stress, hypochondria). You then check the easy causes first, not a brain tumor for a sinus headache.
  • Telephone switching installations ran late. The tree split the effect into fewer men per rack, more hours per man per rack, or fewer hours on duty per week. More hours per man split again into slower work, a job demanding more time, or unexpected delays. Industry experience said where to look first.
  • A retailer's falling sales were classified into semi-fixed factors (falling market, store number and location, size, accessibility) and variable ones (wrong goods, layout, price, display, staff, advertising). For each, you state what data would prove it, such as store coverage not matching the market.
  • A choice structure follows the selling sequence. Ineffective sales support fails at retail or at headquarters. At retail you are in the right stores or not, at the right call frequency or not, doing the right activities or not.
  • B. Robert Holland's sequential marketing structure orders decisions so weaknesses on the left get fixed first. There is no point pushing more frequent use before promotion works. If no branch finds a problem, recheck the target market and consumer benefit.
• Logic trees take over once causes are known, laying out actions and exposing weak lists. Ch. 9 · Structuring the Analysis of the Problem
  • Problem solving asks five questions in order: is there a problem, where, why, what could we do, what should we do. Diagnostic frameworks answer the second and third. Logic trees answer the last two.
  • Split a cost mathematically, list the MECE ways to move each factor, then weigh each action's benefit and risk.
  • Mapping any list of points onto a tree shows items that don't belong and branches nothing covers.
  • A cigarette company's labor cost per cigarette became cost per hour times hours per million cigarettes. Cost per hour falls through less overtime, cheaper labor or smaller wage awards. Hours fall through fewer people per machine, higher machine speeds or better machine efficiency.
  • A factory listed ten Major Issues for cutting energy costs. A tree split the goal into fewer BTUs used and lower cost per BTU. Issues 7, 8 and 9 had nothing to do with the subject. Issue 10 restated the whole goal.
  • For the Texas distributor, the proper issues became two. Does the central system order properly? Does it keep too much obsolete and slow-moving stock? The proposal's so-called issues were really the steps of the study.
• Issues must come from the structure behind R1, and no framework replaces deep knowledge of the field. Ch. 9 · Structuring the Analysis of the Problem
  • An issue demands a yes or no answer. A plain list of client worries should be called concerns.
  • The client's question states R2, so it cannot generate the issues.
  • Experience decides which framework to use and where to look first.
  • David Hertz and Carter Bales at McKinsey & Company coined Issue Analysis in a 1960s study for New York City under Mayor John Lindsey. For middle income housing they mapped Major Decision Variables, such as tenant selection policy, onto the physical system. The method proved too complex and was dropped, but its name survived.
  • A UK retail bank sought a European strategy once the EC allowed cross-border banking. Its consultants taught seven steps starting from the client's question. The steps look close to right but confuse young consultants, because they start from R2.
  • A separate hypothesis step adds nothing, since it doesn't matter whether you guess yes or no. A separate Issues section in a proposal is never needed either. Decision trees and PERT diagrams are different tools, because they show a need for action rather than generate questions.
• When a result can't be explained at all, invent a structure and test it with an experiment that gives a clean yes or no. Appendix A · Problem Solving in Structureless Situations
  • The structure may be absent (the telephone), invisible (the brain, DNA) or wrong (Aristotle's force and the cannonball).
  • Reasoning links a Rule, a Case and a Result. Analytical problem solving starts from a bad Result and a known Rule, then checks the Case.
  • Hypotheses come from the situation's structural elements, often through an analogy.
  • Falling sales show all three forms. Deduction says a price set too high must lower sales. Induction sees a price rise followed by falling sales and infers the price is probably the cause. Abduction starts from falling sales, recalls that high prices often cut sales, and checks the price.
  • Alexander Graham Bell saw that the ear's delicate membrane moves relatively massive bones. He reasoned that a thicker, stouter membrane could move his piece of steel.
  • Galileo listed weight, distance fallen and time fallen as hypotheses for what force is proportional to. If it was time, distance would grow with the square of time: 1, 4, 9. An inclined plane slowed the ball enough to measure, and the relation held.
  • Just varying conditions to see what happens is not enough. The test must say plainly whether to keep or drop the hypothesis. Charles Sanders Peirce coined the term abduction in 1890.

Put it to work

Before you write the next spec, memo or deck, you name the reader's question and put your answer to it at the top.

1. Write down the gap before anything else. On one page, sketch the Opening Scene, the Disturbing Event, the result you get now (R1) and the result you want (R2). State R2 with a number or a clear end state so you can tell when it is reached. Ch. 8 · Defining the Problem
   • If you cannot state R2, finding it becomes the first piece of work. Without an end-product R2 you cannot choose between possible solutions.
   • A household goods distributor had warehouses meant to serve 490 stores, and they already struggled with the present 438. With 198 new stores planned, it expected to run out of capacity in 2 years.
2. Turn the problem into yes-no questions before collecting data. Draw the structure of the area where the problem arose and split it into MECE branches. Phrase each possible cause as a yes-no question, then gather only the data that answers it. Ch. 9 · Structuring the Analysis of the Problem
   • At Barrows, a consultant first planned to gather eight broad data areas through interviews, which would have buried him. A flow diagram of current operations gave six yes-no questions instead, on lead times, purchased items, stock availability, capacity, system costs and management reports.
   • The trap is drawing issues from the client's question. That question only states an R2. Issues come from the structure that produced R1.
3. Open with a story the reader already knows. Give a Situation the reader accepts, then a Complication that raises one Question, then your Answer. The last thing the reader already knows is the Complication. Ch. 4 · Fine Points of Introductions
   • A project team memo opened with a list of six section names, from Background to Prerequisites for success. The list carried no message and hid that the ideas underneath were jumbled.
   • Keep proof and exhibits out of the opening. An introduction reminds the reader. It does not inform.
4. Put the action on the Key Line and the reasoning beneath it. Answer the reader's question directly, then group the findings and causes under each recommendation. Lead with a deductive chain only when your answer will surprise the reader, or when the action makes no sense without the reasoning. Ch. 5 · Deduction and Induction: The Difference
   • A client who asks how to cut costs can simply be told to do A, B and C. A client told to forget cutting costs and sell the business will ask why, so that answer needs the argument first.
   • Lay the findings, conclusions and recommendations out as three columns. Then turn the sheet 90 degrees so the recommendations become the Key Line.
5. Give every list an order before you trust it. Check that each grouping follows time, structural or degree order, and keep it to four or five points. If no order appears, the grouping is broken. Ch. 6 · Imposing Logical Order
   • Eight buyer complaints about a reporting system sorted into three types: bad timing, poor data and unhelpful format. The order of the three then changed with the question being answered.
   • A draft on investment evaluation commented on only two steps of a three-step process: concept, technique, application. Putting the points in time order showed the third step was missing.
6. Build the deck from a storyboard, not from the report. Write the introduction in full. Then put a sentence stating each slide's point across its top. Aim for about 90% exhibits, and keep text slides to about 6 lines or 30 words. Ch. 11 · Reflecting the Pyramid on a Screen
   • The Jackson Foods storyboard opened on a supply chain costing $12m a year, 14% of NPS. It moved through current reality, failed improvements and strategy, with exhibits such as 50% of customers accounting for 95% of orders by value.
   • The trap is the visual recitation: wordy slides read aloud word for word. The speaker carries the message and the slides only stress it.
7. Picture each heavy sentence before you rewrite it. Pick out its concrete nouns and sketch how they relate with simple shapes and arrows. Then write one plain sentence that states that relationship. Ch. 12 · Reflecting the Pyramid in Prose
   • A long sentence about the cost-effectiveness of sales-force deployment and redefined selling missions came down to salesman and store. The rewrite simply said to redeploy the sales force to match the new trading environment.

The book's words

A handful of terms cover the whole method, from defining the gap to titling the slide.

• Pyramid structure Ideas grouped under one governing thought, where each level summarizes the group below it.
  • Three rules test it before drafting. Each point summarizes its group. Each group holds one kind of idea. Each group is in logical order.
• Key Line The row of points directly under the main point, which answer the question the main point raises.
  • State these points in the introduction so the reader gets the whole argument in about 30 seconds.
• Situation, Complication, Question The story an introduction tells to lead the reader to the question the document answers.
  • David Hertz on risk analysis: executives must choose among capital investments, they cannot judge the risk of uncertain assumptions, so is there a realistic way to measure it? The answer is yes.
• R1 and R2 R1 is the result the reader now gets and dislikes. R2 is the result the reader wants instead.
  • Where the reader stands on a solution, untried, chosen, or failed, decides which of seven introduction patterns fits.
• MECE Mutually exclusive, collectively exhaustive: the parts of a whole overlap nowhere and leave nothing out.
  • A headache's causes split into physical or mental, so you can rule out easy causes first and skip the brain tumor test for a sinus headache.
• Intellectually blank assertion A heading like "five problems" that names the kind of idea coming but states no idea.
  • Such labels bore the reader and cover up incomplete thinking. Writing the real summary completes the thinking.
• End product The tangible result you could hold once an action is done, used to word steps and spot missing ones.
  • "Reduce accounts receivable" became "Establish a system for following up overdue accounts," which exposed a missing step about chronic nonpayers.
• Issue A question phrased so it can only be answered yes or no.
  • It comes from the legal phrase "at issue," meaning two sides with one prevailing. A plain list of worrying topics is better called concerns.
• Abduction Reasoning from an observed result to a possible cause, then checking whether that cause holds.
  • Charles Sanders Peirce coined the term in 1890. Galileo used it when he guessed force related to time and tested the guess by rolling a ball down an inclined plane.
• Visual recitation A slide show where wordy slides are read aloud, instead of a real visual presentation.
  • The name is Gene Zelazny's. A typical example is a slide of seven ungrouped guiding principles with no summarizing insight.

Chapter by chapter

Where each idea is in the book.

• 1 Why a Pyramid Structure? Writing is clearest when ideas form a pyramid under one governing thought and reach the reader from the top down. Readers sort what they read into summarized groups anyway, so a writer who presorts and states the summary first spares them the effort and the misreading.
• 2 The Substructures Within the Pyramid Writers rarely know their ideas fully before writing, but the pyramid imposes three substructures that speed up discovering them: a vertical question/answer dialogue, horizontal groupings that are either deductive or inductive, and an introduction that tells a Situation, Complication, Question story leading to the Answer.
• 3 How to Build a Pyramid Structure Build the pyramid from the top down first. Use the introduction's Situation, Complication, Question and Answer to find the main point and the Key Line. When the top won't come, work from the bottom up: list your points, find how they relate, and draw conclusions.
• 4 Fine Points of Introductions An introduction tells the reader a story they already know. It sets out a Situation, adds a Complication that raises the one Question the document answers, and then gives the Answer. Documents tend to answer one of a few common questions, so introductions fall into a handful of recurring patterns.
• 5 Deduction and Induction: The Difference Ideas in a grouping relate either deductively, where each point derives from the one before and leads to a therefore, or inductively, where similar ideas are classed under one plural noun and an inference is drawn about them. Use induction at the top of the pyramid and push deduction as low as possible, because induction is easier on the reader.
• 6 Imposing Logical Order Every inductive grouping of ideas must follow one of three logical orders: time, structural, or degree. Each order reflects the analysis the mind used to form the group. Finding the order, or failing to find it, shows whether the grouping is valid and complete and reveals the real message.
• 7 Summarizing Grouped Ideas Every grouping of ideas implies a point, and the writer must state it. Summarize action groupings by stating the specific effect the actions produce, and situation groupings by stating the insight their similarity implies.
• 8 Defining the Problem A problem is the gap between the result a situation now produces (R1) and the result the reader wants (R2). Laying out the Opening Scene, Disturbing Event, R1 and R2, and noting where the reader stands on a solution, tells you the reader's question and gives you the Situation, Complication and Question of your introduction almost directly.
• 9 Structuring the Analysis of the Problem Structure the analysis before gathering any data. Diagnostic frameworks turn the problem area into yes-no questions that find the causes, and logic trees lay out solutions and test grouped ideas. Both make the work faster and the final pyramid easy to build.
• 10 Reflecting the Pyramid on the Page Lay out a document so the eye sees the pyramid's hierarchy at once, and link its groups with smooth transitions. The reader should grasp the whole argument (introduction, main point, Key Line points) within the first 30 seconds.
• 11 Reflecting the Pyramid on a Screen A visual presentation is not a report turned into slides. It is a show for a live, distractible audience, so the speaker carries the message and the slides only emphasize the pyramid's major points through sparse text and clear, titled exhibits built from a storyboard.
• 12 Reflecting the Pyramid in Prose Clear prose comes from seeing what you mean before you write it. Picture the relationships among your ideas as a simple image, then copy that image into words so the reader can rebuild it.
• Appendix A: Problem Solving in Structureless Situations Some problems are not about disliking a result but about being unable to explain it, because the structure is absent, invisible or wrong. These are solved by scientific abduction: invent a structure that could explain the result, then test it with an experiment that gives a clear yes or no.
• Appendix B: Examples of Introductory Structures Most document introductions answer one of four standard reader questions: what should we do, should we do what we plan, how do we do something, or why did it happen. A few hard cases (proposals, alternatives, process changes) become easy once the writer defines the problem precisely, often by picturing the before and after.