The Communicator's Mindset: Audience, Message, Medium

Presentation track, branch pr-foundations (Foundations) · Node PR0.01 · ~5 focused hours · Mastery gate ≥ 85%

A founder builds a forty-slide deck for her board over eleven hours. The board gives it nineteen minutes. Eleven of those nineteen go to slide 6, a chart she assembled in four minutes because the section looked thin. Nothing on the other thirty-nine slides changes a single thing anyone does on Monday.

The design was fine. Every slide was aligned, the fonts matched, the logo sat in the same corner throughout. The failure happened before the software opened, in the ninety seconds when she should have asked who is in the room, what they're deciding, and what one sentence would decide it. She skipped those ninety seconds and paid for them with eleven hours.

That's a poor rate of exchange, and it's worth asking yourself whether you've ever made the same trade.

The whole Presentation track is built on a claim you can test against your own experience. The quality of a deck is set almost entirely by decisions taken before any slide exists, and the decisions taken inside the tool mostly determine whether that earlier thinking survives contact with a reader. Sixteen hours is a realistic build for a twelve-slide decision deck. Eight of those sixteen come before the first slide is created. That split says nothing about how slow people are at PowerPoint. It shows where the difficulty of making a deck lives.

So the first thing taught here is a sequence, and a tool only enters at its end: takeaway, audience, decision, medium, and only then artefact. Every later branch assumes you have run that sequence. The mechanics branch teaches you to build what the sequence chose. The chart-craft branch teaches you to draw the evidence it demands. The storytelling branch teaches you to order the claims it produced. None of them can rescue a deck that started with the wrong takeaway, because there's no arrangement of slides that answers a question nobody asked.

There's a limit to what can be taught here, and stating it now saves a lot of disappointment later. Taste doesn't come from a module. It comes from reps, from feedback that stings, and from having seen four hundred decks. What can be taught is grammar: the rules that separate a chart that reads from one that does not, the structures that make an argument follow, the checks a machine can run on a slide before a human wastes attention on it. Grammar isn't style, and knowing it won't make you interesting. It will make you correct, which is the part that's teachable and the part that most working analysts are missing.

One more thing about the shape of what follows. Every exercise you'll meet is graded twice, once by a machine and once by you against a rubric you can't see until you have committed your own score. The machine checks seven rules on a slide: alignment, one message, chart choice, the title stating the message, colour discipline, declutter, and hierarchy. It's a floor. A slide can pass all seven and still say nothing worth saying. The rubric exists for that slide, and so the two scores are blended, with neither replacing the other.

Learning objectives

You can:

1. Write a one-sentence takeaway for any piece of analysis before opening any tool, and test it against three properties: it is a claim, it is falsifiable by the data you hold, and it names or implies the action it supports.
2. Build an audience profile on three axes (what they already know, what they are deciding, what they are afraid of), and show how a change on any one axis changes the takeaway.
3. Find the decision a communication exists to serve when the requester has not stated one, using the four questions that recover it, and state what to do when the honest answer is that no decision is pending.
4. Choose a medium on purpose from the six standard forms, pricing each in reader-minutes for a stated audience size, and justify the choice against exactly one rejected alternative.
5. Distinguish a reading deck from a presented deck, size each correctly in words per slide, and compute the attention cost of the hybrid that tries to be both.
6. State what a slide is: one claim with the evidence that supports it, and apply that test to reject slides that are containers for what you know.
7. Apply the seven machine-checkable rules as a floor, compute a blended machine-plus-rubric score, and say what the machine cannot see.
8. Budget the hours for a deck of a stated size, name where they go, and defend the share that is spent before any slide exists.
9. Diagnose the three arrival failure modes (the data dump, the beautiful empty deck, and the deck that answers a question nobody asked) from symptoms, and name the repair each one needs.
10. Ask for feedback in a form that produces useful answers, and triage a redline into substance, structure and style so the expensive fixes happen first.

Prerequisites & connections

Builds on. Nothing. This is a day-one entry point, and it assumes no software, no prior deck and no finance beyond arithmetic. If you have already worked through M8.02, the business teardown, or any of the CN case modules, you'll have an analysis of your own to use for the mini-project, which makes the hour worth more. If you have not, the project supplies a fallback brief.

Feeds forward. PR0.02 takes the artefact choice you make here and teaches the tool that builds it. It stops at using a template and leaves authoring one for later. PR2.01 owns chart-form choice in full. The compact grammar table below exists only so the worked examples can name a chart correctly, and it's deliberately shallow. PR4.01 and PR4.02 own the executive storyline and title logic across a whole deck, taking the single takeaway written here and turning it into an ordered argument. PR5.01 to PR5.03 own the archetypes (the consulting deck, the board deck, the research report) that the medium choice here points at. PR6.01 owns delivery and Q&A. PR7.01 is the capstone that runs the whole sequence end to end.

Six neighbours own material that touches this one, and the non-overlaps are deliberate. None of their material is rebuilt here. M10.03 owns the analyst's written deliverables: the one-page memo, the longer research report, the exhibit-design drill and the action-title convention as applied to written work. What is owned here is the choice that precedes all of them, which is whether a memo is the right artefact at all. CN1.03 owns answer-first synthesis, the governing thought, the pyramid and the so-what ladder. The single takeaway taught here is the input that synthesis produces, and the consulting module's own hand-off note draws the boundary from the other side.

The craft neighbours divide the same way. CN1.02 owns hypothesis-driven problem solving and the ghost deck as a consulting artefact, while PR3.02 owns the ghost deck as a build technique. PR2.02 owns decluttering and colour as data-visualisation craft, and PR1.02 owns alignment, grids and type. Where a section below names one of those ideas, it names it and moves on without teaching it.

Core teaching content

1. The one-sentence takeaway, written before any tool opens

Open a blank document, not a blank deck, and write one sentence that a reader could disagree with. That sentence is the takeaway, and until it exists there's nothing to build.

Three properties separate a takeaway from a topic. It's a claim, meaning someone could hold the opposite view and be wrong. It's falsifiable by the data you actually hold, meaning you can point at the numbers that would break it. And it names or implies an action, meaning a reader who accepts it knows what changes.

Run those tests on four sentences about the same quarter at Chandra Naturals, a synthetic Indian personal-care brand whose numbers carry most of the examples here.

The fourth sentence is longer than the others and that's fine. Length isn't the test. The test is whether a board member who reads it knows what she is being asked to agree to, and whether she could argue against it using the same numbers.

Notice what the fourth sentence costs you. It commits. "Growth is strong but cash is tight" survives any outcome, and that's what makes it worthless. A sentence that can't be wrong can't be useful either. The moment you write November rather than February, you have exposed a claim to attack, and the attack is the point. If the board disagrees, you have found the real conversation in the first minute instead of the fortieth.

Here's a working rule, the one habit from this section worth installing permanently. If you can't write the takeaway, you aren't ready to build. Sometimes that means the analysis is unfinished. More often it means the analysis is finished and you haven't yet decided what it means, which is a different problem with a faster fix. Sit with the numbers and ask what you'd tell a colleague in a corridor if you had eleven seconds. Whatever comes out of your mouth is close to the takeaway, and it's almost always sharper than what comes out of your keyboard.

Revise the sentence last, as well as writing it first. When the analysis moves, the sentence moves with it. Every slide built on the old sentence is now wrong in a way that alignment can't fix. Analysts who build first and write the takeaway last discover this at eleven at night.

3. Finding the decision when nobody states it

Most requests arrive without a decision attached. "Can you put something together on our pricing for Thursday?" contains no decision, no audience and no scope, and the analyst who starts building from it will produce twenty-two slides of competitor price tables that answer nothing.

Four questions recover the decision, and they're asked of the requester, not of yourself. You'll feel a pull to answer them from your own desk. Resist it, because from your desk you'd only be guessing.

The first is what happens after Thursday. If the answer is "we discuss it", ask what would have to be true to stop discussing it. Discussion isn't a decision, but there's usually one close by.

The second is who else is in the room. The requester is rarely the decider. Knowing that the sales head will be present, and that she owns the April price list, converts a vague pricing question into a specific one about the April price list.

The third is what would change your mind. This is the sharpest of the four because it forces the requester to name the evidence that matters, which is the evidence you should be building. If the answer is "nothing", you have learned that the decision is already made and you're being asked for justification, which is a different job that you should do knowingly or decline.

The fourth is what happens if we do nothing. The status quo is always an option and is usually the default outcome, so knowing its cost tells you how large your recommendation has to be to be worth the disruption.

Run those four on the pricing request and it collapses into something buildable. The decision is whether to move the 200 ml face wash from ₹349 to ₹399 in the April price list. The decider is the sales head. What would change her mind is evidence about volume loss, and doing nothing costs the margin that a 14.3% price rise would have earned. That fits on one page, with no need for twenty-two slides of competitor tables, and the fifth worked example builds it.

Sometimes the four questions return an uncomfortable answer: no decision is pending. That happens more often than anyone admits, and the professional response is to say so before building. A status update with no decision attached is an information transfer, and information transfers belong in writing, cheaply. Having no decision to serve is fine, as long as you say so. Spending eleven hours as though one existed is the failure.

There's a variant worth naming because it's common in consulting and in equity research. The requester doesn't know what they're deciding, and your first contribution is to tell them. "Here are the three decisions this analysis could support; which one are we actually facing?" is a better opening than any deck, and it takes one email.

4. Medium on purpose, priced in reader-minutes

Six forms carry almost all analytical communication, and each has a cost to the reader that you can compute before choosing. Reading runs at roughly 200 to 240 words a minute for ordinary business prose. The figure used throughout is 220. Speaking runs at about 130. Those two rates do most of the work.

The column that matters is the last one. A presented deck costs nearly forty-nine times what an email costs, measured in the only currency your audience spends. Forty-nine times is a big multiple to pay for the same information. The presented deck can still be the right choice. It's expensive, though. Expensive things need a reason, and the presented deck has to show one.

The reason is almost always one of three. You need the room, because the decision requires several people to hear the same thing at the same time and commit in front of each other. You need the argument, because the case is long enough that a reader left alone would stop halfway. Or you need the questions, because you can't anticipate the objections and need to handle them live. If none of those three applies, the meeting is a habit.

Notice what the table doesn't say. It doesn't say shorter is better. A two-page memo that takes five minutes and lands is worth more than an email that takes fifty seconds and gets misread, and the analyst who compresses a complex argument into 180 words hasn't saved anyone time. The cost figure is one input to a judgement, not the judgement.

The choice is also reversible in one direction only. An email can escalate into a meeting when someone asks a question you can't answer in writing. A meeting can't be un-held. Default to the cheaper medium, and let the audience pull you upward.

7. What is teachable here, and what is not

Two things get confused constantly, and separating them tells you what to expect from the next seven branches.

Taste is the ability to look at a slide and know it's wrong before you can say why. It comes from volume and from feedback: hundreds of decks built, shown, criticised and rebuilt. No module produces it. What a module can do is shorten the loop, by giving you rules that catch the common errors so your feedback is spent on the interesting ones.

Grammar is the set of rules that are stable, checkable and worth memorising. A bar chart sorted by value reads faster than one sorted alphabetically. A title that states a claim is more useful than a title that states a topic. Two series on different units don't go on two axes. None of that's a matter of opinion, and all of it can be tested against a slide by a person or by a machine.

The track teaches grammar and creates the conditions for taste. That division admits the limits, and it also sets your expectations correctly about the exercises. When a drill marks your chart choice wrong, it's applying a rule, and the rule is defensible. When your rubric self-score disagrees with the machine on whether a slide has one message, the disagreement is the lesson, because the machine can't make that judgement and you must.

The analogy with language holds well enough to be worth a sentence. Grammar doesn't make you a writer, and its absence guarantees you aren't one. You can be grammatical and dull. You can't be ungrammatical and trusted.

8. The seven-rule floor, and what a machine cannot see

Every exercise in this track is scored twice. A machine runs seven rules against your slide. You separately score yourself against a rubric, and the rubric stays hidden until your self-score is committed. The same write-before-peek convention runs through every exercise in the course.

The seven rules are these. Alignment: every element sits on a shared edge or a grid column, with no orphan positioned by eye. One message: the slide makes exactly one claim. Chart choice: the form matches the message type, per the grammar table. Title is the message: the title is a sentence with a verb, not a noun phrase. Colour discipline: colour encodes something, or it's neutral, with one accent at most. Declutter: no element that encodes nothing survives. Hierarchy: size, weight and spacing put the most important thing first in the reading order.

The two scores blend at 0.6 machine and 0.4 rubric, and the pass mark is 0.80, rising to 0.90 for the crown nodes and the capstone. Try one. A slide that scores 0.86 on the machine and 0.72 on your rubric blends to 0.804, which passes the ordinary gate and fails a crown by a wide margin. A slide at 0.95 machine and 0.60 rubric blends to 0.810, barely better. That tells you something about the weighting. A technically clean slide with a weak argument doesn't buy its way through. Working the other direction, a rubric score of 0.70 needs a machine score of 0.867 to reach 0.80 at all. That's a high bar, and it's meant to be.

What the machine can't see is the part that decides whether the deck was worth building. It can't tell whether the claim in your title is true. It can't tell whether the decision you targeted is the decision that was pending. It can't tell whether the chart's message is the message that matters, only that the chart's form fits the message you declared. A slide can pass all seven rules and be a beautifully rendered answer to a question nobody asked. The rubric carries 40% for that reason, and every exercise asks you to state the decision before you build.

Treat the seven as a floor in the strict sense. Falling below them is a defect you should never ship. Clearing them isn't an achievement, and a deck that clears them and changes nothing has still failed.

9. Where the hours actually go

Ask someone how long a deck takes and they will estimate the time spent in the software. That's the smallest piece.

Here's a realistic budget for a twelve-slide decision deck built once, on analysis you already hold.

Half the time goes before a slide exists. About a third goes inside the tool, and that third is the part every beginner budgets for and the only part most people count. The last sixth, rehearsal and revision, is the part that gets cut when the deadline moves, and cutting it's why so many decks are shown for the first time to the audience that matters.

Two practical implications. First, if you're three hours from a deadline with no takeaway written, you're three hours from a bad deck, however hard you work in those three hours. The correct move is to buy time or to switch to a cheaper medium. Second, the data-checking hours aren't compressible in the way the building hours are. A wrong number on a beautiful slide destroys the whole artefact, and the recovery costs far more than four hours of rebuilding. It costs the next three months of everything you present being checked twice.

The numbers above scale roughly linearly in slides for small decks and worse than linearly for large ones, because a thirty-slide deck has an ordering problem a twelve-slide deck does not. That's one of several reasons the archetype templates later in the track are sized where they are.

11. Feedback: asking for it, and taking a redline

Feedback that arrives as "looks good" is worth nothing and is your fault, not the reviewer's. The question you asked produced it. That sounds harsh, but it's good news too, because the question is the part you control.

Ask for feedback on a named axis and at a named stage. Three questions that reliably produce useful answers: "Read only the titles and tell me whether the argument follows." "Which slide would you cut if you had to cut one?" "What will the first question in the room be?" Each one is answerable in under a minute, each targets a specific failure, and none of them can be answered with "looks good".

Timing matters as much as phrasing. Show the titles-only storyline before you build anything, because that's when structural feedback is cheap. Show the built deck when you want to know whether it reads. Asking for structural feedback on a finished deck puts the reviewer in an awkward position and puts you in a worse one, because you now have to choose between rebuilding and defending work you know is wrong.

Receiving a redline is a separate skill. Comments arrive mixed, and the order you fix them in decides whether you finish. Sort every comment into three tiers.

Substance comments say a claim is wrong or unsupported, or that it's overstated. These are expensive and they come first, because a structural or stylistic fix applied to a wrong claim is wasted twice.

Structure comments say the order is wrong, a slide is missing, or two slides are one. These come second, and they often dissolve after the substance fixes, because a corrected claim frequently changes what belongs where.

Style comments say the font is inconsistent, the chart is cluttered, the title is a label. These come last. They're also the ones that arrive in the largest number, so an untriaged redline feels overwhelming. Five comments might be two substance, one structure and two style, and the two substance comments carry all the work.

One rule about disagreement, and it's worth holding to. You may reject a comment. You may not reject it without a word. Reply with the reason, once, and if the reviewer holds their position after that, they're probably seeing something you're not, because they have the reader's view and you have the author's.

12. The placement diagnostic, and how to skip ahead

The diagnostic is twenty items in four clusters of five. The clusters are message and audience, tool mechanics, chart grammar, and storyline structure, which map onto the branch groups you'll meet.

Each item is 5% of the whole and 20% of its cluster. Four correct out of five in a cluster, which is 80%, recommends starting that branch at its second node. The recommendation is advisory, and whether to follow it's your call, because the branches are sequential chains. Skipping the first node of the mechanics branch means the second node assumes tool habits you may have from elsewhere, and it's your judgement whether you actually have them.

Two cautions about skipping. A high score on chart grammar usually reflects reading rather than practice, and reading about chart choice and choosing correctly under time pressure are different skills that the drills separate quickly. And a high score on tool mechanics is the least predictive of the four, because knowing where a menu item lives isn't the same as having the keyboard habits that make a rebuild cheap.

The diagnostic exists to save you hours. It doesn't rank you. If a cluster comes back at 40%, that's a branch worth its full time, and a low score means the material is new to you. It doesn't mean you're slow.

13. Where the assistants help here, and where they are dangerous

An assistant is very good at some parts of this sequence and actively harmful at others, and the split is sharp enough to state as a rule.

It's useful for phrasing. Give it your takeaway and ask for six alternative wordings, and two of them will be sharper than yours. Picking which two is still your job. It's useful for the missing-question check: paste your title stack and ask what a sceptical reader would ask that the stack doesn't answer. It's useful for compression, turning a rambling paragraph into a slide's worth of words without losing a figure, provided you check the figures afterwards.

It's dangerous in three specific places. It will invent a decision when you haven't found one, producing a confident recommendation that fits the data and serves no actual choice, which is the wrong-question failure mode arriving faster than before. It will state numbers with false precision, because a plausible number and a correct number look identical in fluent prose. And it will produce a chart choice that matches the words you used rather than the message your data supports, which is a subtle failure that passes a casual read.

The rule that covers all three: an assistant may propose, and it may never decide. The takeaway is yours because you're the one who will defend it. The numbers are yours because you computed them. The decision the communication serves is yours because you asked the four questions and got answers from a person.

{
  "skill": "Pressure-testing a one-sentence takeaway and an audience profile before any slide is built",
  "use": "After you have written your own takeaway and filled in the three audience lines by hand, paste both into a chat assistant and ask it to attack them: name what the sentence claims, what evidence would falsify it, what action it implies, and which of the three audience lines is doing no work. Use the attack to sharpen your sentence; never adopt a takeaway the assistant wrote.",
  "tools": ["Chat assistants (Claude, ChatGPT, Gemini)", "A plain text editor for the takeaway, so no slide exists yet"],
  "prompt": "Here is my one-sentence takeaway: <SENTENCE>. Here is my audience profile: knows <X>; is deciding <Y>; is afraid of <Z>. Do four things. 1) State the claim my sentence makes, in your own words, and say whether a reasonable person could hold the opposite. 2) Name the specific data that would falsify it. 3) Name the action it implies for the person deciding <Y>, or say it implies none. 4) Tell me which of my three audience lines is generic enough that it would fit any audience. Do not rewrite my sentence.",
  "verify": "Every figure the sentence carries must be traced back to your own computation before the sentence ships: recompute it, and check that the rounding in the sentence matches the rounding in the exhibit that supports it. If the assistant asserts the claim is falsifiable, name the exact table or chart that would falsify it; if you cannot name one, the sentence is a slogan. If the assistant names an implied action you did not intend, that mismatch is the finding, and the fix is yours to make in the sentence, not the assistant's.",
  "diy": "You must be able to write the takeaway and the three audience lines cold, from a fact set, with no tool open. That is the gated skill and every later branch assumes it: the storyline branch orders takeaways, the chart branch draws their evidence, and neither has anything to work with if the sentence was generated rather than decided. An assistant that writes your takeaway has taken the one judgement you will be asked to defend in the room.",
  "market": "IN"
}

Common mistakes & how experts think differently

1. Opening the software first. Tempting because a blank slide feels like progress and a blank document feels like work. The expert writes the takeaway in a text file, then the titles, then opens the tool with the argument already settled. Half the hours in a deck are spent before the tool opens, and the people who look fastest at building are usually the ones who did that half properly.

1. Treating the request as the brief. "Put something together on pricing" is a request. The brief is the decision, the decider and the evidence that would change her mind, and none of the three is in the request. Novices build from requests and are surprised by the first question. Experts spend one email recovering the brief and treat that email as billable work, because it is.

1. Confusing a fact with a takeaway. "Revenue grew 42.6%" can't be disagreed with, so it can't be argued for, so it can't organise a deck. The expert's tell is the "so what" reflex: every fact gets one more sentence beginning "which means", and the deck is built from the second sentences.

1. Adapting substance to the audience when only the framing should change. The bank hears a different sentence from the board because it's deciding something different. It doesn't hear a different runway. Analysts who blur that distinction lose the ability to be believed by anyone, and the loss is permanent in a way that a bad chart is not.

1. Building one artefact for two jobs. The dense slide that a speaker reads aloud is the commonest single defect in corporate decks, and its cost is measurable: at 180 words a slide the audience finishes 34 seconds before the speaker, 41% of every slide. Experts build the reading version and the presented version, or they pick one job and accept that the other reader isn't served.

1. Leaving the title as a noun phrase. "Revenue by state" is a filing label. It's faster to write, it's never wrong, and it does no work. The expert writes the title first, before the chart, and treats a title they can't write as evidence that the analysis underneath isn't finished.

1. Confusing a full deck with a thorough one. Thirty-eight of forty-one charts show an unwillingness to choose. The expert puts the unchosen material in an appendix, which preserves everything and costs the audience nothing, and then answers the one question that reaches for it with a page number.

1. Asking for feedback with an open question. "What do you think?" gets "looks good", and then the objection that matters arrives in the meeting from someone senior. The expert asks a closed question at the stage where the answer is cheapest, usually "read only the titles and tell me whether the argument follows", asked before a single chart exists.

1. Fixing style comments first. They're the most numerous and the least expensive, so clearing them feels productive. It's the wrong order: a substance correction frequently deletes the slide whose font you just fixed. Experts triage the redline into substance, structure and style, and work strictly in that order even when the style fixes would take four minutes.

1. Believing the machine score is the standard. Seven rules passed is a floor. Nothing in the seven asks whether the claim is true or whether anyone needed it. The expert reads a clean score as permission to start the real review, not as the review's result.

Worked examples

Every figure below comes from the same computation, and the Verify block at the end of this section restates each one at full precision. All companies are synthetic composites.

Worked example 1: One quarter, three audiences, three artefacts (India, ₹)

The setting. Chandra Naturals (synthetic) sells personal care direct to consumers and through retail. The quarter ended 30 September 2026 produced revenue of ₹18.40 crore against ₹12.90 crore a year earlier and ₹16.00 crore in the previous quarter, at a gross margin of 62.5% against 58.0% a year earlier. Marketing was ₹6.20 crore, other operating costs ₹6.90 crore, working capital absorbed ₹1.90 crore and capital spending was ₹0.40 crore. Cash at quarter end was ₹9.60 crore. The working-capital facility is ₹8.00 crore sanctioned, ₹6.90 crore drawn.

The rest of the inputs the three audiences will ask for, stated once. Prospecting is 72% of marketing spend in both quarters, and marketing was ₹5.10 crore in the previous quarter; new customers were 41,500 this quarter against 39,800 last. Trade receivables stand at ₹9.40 crore against ₹8.69 crore a year earlier, in a 92-day quarter. For the facility, eligible inventory is ₹8.20 crore, eligible book debts ₹6.40 crore of the ₹9.40 crore total, creditors ₹4.10 crore, and the bank's margin is 25%; current assets are ₹27.30 crore against current liabilities of ₹13.40 crore, and the covenant is a current ratio of 1.25.

Three people want the quarter. Same data, three different jobs.

The wrong version. One deck, twenty-six slides, titled "Q2 FY27 Business Review", shown to all three audiences in three meetings across nine days. Slides 1 to 8 are revenue cuts by channel, state, product and month. Slides 9 to 14 are marketing metrics. Slides 15 to 19 are the profit and loss. Slides 20 to 24 are working capital. Slide 25 is "Key Risks", six bullets. Slide 26 is "Thank You". The founder's meeting spends its time on slide 11. The board's meeting spends its time asking what the cash position implies. The bank's relationship manager asks for the receivables ageing, which isn't in the deck.

The twenty-six slides are a symptom. The defect is that no audience profile was written, so the deck was built for the union of three audiences, which is an audience that doesn't exist.

The corrected version. Three profiles, three takeaways, three artefacts.

Walk through the arithmetic with me, one takeaway at a time. Revenue growth is 18.40 / 12.90 − 1 = 42.6357%. Gross profit is 18.40 × 0.625 = ₹11.5000 crore against 12.90 × 0.580 = ₹7.4820 crore a year earlier, so gross profit grew 53.7022% on 42.6357% revenue growth, and the gap is the 4.5 margin points. EBITDA is 11.5000 − 6.20 − 6.90 = −₹1.6000 crore; adding the ₹1.90 crore of working capital and ₹0.40 crore of capital spending gives a quarterly cash burn of ₹3.9000 crore, or ₹1.3000 crore a month, so 9.60 / 1.30 = 7.3846 months of runway. Acquisition cost is prospecting spend over new customers: 6.20 × 0.72 × 10,000,000 / 41,500 = ₹1,075.66 this quarter against 5.10 × 0.72 × 10,000,000 / 39,800 = ₹922.61 last, a rise of 16.5887%. Receivable days are 9.40 / 18.40 × 92 = 47.0000 against 8.69 / 12.90 × 92 = 61.9752 a year ago. Drawing power is (8.20 + 6.40 − 4.10) × 0.75 = ₹7.8750 crore, which exceeds the ₹6.90 crore drawn by ₹0.9750 crore. The current ratio is 27.30 / 13.40 = 2.0373.

Notice the mediums. The founder gets a conversation because the decision is small, reversible and needs one question answered in real time. The board gets a reading deck because five people need the same facts before the meeting and the argument is long enough that a listener would lose it. The bank gets a letter because a credit file is a written record and a deck isn't a document a credit committee can file. The medium changed twice across three audiences, and each change was forced by the decision rather than by preference.

The checklist to apply. Run it before you build any of the three.

1. The takeaway is a claim someone could dispute.
2. The decision is named, in the reader's own terms.
3. Every exhibit carried moves that decision.
4. The medium is the cheapest one that does the job, priced in reader-minutes.
5. The substance matches across all three artefacts, with only framing and depth changing.

Worked example 2: The status deck that should have been an email (US, $)

The setting. Meridian Health Partners (synthetic) is running a twelve-week revenue-cycle improvement project for a hospital system. The team ships a fifteen-slide status deck every Friday into a 45-minute call with nine attendees. Preparing it takes 3.5 hours a week.

The message that actually exists each week. Three things: whether the project is on schedule, what one decision the sponsor owes the team, and what one risk has changed. Everything else is history the attendees lived through.

Here's the wrong version, slide by slide, with the reason each one goes.

Fourteen of fifteen slides are cut, and the surviving one is slide 14.

The corrected version. An email of 168 words, sent Friday morning, with the deck's remaining content as an attached one-page log for anyone who wants it. The call itself survives, but once a month instead of once a week, and it's kept for the conversations that need a room.

The email's shape is fixed each week. The subject line carries the schedule status and the decision owed. The first line states the decision and the date it's needed by. The second paragraph is the one changed risk with its new mitigation, and the third is a single sentence on schedule with the variance in days. A closing line offers a fifteen-minute call to anyone who wants one.

The cost, computed. Start with the meeting. It costs 45 / 60 × 9 = 6.7500 person-hours a week, which is 81.0000 person-hours across twelve weeks, plus 42.0000 hours of preparation, for 123.0000 person-hours in total. At a stated planning rate of $95 an hour that's $11,685.00. Now the email, which costs 168 / 220 = 0.7636 minutes to read, which is 45.82 seconds; nine readers spend 0.1145 person-hours a week and 1.3745 across twelve weeks. Preparation falls to 0.5 hours a week, so 6.0000 hours in total, and the whole apparatus costs 7.3745 person-hours instead of 123.0000. The saving is 115.6255 person-hours, or $10,984.42, and the audience-time ratio between the two is 58.93 to 1. Most of the apparatus is gone, and the decision still reaches the sponsor every Friday.

What the numbers don't say. They don't say the meeting was worthless. A weekly call builds the working relationship that makes a hard conversation possible in week 9, and an email can't do that. What the numbers say is that the relationship was being bought at 81 person-hours across the quarter and that a monthly call buys most of it for a quarter of the price. The right answer here was a monthly call and eleven emails, and the argument for it's a number. That's what got it accepted.

The checklist to apply.

1. Name the decision each recurring communication serves. If it serves none, it's an information transfer.
2. Price the current medium in person-hours for the actual attendee list.
3. List every slide and write the reason it survives, not the reason it exists.
4. Keep the cheapest medium that carries the decision, and schedule the expensive one for the cases that need a room.
5. Put the cut material somewhere retrievable, and don't delete it.

Worked example 3: The pie chart that cannot show the thing being claimed (India, ₹)

The message. The three states that are largest today carried 58.0% of revenue two years ago and carry 71.0% now, a gain of 13.1 points, while total revenue grew 54.5%. Concentration is rising, and the question for the board is whether that's a distribution win or a dangerous dependency.

The data. Chandra Naturals revenue by state, FY24 and FY26, in ₹ crore.

The wrong version. A single pie of the FY26 column, six slices, a legend on the right, percentages in a data-label ring, title "Revenue by State, FY26". It's wrong three ways. It shows one period, so it can't express a claim about change, which is the entire message. Its six slices exceed what angle comparison supports: the Gujarat and Rest of India slices differ by 5.82 degrees out of 360, which no reader will resolve. And its title is a filing label that commits to nothing. That's three defects in one chart.

Two pies side by side, FY24 and FY26, look like a fix and are not. The reader now has to hold six angles in memory and compare them to six others, which is harder than the original, and the flat ₹5.20 crore of Rest of India will read as a decline because its slice shrank.

The corrected version. A slope chart. Two vertical axes, FY24 share on the left and FY26 share on the right, six labelled lines connecting them, sorted by FY26 value, with the three rising states in one accent colour and the three falling states in grey. Direct labels at both ends carry the share to one decimal, so no legend exists. The title reads: "The three largest states went from 58.0% to 71.0% of revenue in two years."

In detail, the layout runs as follows. The title sits across the top in a single line. The plot occupies the middle two thirds, with FY24 and FY26 as the only two x positions and the state names and share values at both ends of every line. A single footnote gives the totals, ₹44.00 crore and ₹68.00 crore, because a share chart without its base invites the wrong reading. One annotation sits on the Rest of India line: "flat at ₹5.20 cr, so its share fell without its revenue falling".

That annotation is the slide's most valuable element and costs nine words. Without it, a reader concludes that business outside the top five states shrank. The revenue was identical in both years.

Why a slope chart and not a sorted bar. A sorted bar of FY26 share answers "who is biggest", which is a comparison message. The message here is a change between exactly two periods across a few categories, which is what a slope chart is for. If the message had been "Maharashtra is the largest state at 32.2% of revenue", the sorted bar would be correct and the slope chart would be over-built. The form follows the message, not the data.

The checklist to apply.

1. State the message as a sentence with a verb before choosing a form.
2. Classify the message type, which here is change between two periods.
3. Check the form against the grammar table and name the rejected alternative.
4. Confirm the title states the message and carries the number.
5. Check that shares are accompanied by their base.
6. Look for the reading a viewer would make that's false, and annotate it.

Worked example 4: The dual-axis chart that hid a 6.5-fold divergence (US, $)

The setting. Ridgeline Software (synthetic) is a US business-software company. Eight quarters of results, in $ millions.

The message. Operating income grew 348.1% while revenue grew 53.6%, so operating income grew 6.5 times faster than revenue. The 11.69-point gain in operating margin splits into 5.60 points from gross margin and 6.09 points from operating leverage, which sum exactly to 11.69.

The wrong version. One chart. Revenue as a line on a left axis running 0 to 70, operating income as a line on a right axis running 0 to 12, eight quarters across the bottom, title "Revenue and Operating Income". The two lines rise together and appear to move as one. That appearance is manufactured by the axis choice. At Q1 revenue sits at 60.00% of the plot height and operating income at 21.33%; by Q8 revenue is at 92.14% and operating income at 95.60%, so the operating-income line ends above the revenue line and the reader concludes the two grew in step. The divergence carrying the whole message has been scaled out of view.

Change the right axis to 0 to 20 and the story changes without a single number changing. Sketch it on paper if you don't believe it. That's the objection to dual axes, and it isn't aesthetic. The author picks the scaling, the scaling picks the story, and the reader has no way to see that a choice was made.

The corrected version, option one: index both series. Set Q1 to 100 and plot the two series on a single axis, so the units are gone and only relative growth remains.

Two lines, one axis, direct labels at the right-hand end, no legend. The gap opens visibly from Q2 and reaches 294.55 index points by Q8. Title: "Operating income grew 348% against revenue's 54%: six times faster." A footnote states the base, Q1 = 100, revenue $42.0m, operating income $2.56m, because an indexed chart without its base is unreadable.

The corrected version, option two: small multiples. Three small line charts in a row on a shared 0 to 80% vertical scale: gross margin rising 68.0% to 73.6%, operating expense as a share of revenue falling 61.9% to 55.8%, and operating margin rising 6.1% to 17.8%. Each carries its own one-line title, and the row carries the claim: "Half the margin gain is pricing and product cost, half is operating leverage." Choose this version when the audience needs the decomposition; choose the indexed version when the audience needs the headline.

One caveat. The indexed chart flatters the smaller series by construction. Operating income started at $2.56 million, so a $8.91 million improvement reads as 348%. Say so in the footnote. A chart that's technically correct and predictably over-read is still a chart that misleads, and the fix costs one sentence.

The checklist to apply.

1. Name the message and the number that carries it.
2. If two series have different units, index them or split them, and never reach for a second axis.
3. State the base of any indexed series.
4. Check whether the form flatters one series, and annotate it if it does.
5. Order by something intrinsic, which here is time.
6. Confirm that the decomposition sums exactly, 5.60 + 6.09 = 11.69, before any of it goes on a slide.

Worked example 5: The pricing request with no decision in it (India, ₹)

The request. "Can you put together something on our pricing for Thursday?"

The wrong version. Twenty-two slides. Competitor price tables for eleven brands across four pack sizes, a price-index chart, a margin waterfall, a category growth chart, three slides on premiumisation, and a closing slide with four options and no recommendation. Build time about eleven hours. In the meeting, the first question is "so do we take the 200 ml to ₹399 in April or not?", and nothing in the twenty-two slides answers it, because the analyst never asked.

Recovering the decision. The four questions took about four minutes to write into one email. After Thursday the April price list is locked. The sales head owns it and will be in the room. What would change her mind is evidence about how much volume a ₹50 increase would cost, and doing nothing means holding ₹349 through the festive quarter.

That converts the request into a single question with a single number attached: how much volume can the ₹399 price afford to lose before it's worse than ₹349?

The corrected version. One page. One chart, one table, three sentences, and a recommendation.

Now the arithmetic, starting from today's price. At ₹349 with a 62.5% gross margin, variable cost is 349 × 0.375 = ₹130.875 a unit and contribution is 349 − 130.875 = ₹218.125. At ₹399 the same variable cost gives contribution of 399 − 130.875 = ₹268.125. Breakeven volume is the ratio of the two contributions, 218.125 / 268.125 = 0.813520, so volume can fall 18.6480% before the higher price is worse. The price rise itself is 399 / 349 − 1 = 14.3266%.

The evidence about what volume will actually do comes from two past moves on the same product. From ₹299 to ₹329 was a 10.0334% price rise and volume fell 6.20%, an implied elasticity of −0.6179 measured off the starting price. From ₹329 to ₹349 was a 6.0790% rise and volume fell 3.40%, an elasticity of −0.5593 on the same basis. The midpoint convention would give slightly smaller numbers, so say which you used. The average is −0.5886, which predicts a volume fall of −0.5886 × 14.3266% = −8.4329% for the proposed move: 130,025 units against 142,000 today. Contribution rises from 142,000 × 218.125 = ₹3.0974 crore to 130,025 × 268.125 = ₹3.4863 crore, a gain of ₹0.3889 crore, or 12.56%, in a quarter.

The slide's chart is one horizontal bar with two markers on it: the breakeven volume fall at 18.65% and the predicted volume fall at 8.43%, with the 10.22 points between them shaded and labelled "headroom". Title: "Volume can fall 18.6% before ₹399 is worse than ₹349; history says it falls about 8%."

The table beneath it's four rows: price, contribution per unit, units, quarterly contribution, for the two scenarios. The three sentences state the elasticity evidence, the assumption that past moves predict this one, and the one condition that would break it, which is a competitor holding price while Chandra Naturals moves.

What makes this the right artefact. The decision is binary and the evidence is one number against another number. That's a one-pager, and the presented deck it replaced would have cost the room forty minutes to reach the same place. Total build time for the one-pager is about two and a half hours, of which the four questions were four minutes and saved roughly eight hours.

The checklist to apply.

1. Ask the four questions before building anything.
2. Restate the decision in one sentence and get it confirmed.
3. Compute the breakeven before gathering evidence, because it tells you what evidence would matter.
4. Name the assumption that carries the recommendation, here that two past moves predict a third.
5. Name the one condition that would break it.
6. Choose the medium after the analysis is scoped, not before.

Worked example 6: The board pack that had to be two artefacts (US, $)

The setting. Ridgeline Software's quarterly board meeting has a thirty-minute slot. The chief financial officer has one argument to land: operating income grew 348.1% against revenue's 53.6%, so the company should raise the sales-hiring plan for next year rather than protect margin.

The wrong version. Fourteen slides at about 180 words each, projected, read aloud. The arithmetic of that choice: 180 words takes 49.09 seconds to read silently at 220 words a minute and 83.08 seconds to say aloud at 130. The audience finishes 33.99 seconds early on every slide, which is 40.91% of the time the speaker is talking. Across fourteen slides the deck takes 19.38 minutes to say and 11.45 minutes to read, and the board spends nearly eight minutes of a thirty-minute slot ahead of the speaker. Directors read ahead, form objections privately, and arrive at question time having already decided.

The corrected version is two artefacts built from one argument.

The reading pack, circulated 48 hours ahead: nine pages at about 340 words each, 3,060 words in total, read in 13.91 minutes by any director who chooses to. It carries the full argument in prose with the exhibits inline, including the indexed chart from the previous example and the margin decomposition. It survives forwarding to a director who missed the meeting, because nothing in it depends on a speaker.

The presented deck, shown in the room: fourteen slides at about 31 words each, 434 words on screen. In a thirty-minute slot at 130 words a minute the speaker says roughly 3,900 words, so about 11% of the argument is in front of the eye and the rest is in the ear. Each slide carries an action title and one exhibit. The speaker notes carry the rest.

The third artefact nobody plans for. After the meeting, someone forwards the presented deck to a director who was travelling. Fourteen slides of 31 words each will tell her nothing. The rule that solves this: whichever artefact is likely to be forwarded gets speaker notes dense enough to reconstruct the argument, or the forwarding is done with the reading pack instead. Deciding which, in advance, belongs to choosing the medium and shouldn't be left as an afterthought.

The cost of building both. Roughly three extra hours against the sixteen-hour budget. The alternative is one artefact that costs the board 40.91% of its attention on every slide, and a decision taken by directors who read ahead and stopped listening. Three hours is the cheaper option and it isn't close.

The checklist to apply.

1. Decide whether the artefact is read or presented, and never both.
2. Size it accordingly: roughly 30 words a slide presented, 300 or more per page read.
3. Compute the read-versus-speak gap if the slide is dense, and treat anything above about 60 words a slide as a warning.
4. Decide in advance what gets forwarded, and make that artefact self-sufficient.
5. Confirm the substance is identical across both, with only depth and framing differing.

Verify

Every figure printed above, at full precision, in the order it appears.

Revenue growth 18.40/12.90 − 1 = 0.42635659 (42.6357%). Quarter on quarter 18.40/16.00 − 1 = 0.15 (15.0000%). Gross profit 18.40 × 0.625 = 11.50000 and 12.90 × 0.580 = 7.48200; gross profit growth 0.53702219 (53.7022%); margin change 4.5000 points. EBITDA 11.50000 − 6.20 − 6.90 = −1.60000. Quarterly burn 1.60000 + 1.90 + 0.40 = 3.90000; monthly 1.30000; runway 9.60/1.30 = 7.38462 months. Acquisition cost 6.20 × 0.72 × 10^7 / 41,500 = 1,075.66265 and 5.10 × 0.72 × 10^7 / 39,800 = 922.61307; change 0.16588668 (16.5887%). Receivable days 9.40/18.40 × 92 = 47.00000 and 8.69/12.90 × 92 = 61.97519; improvement 14.97519 days. Drawing power (8.20 + 6.40 − 4.10) × 0.75 = 7.87500; headroom 7.87500 − 6.90 = 0.97500. Current ratio 27.30/13.40 = 2.03731.

Meeting cost 0.75 × 9 = 6.75000 person-hours weekly, 81.00000 over twelve weeks, plus 42.00000 preparation, total 123.00000, at $95 giving $11,685.00. Email 168/220 = 0.76364 minutes, 45.81818 seconds; nine readers 0.11455 person-hours weekly, 1.37455 over twelve weeks; new preparation 6.00000; new total 7.37455; saving 115.62545 person-hours and $10,984.42; ratio 81.00000/1.37455 = 58.92857.

State shares FY24 26.36364, 18.86364, 12.72727, 16.81818, 13.40909, 11.81818, summing to 100.00000. FY26 32.20588, 22.50000, 16.32353, 12.05882, 9.26471, 7.64706, summing to 100.00000. Top three 57.95455 and 71.02941; change 13.07487 points. Total growth 68.00/44.00 − 1 = 0.54545455 (54.5455%). FY26 slice angles 115.94, 81.00, 58.76, 43.41, 33.35, 27.53 degrees; smallest pair differ by 5.82 degrees.

Ridgeline gross profit 28.56000, 31.83180, 35.17920, 38.12400, 40.89800, 43.38000, 45.55200, 47.47200; operating income 2.56000, 3.83180, 5.17920, 6.52400, 7.89800, 9.18000, 10.35200, 11.47200. Revenue growth 64.5/42.0 − 1 = 0.53571429; operating income growth 11.47200/2.56000 − 1 = 3.48125; ratio 6.49833. Operating margin 6.09524% to 17.78605%, a gain of 11.69081 points, of which gross margin contributes 5.60000 and the operating-expense ratio 61.90476 − 55.81395 = 6.09081; the two sum to 11.69081. Index values as tabled, with 448.12500 − 153.57143 = 294.55357 index points of gap at Q8.

Pricing: variable cost 130.87500; contribution 218.12500 and 268.12500; breakeven ratio 0.81352, fall 18.64802%; price rise 14.32665%. Elasticities −0.61793 and −0.55930, average −0.58862; predicted change −8.43290%; units 130,025; contribution ₹3.09738 crore and ₹3.48630 crore; gain ₹0.38892 crore, 12.55645%; headroom 10.21512 points.

Reading and speaking: 180/220 × 60 = 49.09091 seconds and 180/130 × 60 = 83.07692 seconds; gap 33.98601 seconds, 40.90909% of the spoken time. Fourteen slides: 11.45455 minutes read, 19.38462 minutes spoken. Reading pack 9 × 340 = 3,060 words, 13.90909 minutes. Presented deck 14 × 31 = 434 words against 30 × 130 = 3,900 spoken, 11.13%.

Blended scores: 0.6 × 0.86 + 0.4 × 0.72 = 0.80400; 0.6 × 0.95 + 0.4 × 0.60 = 0.81000; (0.80 − 0.4 × 0.70)/0.6 = 0.86667. Hours: 8.0/16.0 = 50.00% before the tool, 5.5/16.0 = 34.375% inside it. Data dump: 30 × 60/34 = 52.94118 seconds a slide and 30 × 60/41 = 43.90244 seconds a chart; 41/3 = 13.66667 charts a claim.

Practice set

Try each one before you read the solution. A calculator is assumed. Numeric answers within 2% of the stated figure count as correct unless a problem says otherwise, and where a problem asks for a sentence, the test is whether a reader could disagree with it.

P1 (guided). Write the takeaway (India). Chandra Naturals reports revenue of ₹18.40 crore for the quarter against ₹12.90 crore a year earlier, gross margin of 62.5% against 58.0%, cash of ₹9.60 crore and a monthly burn of ₹1.30 crore. The board is deciding when to start the Series B. Write the one-sentence takeaway and show the two computations it rests on.

Solution. Revenue growth is 18.40/12.90 − 1 = 42.6357%, and runway is 9.60/1.30 = 7.3846 months. A takeaway that passes all three tests: "Revenue grew 42.6% and gross margin is up 4.5 points, but 7.4 months of runway means the raise starts in November, not February." It's a claim, because a director could argue for February. It's falsifiable, because a lower burn rate or a bridge facility would break it. And it implies an action, which is starting the process in November. A common wrong answer is "Growth is strong but cash is tight", which fails all three: nobody can disagree with it, no number could refute it, and no date follows from it.

P2 (guided). Recover the decision (US). A partner asks: "Can you look at our customer churn and come back to me next week?" Write the four questions you would ask, and state what each answer would change about what you build.

Solution. What happens after next week? If the answer is a pricing committee, you're building for a price decision; if it's a board update, you're building for information. Who else will see it? If the head of customer success is present, the analysis needs to survive someone who knows the accounts individually, which raises the evidence bar. What would change your mind? If the answer is "evidence that churn is concentrated in one segment", you now know the cut that matters and can ignore the other eleven. What happens if we do nothing? If churn costs $2.3 million a year at the current rate, that number sizes how large a recommendation has to be to be worth the disruption. The four answers together turn a week of work into two days of the right work.

P3 (guided). Price the medium (India). A team proposes a 40-slide deck at about 95 words a slide for six readers. The alternative is a two-page memo of 1,100 words. Reading runs at 220 words a minute. Compute the reader cost of each and state the choice you would defend.

Solution. The deck is 40 × 95 = 3,800 words, read in 3,800/220 = 17.2727 minutes, which is 103.6364 person-minutes across six readers. The memo is 1,100/220 = 5.0000 minutes, or 30.0000 person-minutes, saving 73.6364 person-minutes. The memo wins unless one of the three reasons for the expensive medium applies: the room, the length of the argument, or unanticipated questions. Six readers reading alone isn't a room, and 3,800 words is a padded argument rather than a long one. Defend the memo, and offer a meeting if it raises questions.

P4. Three profiles, one quarter (US). Ridgeline Software's Q8 results are revenue $64.5m, operating income $11.47m, operating margin 17.79%. Write the "decides" and "fears" lines for three readers: the board, the head of sales, and a lender reviewing a revolving credit facility. Then write one takeaway each.

Solution. The board decides next year's hiring plan and fears over-hiring into a slowdown; its takeaway is that operating income grew 348.1% against revenue's 53.6%, so the margin can fund a larger sales plan without breaching the operating-margin floor. The head of sales decides quota and territory design and fears a plan she can't hit; her takeaway is that with operating margin at 17.79% against 6.10% two years ago, the company can carry a longer ramp for new hires than the current three-month quota schedule assumes. The lender decides whether to renew the revolver and fears a covenant breach; its takeaway is that operating income has risen in all eight quarters to $11.47m, so coverage has improved every period since the facility was signed. Same eight quarters, three decisions, three sentences.

P5. Choose the form from the message (India). Chandra Naturals' channel mix, in ₹ crore: marketplaces 19.80 to 22.40, own website 8.20 to 12.90, modern trade 6.10 to 16.30, general trade 6.60 to 7.50, quick commerce 3.30 to 8.90, FY24 to FY26. The message is "marketplaces fell from 45.0% to 32.9% of revenue while quick commerce and modern trade grew about 170% each". Name the chart form, the rejected alternative, and the rule that decided it. Compute the two growth figures and the share change.

Solution. Totals are 44.00 and 68.00, which tie to the state table, so the two views are consistent. Marketplaces' share is 19.80/44.00 = 45.000% and 22.40/68.00 = 32.941%, a fall of 12.059 points. Quick commerce grew 8.90/3.30 − 1 = 169.697% and modern trade 16.30/6.10 − 1 = 167.213%. The message is a change between exactly two periods across five categories, so a slope chart of share carries it, with the two rising channels accented. The rejected alternative is a 100% stacked bar for the two years, which shows the composition but makes the individual channel movements hard to read because only the bottom segment shares a baseline. The rule: change between two periods across a few categories goes to a slope chart, and a stacked bar is for composition at a point rather than movement between two.

P6. Repair a dual axis (US). Ridgeline's revenue runs 42.0 to 64.5 and operating income 2.56 to 11.47 across eight quarters. Someone has put them on two axes. Produce the indexed series for both, state the Q8 gap in index points, and give the action title.

Solution. Indexed to Q1 = 100, revenue runs 100.00, 110.00, 120.00, 128.57, 136.19, 142.86, 148.57, 153.57 and operating income runs 100.00, 149.68, 202.31, 254.84, 308.52, 358.59, 404.38, 448.13. The Q8 gap is 448.125 − 153.571 = 294.55 index points. Title: "Operating income grew 348% against revenue's 54%: six times faster." The footnote must give the base, Q1 = 100 on revenue of $42.0m and operating income of $2.56m, because 348% on a $2.56m base is a smaller achievement than the percentage suggests, and the footnote should say so.

P7. Cut twelve slides to four (India). A twelve-slide update contains: title, agenda, company overview, quarterly revenue, revenue by state, revenue by channel, gross margin bridge, marketing spend, acquisition cost, cash and runway, risks, thank you. The decision is when to start the Series B. Name the four slides that survive and state what each cut costs.

Solution. Survivors: quarterly revenue with growth (the credibility of the trajectory), gross margin bridge (the quality of that growth), cash and runway (the constraint that drives the timing), and a recommendation slide stating November with its two conditions. That's a 66.7% cut. Now the costs. Dropping revenue by state and by channel loses the concentration story, which belongs in an appendix because a director may ask for it. Dropping the company overview costs nothing with a board that already knows the company. Dropping marketing spend and acquisition cost loses operational detail that's the founder's decision rather than the board's, and it belongs in the ops review. Dropping risks is the only uncomfortable cut, so fold the two live risks into the recommendation slide as its conditions instead of deleting them. Agenda, title and thank-you're furniture.

P8. Compute the collision (US). An eighteen-slide deck averages 145 words a slide and will be presented. Compute the silent reading time, the spoken time, and the gap, at 220 and 130 words a minute. Then say what you would change.

Solution. Total words are 18 × 145 = 2,610. Reading takes 2,610/220 = 11.8636 minutes; speaking takes 2,610/130 = 20.0769 minutes; the gap is 8.2133 minutes, which is 40.91% of the spoken time, the same share as any dense slide because the ratio of the two rates is fixed. The change: rebuild as a presented deck at about 30 words a slide with the argument in the speaker notes, or accept that it's a reading deck and stop presenting it. Halving the word count leaves the ratio where it's and only shortens both sides.

P9. Title repair (India). Rewrite these as action titles, using the figures given: (a) "Revenue by State" for a chart where the top three states moved from 58.0% to 71.0% of revenue; (b) "Marketing Update" for a quarter where acquisition cost went from ₹923 to ₹1,076; (c) "Q2 Financials" where gross profit grew 53.7% on 42.6% revenue growth. Then state the test each rewrite passes.

Solution. (a) "The three largest states went from 58.0% to 71.0% of revenue in two years." (b) "Acquisition cost rose 16.6% in one quarter, to ₹1,076." (c) "Gross profit grew 53.7% on 42.6% revenue growth, because margin added 4.5 points." Each is a sentence with a verb, each carries the number that makes it checkable, and each could be contradicted by someone holding different data. The check on (b): 1,075.66/922.61 − 1 = 16.59%, so 16.6% is the correct rounding and "sharply" wouldn't be.

P10. Price a data dump (US). A 47-slide deck averaging 120 words a slide is circulated to eleven people. Compute the reader cost in person-hours at 220 words a minute, and state the two-line diagnosis.

Solution. Words total 47 × 120 = 5,640, read in 5,640/220 = 25.6364 minutes each, which is 282.0000 person-minutes or 4.7000 person-hours across eleven readers. The diagnosis: the deck has no takeaway, because a deck with a takeaway would have kept the slides that support it and moved the rest to an appendix; and the author is protecting themselves against the accusation of having missed something, which is a reasonable fear with an unreasonable remedy. The repair is upstream, not editorial.

P11. Triage a redline (India). Five comments arrive on a nine-slide deck.

• (a) "slide 4's margin number does not match slide 7"
• (b) "the font on slide 2 is different"
• (c) "slides 5 and 6 make the same point"
• (d) "the recommendation assumes the competitor holds price, which we have no evidence for"
• (e) "the chart on slide 8 has gridlines you do not need"

Sort them and order the work. Then say why the order matters.

Solution. Substance: (a) and (d). Structure: (c). Style: (b) and (e). Work in that order. (a) is a contradiction inside the deck, which destroys trust in every other number, so it's first. (d) is an unsupported assumption carrying the recommendation, and the fix may change what the deck recommends. (c) is a merge, and it may dissolve on its own once (d) is resolved, because one of the two slides may no longer be needed. (b) and (e) take four minutes and go last. Fixing (b) and (e) first feels productive and is the reason redlines take three passes instead of one: you would have spent the four minutes on a slide that (d) later deleted.

P12. Diagnose the wrong question (US). A team spent 22 person-hours building a 26-slide deck, then presented it for 60 minutes to seven people. In the meeting it emerged that only two slides addressed the question the sponsor cared about. Compute the total person-hours committed and the share of the deck that was on target, then name the failure and the four-minute fix.

Solution. Build 22.0 person-hours plus 60/60 × 7 = 7.0 person-hours of meeting is 29.0 person-hours. On-target slides are 2/26 = 7.7%. The failure is the third arrival mode, a deck that answers a question nobody asked, and it's a briefing failure rather than an analysis failure. The fix is the four questions asked in one email before any work started: what happens after the meeting, who else is in the room, what would change your mind, and what happens if we do nothing. Four minutes against 29 person-hours is the highest-return four minutes in this branch.

Applied mini-project

One page, written before anything is built. It's short on purpose, and you should budget about 45 minutes for it. Take one analysis you have already completed, a business teardown, a case answer, a valuation, a market study, and produce the page that should have preceded it. If you have no prior analysis, use the fallback brief given after the five steps.

Step 1 names the reader and the decision, and it takes about ten minutes. One named role, not a category. "The head of category at a mid-size Indian personal-care brand" is a role; "management" is not. Write the three profile lines: what this reader already knows about your subject, what they're deciding, and what they're afraid of. If you can't fill the decision line from what you were told, write the four recovery questions you would send and mark the line as pending. A decision marked as pending scores better than an invented one.

Step 2 writes the takeaway, in another ten minutes. One sentence. Then, underneath it, write the three tests and your answer to each: what claim is being made, what data would falsify it, and what action follows. Then write the strongest sentence someone could say against it. If you can't write the counter-sentence, the takeaway isn't a claim yet.

Step 3 chooses the medium and prices it, ten minutes again. Name your choice and a single rejected alternative. Price both in reader-minutes for your actual audience size, using 220 words a minute for reading and the meeting length for anything presented. State which of the three reasons for an expensive medium applies, or state that none does and that you chose the cheap one.

Step 4 lists the evidence and what it excludes, in the last of the ten-minute blocks. Three to six items, each one a claim you can support, each with the number that supports it. Then list what you're deliberately leaving out and where it goes. An appendix line is a decision. Treat it as more than a dumping ground, and name what question each excluded item would answer if asked.

Step 5 is the one-line self-check, and five minutes covers it. Write the first question you expect from your named reader. If your one page doesn't answer it, that's the finding, and fixing it now costs five minutes. Later it costs a rebuild.

A note on the artefact itself. Deliver the page as a document rather than as a slide, because the point of the exercise is that the artefact choice comes last. If you export to PDF, keep it under the size your mail system will carry: most corporate systems reject attachments above roughly 20 to 25 MB, and the ceiling is set per organisation (checked September 2026; confirm with your own administrator instead of assuming).

The fallback brief. Chandra Naturals is considering closing its general-trade distribution, which is ₹7.50 crore of FY26 revenue growing 13.6%, at a gross margin 11 points below the company average, to redirect the working capital into quick commerce, which is ₹8.90 crore growing 169.7%. The founder wants a view by Friday. Nothing else has been told to you. Run the five steps on that.

The rubric runs to twenty points, passing at 16 with no zero on any row.

Any row where the evidence is missing scores zero regardless of how well the page reads. The page is graded on whether it would have made the analysis land, not on its prose.

Reading & resources

Core (do these).

• Barbara Minto, The Pyramid Principle: the structural spine of every consulting deliverable, and the source of the answer-first habit that the storytelling branch builds on. Read the first part on the pyramid and the introduction structures; the later material on deduction and induction repays a second pass. [Paid] [Intermediate]
• Cole Nussbaumer Knaflic, Storytelling with Data: the most useful single book for the gap this track fills, and the closest published treatment of the chart-choice and declutter grammar taught here. Its exercise volume, Storytelling with Data: Let's Practice, is the better buy if you want reps. [Paid] [Beginner]
• Gene Zelazny, Say It With Charts: the original message-to-chart-form mapping, written for a consulting audience and still the tightest statement of the idea that the message chooses the chart. [Paid] [Beginner]
• Stephen Few, Show Me the Numbers: the reference treatment of table and graph design, and the place to go when you want the reason behind a rule rather than the rule. [Paid] [Intermediate]
• Edward Tufte, The Visual Display of Quantitative Information: the origin of the data-ink argument and the lie factor. Read it for the standard of care rather than for the specific prescriptions, some of which the field has since argued with. [Paid] [Intermediate]

Going deeper.

• Alberto Cairo, How Charts Lie: the best short treatment of chart deception, including axis manipulation and the dual-axis problem ruled out above. [Paid] [Beginner]
• Nancy Duarte, Resonate and slide:ology: the audience-first case, made from a design rather than an analytical direction. Resonate is the one to read for the audience profile; slide:ology for the artefact. [Paid] [Beginner]
• Garr Reynolds, Presentation Zen: the argument for restraint on a presented slide, and the clearest published statement of why a dense projected slide competes with its own speaker. [Paid] [Beginner]
• Jean-luc Doumont, Trees, Maps, and Theorems: a compressed, opinionated treatment of professional communication that covers documents, slides and speech in one frame. Short, dense, and worth rereading. [Paid] [Advanced]
• The US Securities and Exchange Commission, A Plain English Handbook: free from sec.gov, written for disclosure documents, and the single cheapest improvement available to anyone whose prose is dense. [Free] [Beginner]

Standards and tools (free).

• W3C Web Content Accessibility Guidelines: at Level AA the contrast minimum for normal-size text is 4.5 to 1, relaxing to 3 to 1 for large text and rising to 7 to 1 at Level AAA, and the 4.5 figure is the one the accessibility rules in the mechanics branch are written against (WCAG 2.2, checked September 2026; confirm, since the guideline version advances and the ratios can move with it).
• Your own presentation software's defaults: the widescreen slide in current desktop PowerPoint is 13.333 by 7.5 inches at a 16:9 ratio (Microsoft 365 desktop, checked September 2026; confirm against your own installation, since the default follows the template and the release). Google Slides and Keynote use the same ratio with different measurements. The mechanics branch teaches one of these tools in depth and states the equivalents in the others.

Do this, not just read. Take the last deck you built or received and read only its titles, in order, with the slides hidden. Write down the argument they make. If they make none, you have just measured the gap this whole track exists to close, on a real artefact, in four minutes.

Teach it back & journal

Feynman prompt. Explain to a competent colleague who has never thought about this why two decks with identical data can have opposite value, without using the words design, professional or polish. Use the arithmetic. A projected slide of 180 words takes 49 seconds to read and 83 to say, so the audience spends 41% of every slide waiting. A 45-minute weekly meeting for nine people costs 81 person-hours a quarter, against 1.4 for the email that carries the same decision. And 22 person-hours of building plus a 60-minute meeting for seven is 29 person-hours spent on a deck where 7.7% of the slides touched the real question. Then make the harder point: none of those numbers is about slides. Each one is about a decision that was or wasn't identified before the work started. If your colleague can tell you, unprompted, what the four recovery questions are for, you have taught the idea rather than the arithmetic.

Journal prompt. Find the last deck you built, or the last one somebody sent you, and run three checks on it. First, hide the slides and read only the titles in order: does an argument follow, or is it a table of contents? Second, write down the decision that deck was serving, in the words the decider would use, and note whether you can. Third, price the medium: how many people, how many minutes, and what would the same content have cost as a memo? Write down the three answers. Then write a paragraph on which of the three arrival failure modes is yours, with the evidence from your own artefact rather than from a description. Data dumpers are usually analysts with real depth who are afraid of being caught not knowing something. Empty-deck builders usually have design instincts and no ownership of the argument. Wrong-question builders are usually eager and under-briefed. Being wrong about which one you're is common and costs a branch of frustration, so trust the artefact over your self-image. Finish with one habit you'll install this week, and make it small enough to survive a bad Tuesday: a text file open before the software, one sentence written in it, and a rule that the software stays shut until the sentence could be argued with.

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.