The Board Deck and the Investor Pitch

Presentation track, branch pr-archetypes (Archetypes) · Node PR5.02 · ~12 focused hours · Mastery gate ≥ 85%

A board of eight people meets for two hours, four times a year. That is eight hours of the most expensive attention a company can buy, and a chief executive who mails sixty-eight pages into it has already spent most of that attention before anybody sits down.

Nilgiri Leaf Consumer Brands (synthetic, a composite of Indian direct-to-consumer tea and wellness brands at roughly ₹285 crore of annualised revenue) mailed sixty-eight. The September pack opened on a page titled "Q2 FY27 Performance Update" and reported net revenue of ₹71.20 crore against ₹58.40 crore a year earlier, growth of 21.92%. Every figure on that page was correct. On page 44 of the appendix sat the contribution margin, down from 24.60% to 19.80%, which means contribution in rupees of ₹14.0976 crore against ₹14.3664 crore a year earlier. Revenue was up 21.92% and contribution was down 1.87%, and nobody read page 44. The meeting approved an increase in the performance-marketing cap.

Two quarters later the same board approved an emergency bridge on a Sunday call. The information needed to prevent that was in the September pack, in the correct units, computed correctly, on a page that had been produced by a competent analyst. It failed because it had been filed rather than argued.

Nothing about that is a design problem. The fonts matched. The chart on page 44 was well drawn. What went wrong is upstream of design: the pack was built to the shape of an update when the audience was a body whose entire function is to take decisions. A board hires and fires the chief executive, approves capital above a threshold, signs the accounts, and carries personal liability for the risks it accepted. Every one of those is a choice, made by people who were not in the building last week and will not be in it next week.

Two archetypes are taught here and they share one property that makes them worth teaching together. In both, the reader is deciding something, and in both the reader has far less time than you have material. They differ in the direction of trust. A board has already committed; its members own the outcome and want to know what has moved and what needs deciding. An investor has committed nothing; the default is no, and every slide is being read for a reason to stop reading. That difference changes the order of the pages, the density of each page, and where the bad news goes, and it changes almost nothing about the arithmetic underneath.

The honest scope is narrow and worth stating. Governance obligations, committee structure and the legal duties of directors belong to the finance and ethics material and are not rebuilt here. Market-sizing method is owned elsewhere too, and the treatment below is about what a sizing slide may claim, not about how to size. What is taught here is the craft: which page goes first, and what a decision looks like when it is written for a body that must vote on it. The rest is density, meaning how much a read page can carry before it stops being read, and the arithmetic that ties an ask to a milestone so a reader can check it rather than admire it.

By the end you will have rebuilt one sixty-eight-page pack into a fourteen-page body with a forty-one-page indexed appendix, and one twelve-slide seed pitch whose market slide survives being checked. Both rebuilds use data that ties.

Learning objectives

You can:

1. Convert a board update into a decisions-first pack, stating each decision as a question the board can answer with a vote, with the option set, the recommendation, and the number the choice turns on.
2. Compute and present the figure that contradicts a headline, such as revenue growing 21.92% while contribution falls 1.87%, and place it where a director cannot miss it.
3. Size a read pack against the meeting it serves, in pages, words per page and reading minutes, and defend the split between body and appendix.
4. Build a board dashboard from a decision test, cutting tiles that no decision consumes, and justify every metric that survives.
5. Frame bad news for a board: where it goes in the pack, what must travel with it, and why the version that arrives late costs more than the one that arrives ugly.
6. Structure an investor pitch across the seven questions every investor asks, and map each to one slide with a stated job.
7. Test a market slide by rebuilding it bottom up, comparing the top-down claim to the built figure and reporting the ratio rather than hiding it.
8. Tie an ask to a milestone, decomposing the required net new revenue into the part the existing base supplies and the part new sales must supply, then checking that against demonstrated capacity.
9. Apply the two shipped templates to a live brief, and state for each borrowed page what the page is for and what would justify deleting it.
10. Choose between archetypes under pressure, including the three cases where the obvious choice is wrong, and say what changes when the same content must go to the other audience.

Prerequisites & connections

Builds on. The consulting deck and the project readout. They own the action-title body page, the executive summary written last and the appendix contract, and everything below assumes them without teaching them again. You should already be able to write a one-sentence takeaway, profile an audience on three axes, and choose a chart form from a message. You'll be working with contribution margins, break-evens, payback periods and recurring-revenue retention. It's all school algebra, but you'll move faster if the finance vocabulary is already familiar.

Feeds forward. The equity-research deck and the model walkthrough carry the same density rules into a sell-side artefact, where compliance furniture is mandatory. The delivery branch owns the Q&A that a decisions page provokes, and that's the part of a board meeting the deck can't control. In the capstone you run the whole sequence, from an analysis to a built deck, and either archetype below is a legitimate choice there.

Deliberate non-overlaps. Six neighbours own material that touches this, and none of it is rebuilt here. From the consulting-deck node come the action-title body page, the executive summary page and the appendix index, all of which are used below as known craft. The research-deck node owns the thesis page, the valuation bridge, the scenario page and the compliance furniture, so nothing below teaches a rating or a target price. Market-sizing method, meaning the top-down and bottom-up construction and the sanity triangulation, belongs to the capital-markets orientation node and to the consulting sizing node. All you'll get below is what a sizing slide may claim and how a reader checks it. Chart-form choice belongs to the chart-craft branch and appears here only as applied grammar. Storyline has its own branch, which owns the pyramid, the horizontal and vertical logic and the title stack. Governance duties, director liability and the disclosure perimeter belong to the ethics and standards node. Where a page below needs one of those, it names the craft and moves on.

Core teaching content

4. The pre-read, and the meeting that assumes it

A pre-read is a pack circulated far enough ahead that directors can read it. A meeting that assumes the pre-read doesn't present it. Those two belong together, and it's usually the second that collapses.

Circulate 72 hours ahead as a working default, not 24. A director with a day job needs to find eleven minutes plus thinking time, and if that has to happen the evening before a morning meeting, it won't happen. Where directors are spread across several time zones, 96 hours is safer. An extra day costs you a day of freshness in your figures, and that's almost always the cheaper trade.

Then hold the meeting as though the reading happened, because otherwise it never will. A chief executive who opens by walking through the pack has told every director that reading it was optional, and by the third meeting nobody reads. Open instead with a two-minute frame: what has changed since the pack was mailed, which decision will take most of the meeting, and one question put to the room. Then go straight to the first decision.

Directors who didn't read will be exposed by that. It feels harsh, and it's meant to. A board where two members are unprepared is taking decisions with a quorum of the informed, and the chair needs to see it. Managing that is the chair's job. Yours is to build a pack that makes it visible.

Two mechanics make the pre-read work. First, version the pack visibly, with a version number and a date on the cover and on every page footer. A director reading v2 while the room discusses v3 produces an argument that looks like disagreement and is really a document control failure. Second, mark the confidentiality tier on the cover in the company's own scheme, with the distribution list named. A board pack that leaks was usually forwarded innocently, because nothing on it said not to.

Late-breaking material gets its own one-page addendum with its own date, mailed separately. Never swap the original file without saying so. A pack that changes without notice is the fastest way to destroy the trust the whole artefact depends on.

7. Bad news: where it goes and how it is framed

Bad news goes early, in the body, with a name attached. You'll feel a strong instinct to bury it in the appendix, and it's always wrong. You don't need ethics to see why. A board that discovers bad news late loses confidence in every other number in the pack, including the ones that were fine, and that loss is permanent in a way the original bad news never was.

Take Nilgiri Leaf's contribution margin. It had fallen in each of eight consecutive quarters, and the fall accelerated in the last two, from 22.70% to 21.10% to 19.80%. That's the most important fact about the company, and it sat on appendix page 44 under a title that named a topic. In the rebuild it's body page 3, titled with the claim, and the page carries the bridge that explains it.

Four things travel with bad news and none of them is optional.

The size, in the unit the board thinks in. "Margins are under pressure" won't do. Write "contribution fell ₹0.2688 crore against a year earlier while revenue rose ₹12.80 crore, because contribution margin fell 4.80 points". A board that hears the vague version will ask how bad, and the meeting then spends its expensive minutes extracting a number that could have been printed.

The cause, decomposed. Those 4.80 points break into five bridge steps: price and mix added 1.4 points, leaf and packaging cost took 3.1, freight and fulfilment took 1.2, quick-commerce trade spend took 2.5, and other variable lines added 0.6. Those sum to −4.80 and reconcile 24.60% to 19.80% with nothing left over. Two lines, leaf cost and quick-commerce trade spend, account for 5.6 points of drag between them. That's more than the whole net fall, because price and mix plus other variable added 2.0 points back.

So say "two cost lines took 5.6 points and the rest of the business gave 2.0 back". It's true, it's more useful than a net number, and it points at where the fix has to come from.

The action already taken, with a date. Boards forgive problems. They don't forgive passivity. If the trade-spend line was renegotiated in August, say so, and say what it's expected to recover.

And what would make it worse, stated as a trigger. "If quick commerce goes from 30% to 40% of revenue at the current trade-spend rate, contribution margin falls a further 2.5 points" turns an anxiety into a number the board can monitor.

Framing has one hard rule. It may set the order and the emphasis, and it may never change the number or soften the claim. "Contribution margin declined 4.80 points, of which 3.1 is input cost we have now hedged through March" is fair framing. "Contribution margin was broadly stable excluding one-off channel investments" is a different assertion, and if the channel investment recurs every quarter it's a false one. Directors have seen the second version many times, and they price it in. So the analyst who writes the first version is the one trusted with harder material later.

There's a timing rule too. Bad news that crosses a materiality threshold doesn't wait for the quarterly pack. It goes to the chair by phone, then in writing, then into the pack for the record. If a board learns something serious from the pack when it should have had a call, it has learned two things. The second one is about you.

12. Working from the two templates

You get two templates: a board deck built decisions-first, and an investor pitch built to the seven questions. Each carries a layout skeleton, a placeholder title storyline where every title is already a sentence, chart slots with the intended form named, and speaker notes stating what each page is for.

A template is a hypothesis about what an audience of that kind needs. Using one well means arguing with it. Anyone can fill one in. Three rules make the difference.

Write your own titles before you look at the placeholder titles. Those placeholders are shaped correctly, but they're about the wrong company, and reading them first anchors your storyline to a generic argument. Write the title stack cold from your analysis, then compare. Where they differ, you learn something. A page you couldn't write a title for is a page you don't need. A title you wrote that has no template page is where your argument really is different.

Delete before you add. Fourteen body pages ship in the board template because that's a workable quarterly shape. Nothing requires fourteen. A quarter with one decision and no bad news is a nine-page pack, and shipping fourteen because the template had fourteen is the failure the template is most likely to cause.

And for every page you keep, state what it's for and what would justify deleting it. That's a two-minute exercise per page, and it's all the method asks of you. A page whose deletion condition is "if the board stops caring about revenue" is holding the pack up. A page whose deletion condition you can't write is decoration that arrived through a template.

Chart slots deserve one specific warning. A slot labelled "trend, eight periods, line" tells you the form the layout expects. If your message compares categories and doesn't describe a change over time, the slot is wrong for your data. Change the chart and keep the message. Templates fail most often by converting a message into whatever the layout was built for, and nobody notices it happening.

13. Choosing the archetype, and three cases where the obvious choice is wrong

An audience and its decision choose the archetype, and usually that makes the choice obvious. A quarterly meeting of a company's own directors takes the board archetype. A first meeting with a fund takes the pitch archetype. A client engagement readout takes the consulting archetype. Three situations break the obvious mapping, and all three are common.

The first is the investor who's already on your board. A fund with a board seat is reading as an owner during the quarterly meeting and as an investor during a raise, and the same person needs different artefacts a fortnight apart. Build one artefact for both and you get a board pack that reads as though the executives are selling to their own directors. That corrodes the monitoring relationship that makes a board useful. Build both. They share the data and almost nothing else.

The second is the bad quarter in the middle of a raise. Everything in the pitch archetype pushes toward the story, and everything in the board archetype pushes toward the decision. A board pack written with a live term sheet in the room drifts toward the pitch. Don't let it. Your board pack stays a board pack, in full, including the margin page. Directors who discover during diligence that the pack they approved was softened have a governance problem, and you're it.

The third is the strategy day, which looks like a board meeting and isn't one. A full-day off-site exists to explore, and deciding can wait. Bring a decisions-first pack and you destroy it, by asking for votes on questions the room hasn't thought about yet. What the day needs is a pre-read document that frames the options without recommending, plus a small number of provocation slides for the room. A decisions page comes later, at the next ordinary meeting, once the exploration has produced something to decide.

One test resolves all three. Name the reader, name the decision they're taking in the next thirty days, and name what they'll do differently if they believe you. Where the answers differ between two audiences, the artefacts differ too, whatever it costs to build both.

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

An assistant is useful on the structural parts of this work and dangerous on the parts that are judgement. That boundary is sharper for board and investor material than for anything else in the track, because the reader is a fiduciary or is writing a cheque.

It's useful for the deletion pass. Give it your page list with the title of each page and ask which pages no named decision consumes. It's useful for the sceptical read of a decisions page: paste the five columns and ask what a director would ask that the row doesn't answer. It's useful for compression, turning a four-paragraph explanation into 190 words for a read page, provided you check every figure afterwards against your own computation.

It's dangerous in three specific places. It will invent a market number, because a plausible market size and a real one look the same in fluent prose, and a sizing slide is where an invented figure destroys a pitch. It will soften bad news, because the training data is full of corporate prose that softens bad news, and "margin compression from strategic channel investments" will arrive without being asked for. And it'll produce a decision that's a topic in disguise, phrased as a question but unanswerable by a vote. That passes a casual read and fails in the room.

One rule covers all three. An assistant may propose structure, and it may never supply a number or a claim. The decisions are yours because you'll be in the room when they're voted on. The market build is yours because you'll be asked where 55% came from. And the bad news is yours, because a board that learns you softened it once will read everything you write differently for as long as you work there.

{
  "skill": "Stress-testing a decisions-first board pack before it is mailed to directors",
  "use": "After you have written the decisions page and the title stack by hand, paste the titles and the five decision columns into an assistant and ask it to act as a sceptical non-executive director. Use its questions to find the missing column, the unpriced deferral and the page no decision consumes. Never let it write a decision row or supply a figure.",
  "tools": ["Chat assistants (Claude, ChatGPT, Gemini)", "A spreadsheet holding the pack's underlying computation, kept separate from the deck"],
  "prompt": "You are a non-executive director of a mid-size consumer brand. Below is the decisions page of a quarterly board pack and the title of every body page. For each decision row: 1) state whether the question can be answered yes or no, or whether it is a topic in disguise; 2) name the one number that would flip the recommendation, and say whether the row names it; 3) name what the row does not tell you about the cost of deferring. Then list any body page whose title does not support one of the decisions or one of the standing sections. Ask questions only. Do not rewrite my rows and do not supply any figures.",
  "verify": "Every figure that survives into the pack is recomputed from your own source before mailing, and the rounding on the page must match the rounding in the exhibit and the appendix page behind it. Any decision the assistant flags as unanswerable is rewritten by you until a yes and a no are both sayable. Any page it flags as unsupported is either deleted or given the decision it serves, in writing. If the assistant asserts a market figure, a benchmark or a covenant threshold, treat it as absent: those come from the company's own records or from a primary source with a date, and never from a model.",
  "diy": "You must be able to write the five decision columns cold from a fact set, and compute the break-even, the payback and the deferral cost by hand. That is the gated skill, because in the meeting you will be asked where the number came from by someone who has read the appendix, and an answer that begins with the tool you used is the end of the conversation.",
  "market": "IN"
}

{
  "skill": "Auditing an investor-pitch market slide and an ask against the plan's own arithmetic",
  "use": "Once you have built the market bottom up and computed the milestone decomposition yourself, use an assistant as an adversarial reader: have it attack the filters, the tiering and the capacity assumption, and have it check that the use-of-funds lines sum to the ask. Treat every number it offers as unusable and every question it raises as a to-do.",
  "tools": ["Chat assistants (Claude, ChatGPT, Gemini)", "A spreadsheet or Python script holding the sizing build and the milestone decomposition"],
  "prompt": "Here is a bottom-up market build and an ask slide from a seed pitch. Build: <counts, filters, tiers, blended contract value, resulting serviceable market>. Ask: <amount, use-of-funds lines, milestone, months, current recurring revenue, retention rate, plan contract value, sellers and ramp>. Do four things. 1) Name every filter whose source I have not stated. 2) Tell me which single assumption the market number is most sensitive to, and what the market becomes if it is wrong by a third. 3) Check that the use-of-funds lines sum to the ask and say what is missing from the categories. 4) State what close rate per seller per month the milestone implies, and what question you would ask about it. Do not supply market data, benchmarks or comparables.",
  "verify": "Recompute the blended contract value, the serviceable market and the implied close rate from your own script and confirm they match what you put on the slide to the stated rounding. Every filter needs a named source and a date on the slide itself. The top-down figure, if you show one, is quoted with its publisher and its year and is stated as a claim rather than a fact, and the ratio between it and your build is printed rather than hidden. Any comparable or benchmark the assistant volunteers is discarded outright.",
  "diy": "You must be able to construct the build from countable units, defend each filter to a reader who disagrees with it, and decompose a milestone into the base contribution and the new-logo requirement without help. An investor's second question is always about a filter, and the founder who cannot answer it has lost the meeting regardless of how the slide looked.",
  "market": "US"
}

Common mistakes & how experts think differently

1. Leading with performance instead of decisions. The pack opens with the quarter, because that's how the internal review is organised and it feels like context. It's tempting because it's the material you have most of, and it postpones the uncomfortable part. An expert opens with the three things needing a vote and treats performance as the standing section that follows. Context a director can't act on is being paid for at director rates.

2. Writing a decision as a topic. "Fulfilment strategy" and "Marketing investment review" look like agenda items and can't be voted on. So the meeting becomes a discussion, and it ends with the executives being asked to come back next quarter. An expert's move is mechanical. Write the row so that both yes and no are sayable, then name the number that would flip your own recommendation. If you can't write a decision's flip condition, you haven't analysed it yet.

3. Sizing a read page like a projected slide. Thirty words per page is right for a room with a speaker. When the reader is alone on a Sunday, it produces a sixty-page pack. It happens because presentation advice is written for the projected case and then gets applied to the read case. An expert sizes the pack against reading minutes: 190 words and one exhibit a page, fourteen body pages, 11.08 minutes, and an appendix that carries everything else.

4. Burying the number that contradicts the headline. Revenue up 21.92% goes on page 1 and contribution down 1.87% goes on page 44, and the analyst who did it can truthfully say nothing was hidden. It's still a failure, because the pack's job was to direct attention, and it directed attention away from the only fact that changed what the board should do. An expert puts the contradiction in the title of an early body page and shows the bridge underneath it.

5. Building a dashboard by accretion. Eighteen tiles and forty-seven numbers arrive one request at a time over four years, and no single addition was unreasonable. An expert applies the decision test once a year and removes what fails it. Here that cut twelve tiles and 61.70% of the printed numbers. Removing a tile is a decision, so it gets minuted like one, and nobody files it as a tidy-up.

6. Using a dual axis to fit two units on one chart. It turns up because two series are interesting and space is short. It's seductive because the result looks informative. The author chooses the right-hand scale, that choice determines the apparent relationship, and no reader can see that a choice was made. Index both series to 100 at the start, or use small multiples with a shared time axis, and accept that this sometimes takes two panels.

7. Applying a share to a top-down market number. "The market is $18.5 billion and we take 1%" appears on seed deck after seed deck, and every experienced reader discounts the slide without reading it. An expert builds from countable units and prints the filters. Where a top-down figure exists too, the expert shows the ratio between the two, here 1.770 times, with the build at 56.49% of the claim. Volunteering the gap buys more credibility than the larger number ever could.

8. Asking for an amount without a milestone. A use-of-funds pie gets drawn and the slide ends there, leaving the reader no way to judge whether the amount is right. An expert states what the money buys in terms of the next financing event. Then they decompose the required growth into the base contribution and the new-logo requirement, and check the requirement against a demonstrated rate: 1.400 logos per seller-month required against 1.8333 demonstrated, a cushion of 23.64%.

9. Softening bad news into a phrase. "Margin compression from strategic channel investments" is easier to write and to say than "contribution margin fell 4.80 points and two cost lines account for 5.6 points of that". That soft version survives one meeting and fails the second, when the trend continues and a director rereads the old pack. Experts treat the ugly sentence as an asset, because the analyst who wrote it is the one who gets asked the hard questions next time.

10. Building one artefact for two audiences. A board pack that has to double as a fundraising deck, or a pitch that has to double as a diligence pack, saves a day of work and costs you both jobs. Both share the underlying data and almost nothing else. Order, density, the placement of bad news and the definition of a good outcome all differ. An expert builds both, reusing the computation and never the slides.

Worked examples

Six rebuilds, three in India and three in the United States, all from two synthetic companies. Every figure comes from one computation, and the Verify block after the last example restates each one at full precision. Slide scores use the seven-rule machine check blended with a committed rubric self-score at 0.6 machine plus 0.4 rubric, with 0.80 the pass mark for an ordinary project.

Worked example 1: The headline that hid the problem (India, ₹)

The setting. Nilgiri Leaf Consumer Brands (synthetic) sells tea and wellness products direct to consumers and through retail. Eight quarters of net revenue, in ₹ crore, run 41.20, 46.80, 52.10, 58.40, 63.90, 68.30, 66.50 and 71.20 from Q3 FY25 to Q2 FY27. Contribution margin over the same eight quarters runs 26.80%, 26.10%, 25.40%, 24.60%, 23.90%, 22.70%, 21.10% and 19.80%. Net burn runs 3.10, 3.60, 4.40, 5.20, 5.90, 6.40, 7.10 and 7.80. Opening cash two years ago was ₹105.90 crore.

The wrong version. Body page 1 is titled "Q2 FY27 Performance Update". It carries a column chart of net revenue by quarter for eight quarters, a callout box reading "+21.9% YoY", and four bullets: revenue ahead of plan, quick commerce scaling, new SKU launched, marketing efficiency under review. The contribution margin appears on appendix page 44, in a table of eleven operating metrics, under the title "Operating Metrics Summary". Nothing on page 1 is false.

This page fails three of the seven rules. Its title states a topic where the message should be, so it scores zero on title-is-the-message. It carries four unrelated claims, so it fails one-message. And the callout emphasises the growth number while the contradicting number is 43 pages away, which fails hierarchy. Machine score is 4 of 7, or 0.5714. Its committed rubric self-score is 0.45, and the composite is 0.6 × 0.5714 + 0.4 × 0.45 = 0.5229. That's a long way below the 0.80 pass.

The corrected version. Two body pages, at positions 3 and 4, both titled with the claim.

Page 3 is titled: "Contribution fell ₹0.27 crore while revenue rose ₹12.80 crore, because contribution margin fell 4.80 points." Its exhibit is a line chart of contribution margin across all eight quarters on a single axis, with the last two points labelled. Underneath, the prose runs to four sentences and 96 words.

The arithmetic. Revenue Q2 FY27 against Q2 FY26 is 71.20 / 58.40 − 1 = 21.9178%. Contribution is revenue times margin: 58.40 × 0.2460 = ₹14.3664 crore a year ago against 71.20 × 0.1980 = ₹14.0976 crore now, a fall of ₹0.2688 crore or −1.8710%. Margin fell 24.60 − 19.80 = 4.80 points. In ₹ crore, the eight-quarter series of contribution runs 11.0416, 12.2148, 13.2334, 14.3664, 15.2721, 15.5041, 14.0315 and 14.0976. It peaked two quarters ago and has been below its peak since.

Page 4 is titled: "Two cost lines took 5.6 points of margin; price and mix gave 2.0 back." Its exhibit is a horizontal bridge from 24.60% to 19.80%, sorted by absolute size within the negative and positive groups.

Those steps sum to +1.40 − 3.10 − 1.20 − 2.50 + 0.60 = −4.80 points, and 24.60 − 4.80 = 19.80, so the bridge ties. Between them the two worst lines take 3.10 + 2.50 = 5.60 points, which is larger than the net fall of 4.80 because the two positive lines return 2.00.

Both pages pass all seven rules: one message each, a sentence title carrying the claim, a line for change over ordered time and a bridge for decomposition, colour used only to separate positive from negative steps, no gridline clutter, and a clear reading order. Machine 7 of 7 at 1.0000, rubric 0.94, composite 0.6 × 1.0 + 0.4 × 0.94 = 0.9760.

Read it. The page doesn't say the company is in trouble, because it isn't. Contribution is roughly flat and cash covers 24.00 months at the current burn. What the page does say is that this quarter's revenue at last year's margin would have produced 71.20 × 0.2460 = ₹17.5152 crore of contribution. So ₹3.4176 crore has been handed back, which is 26.70% of the ₹12.80 crore of revenue the company added. More than a quarter of the growth is going straight into margin, and the giving-back is accelerating.

What would break the reading is a mix effect the bridge hasn't isolated. Suppose the quick-commerce channel simply carries a structurally lower margin. Then the trade-spend line comes from channel mix, and nobody lost a negotiation. Fixing it would take a pricing decision, and procurement couldn't solve it.

The checklist to apply.

1. The title is a sentence with a verb and carries the number.
2. The exhibit shows the whole series, never just the two periods that flatter.
3. Any decomposition sums to the change it is decomposing, shown.
4. The contradicting number goes in the body, never in the appendix.
5. What would break the reading is stated on the page.

Worked example 2: The decisions page, rebuilt (India, ₹)

The setting. Nilgiri Leaf's board meets for 120 minutes, and three matters need approval. First, an owned fulfilment centre costing ₹14.00 crore of capital with a fixed cost of ₹1.35 crore a quarter. It would replace a third-party arrangement at ₹78 an order with an in-house cost of ₹41 an order, and the company shipped 960,000 orders in Q2 FY27 on net revenue of ₹71.20 crore. Second, a 6% list-price increase on the top three SKUs, which are 62% of revenue and carry a 21.5% contribution margin, against an assumed price elasticity of −0.80. Third, a rise in the quarterly performance-marketing cap from ₹9.50 crore to ₹12.00 crore. It assumes a marginal acquisition cost of ₹910 against a blended ₹652.17, in a company whose board covenant caps acquisition-cost payback at 14.0 months.

The wrong version. Three separate pages at positions 31, 44 and 12, titled "Fulfilment Strategy Update", "Pricing Review" and "Marketing Investment". Each runs seven to eleven bullets. None states an amount to be approved. Marketing comes first because the executive who wants it most built the pack. Directors approve the marketing increase in eleven minutes and run out of time on fulfilment.

Machine score 4 of 7, or 0.5714: the pages fail title-is-the-message, one-message and hierarchy. Rubric 0.40. Composite 0.6 × 0.5714 + 0.4 × 0.40 = 0.5029.

The corrected version. One page at position 2, ordered by size of irreversible commitment, with the supporting analysis on pages 5 to 11 and every row naming its appendix page.

The arithmetic, in order. Average order value is 71.20 crore / 960,000 = ₹741.67. Owning saves 78 − 41 = ₹37 an order. Against the ₹1.35 crore quarterly fixed cost, break-even is 1.35 × 10^7 / 37 = 364,864.9 orders, which is 364,864.9 / 960,000 = 38.01% of current volume, so volume can fall 61.99% before the decision reverses. At current volume the centre saves 960,000 × 37 − 1.35 × 10^7 = ₹2.2020 crore a quarter and ₹8.8080 crore a year. Its ₹14.00 crore of capital pays back in 14.00 / 2.2020 = 6.358 quarters, or 19.07 months.

A board will ask about the runway consequence, and the original pages never computed it. Cash is ₹62.40 crore against a net burn of ₹7.80 crore a quarter, which is 8.000 quarters or 24.00 months. Build the centre and cash falls by the ₹14.00 crore of capital plus two quarters of unchanged burn, leaving 62.40 − 14.00 − 15.60 = ₹32.80 crore. After that, burn drops to 7.80 − 2.2020 = ₹5.5980 crore a quarter. Runway becomes 2 + 32.80 / 5.5980 = 7.859 quarters, or 23.58 months. So the decision costs 0.42 months of runway and buys ₹8.8080 crore a year, and that's the sentence the row exists to support.

For the price rise, top-three revenue is 71.20 × 0.62 = ₹44.1440 crore. A 6% price rise at an elasticity of −0.80 implies a volume change of −0.80 × 0.06 = −4.80%, so revenue becomes 44.1440 × 1.06 × 0.952 = ₹44.5466 crore, a gain of ₹0.4026 crore. Contribution matters more than revenue, though. Variable cost before is 44.1440 × 0.785 = ₹34.6530 crore, and volume falls 4.80%, so variable cost after is ₹32.9897 crore. Contribution after is 44.5466 − 32.9897 = ₹11.5569 crore against ₹9.4910 crore before. That's a gain of ₹2.0659 crore a quarter or ₹8.2638 crore a year, which is 2.9016 points of company contribution margin.

Break-even is the volume loss at which contribution is unchanged. After the rise, contribution per unit of retained volume is 44.1440 × 1.06 − 34.6530 = ₹12.1396 crore, so the retained fraction is 9.4910 / 12.1396 = 0.78182 and volume can fall 21.82%. Against that, the plan assumes 4.80%, a cushion of 17.02 points. To wipe out the gain, elasticity would have to reach −0.2182 / 0.06 = −3.636, far outside anything observed in the category.

For the marketing cap, the six acquisition channels carry costs per new customer of ₹412 (own website, 46,000 customers), ₹520 (marketplaces, 31,000), ₹1,180 (quick commerce, 22,000), ₹735 (paid social, 24,000), ₹648 (influencer and affiliate, 12,000) and ₹880 (retail media, 7,000). Total spend is ₹9.2608 crore across 142,000 new customers, so blended cost is ₹652.17. Contribution per order at the current margin is 741.67 × 0.198 = ₹146.85. At 3.4 orders a customer a year, that's ₹499.29 of contribution a year, or ₹41.61 a month. Blended payback is 652.17 / 41.6075 = 15.67 months, already 1.67 months past the 14.0-month covenant.

Now the request. An extra ₹2.50 crore of spend at a marginal cost of ₹910 buys 27,472 customers at a payback of 910 / 41.6075 = 21.87 months, and that breaches the covenant by 7.87 months.

A counter-proposal turns the row from a refusal into a recommendation. Move ₹1.80 crore into reactivating lapsed customers at ₹190 each, which buys 94,737 reactivations. If each returns 1.9 incremental orders in twelve months, contribution per reactivation is 146.85 × 1.9 = ₹279.02, and payback is 190 / 23.2513 = 8.17 months. That's 52.13% of the blended acquisition payback, and 7.50 months better.

Machine 7 of 7, rubric 0.92, composite 0.9680.

Read it. The page isn't a recommendation to spend less. It recommends spending ₹14.00 crore of capital and raising prices, and declines only the one request whose arithmetic breaches a limit the board itself set. What would break it is the elasticity assumption on decision 2, the only input with no internal evidence behind it. A −0.80 elasticity is borrowed from category data, and the row should say plainly that the company has never tested a 6% rise on these SKUs.

The checklist to apply.

1. Every decision is answerable yes or no as written.
2. The do-nothing option appears in the option set.
3. The number that would flip the recommendation is named and computed.
4. Deferral is priced in rupees or in weeks.
5. Rows are ordered by size of irreversible commitment.
6. Each row names the appendix page behind it.

Worked example 3: The board dashboard, eighteen tiles to six (India, ₹)

The setting. Nilgiri Leaf's inherited board dashboard is one page carrying eighteen tiles and forty-seven printed numbers, of which three tiles are charts with two vertical axes. It has grown by accretion since the Series A, and no metric has ever been removed. This quarter, the board's own decisions are the three above.

The wrong version. Among the eighteen: revenue, revenue growth, orders, average order value, new customers, repeat rate, app installs, website sessions, conversion rate, net promoter score, contribution margin, inventory days, marketing spend, blended acquisition-cost payback, headcount, burn, cash, and a channel-mix pie with seven slices. That pie is the worst object on the page. Seven slices is more than the three the chart grammar allows, and the pie is being asked to carry change between two periods, which a single pie can't express at all.

Machine 2 of 7, or 0.2857: it fails one-message, chart-choice, declutter, hierarchy and colour-discipline. Rubric 0.35. Composite 0.6 × 0.2857 + 0.4 × 0.35 = 0.3114.

The corrected version. Six tiles, chosen by asking of each metric which board decision it changes.

Twelve tiles went, a cut of 66.67%, and printed numbers fell from forty-seven to eighteen, down 61.70%. Inventory days is computed as 21.60 / 29.5480 × 91 = 66.52 on a 91-day quarter, where cost of goods for the quarter is 71.20 × 0.415 = ₹29.5480 crore and closing inventory is ₹21.60 crore. Against 58.00 days a year earlier, the 8.52-day rise is why the fulfilment paper has to carry a working-capital line.

A slope chart of channel share between Q2 FY26 and Q2 FY27 replaces the seven-slice pie. It's the correct form for change between exactly two periods across a few categories.

Shares sum to 100.0% in both periods, and channel revenue sums to ₹58.40 crore and ₹71.20 crore respectively. A slope chart carries the message the pie couldn't. Quick commerce doubled its share while the own website lost ten points, and the own website is the only channel whose absolute revenue fell, by ₹2.768 crore or 13.94%. Marketplaces lost four points of share while growing 0.71%. A share-only chart hides that distinction, and a chart with both share and revenue makes it visible.

Machine 7 of 7, rubric 0.90, composite 0.9600.

Read it. The rebuilt dashboard isn't a smaller version of the old one. Five of the six tiles were chosen because a decision on the agenda consumes them, and the sixth is there because a board has a monitoring duty independent of any decision. What would break it is a quarter with different decisions. A decision-driven dashboard changes composition between quarters, and some boards dislike that. A compromise that works is a fixed standing four plus two that rotate with the agenda.

The checklist to apply.

1. Every tile names the decision it serves, or is one of the standing monitoring set.
2. No chart on the page uses two vertical axes.
3. No pie carries more than three slices.
4. Shares sum to 100 and totals tie to the financial statements.
5. Colour marks a breached threshold and nothing else.

Worked example 4: The traction slide a sceptic can check (US, $)

The setting. Yardpost (synthetic) sells dock-door scheduling software to United States food distributors. Twelve months ago it had $412,000 of annual recurring revenue across 19 customers. Over twelve months it added 36 new logos at contract values rising from $23,400 to $30,000, gained $83,440 of expansion revenue, and lost 4 customers carrying $34,000 of recurring revenue. It spent $784,000 on sales and marketing, runs a 78% gross margin, burned $1,640,000 net over the year, and holds $1,180,000 of cash against a current burn of $152,000 a month.

The wrong version. One slide titled "Traction", carrying a column chart of monthly recurring revenue with customer count on a second vertical axis, a "3.5x" callout, and the sentence "strong momentum across all segments". Churn appears nowhere. Its right-hand axis is scaled so the customer line sits neatly above the revenue bars, and no reader can detect that scaling choice.

Machine 4 of 7, or 0.5714: it fails chart-choice (dual axis), title-is-the-message and hierarchy. Rubric 0.42. Composite 0.5109.

The corrected version. Two panels on one slide, sharing a time axis, no second vertical axis anywhere.

The upper panel is a line of annual recurring revenue across thirteen month-ends, from $412,000 to $1,437,640. The lower panel is a stacked column of quarterly net new recurring revenue, split into new, expansion and churn, which is a part-to-whole message across three components over four periods.

Those four quarters sum to $1,025,640, which equals 1,437,640 − 412,000 to the dollar, so the chart ties to the series above it. Q4 net new is 1.942 times Q1. Showing churn as a visible negative component is what turns the slide from a claim into evidence.

The title carries the message: "ARR grew 3.49x to $1.44m with 112% net revenue retention; every quarter added more than the one before." Three retention figures sit in a small block beneath, each with its formula. Net revenue retention is (412,000 + 83,440 − 34,000) / 412,000 = 112.00%. Gross revenue retention is (412,000 − 34,000) / 412,000 = 91.75%. Logo retention is (51 − 36) / 19 = 78.95%, computed on the nineteen customers who were there at the start, which is the base the gross figure uses too. Put the 36 new logos in the denominator and you'd report 92.73%, comparing a churn count against a base that grew while the churn was happening. Those four lost accounts averaged 34,000 / 4 = $8,500 against an opening-base average of 412,000 / 19 = $21,684, so a fifth of the customers took 8.25% of the revenue with them.

Four efficiency figures go in the footer. A separate slide would be the wrong place, because a sceptical reader wants them next to the growth. Acquisition cost is 784,000 / 36 = $21,777.78. Monthly gross profit per new customer is (976,200 / 36) × 0.78 / 12 = $1,762.58, so payback is 21,777.78 / 1,762.58 = 12.36 months. And the burn multiple is 1,640,000 / 1,025,640 = 1.599. Cash of $1,180,000 against $152,000 a month is 7.76 months of runway. That number explains why the company is raising now, so put it on the slide and don't make anyone extract it in the meeting.

Machine 7 of 7, rubric 0.95, composite 0.9800.

Read it. Growth of 248.94% comes off a small base of $412,000, and a sophisticated reader will say so. It's the retention figures that carry the slide. A 112.00% net revenue retention on a 91.75% gross figure means expansion is outrunning churn, in a product that's been live long enough to churn four accounts. What would break the reading is concentration. Fifty-one customers at an average of $28,189 hides whatever the largest three contribute. A reader who asks for the top-five share and hears it's in the appendix will believe the slide. A reader who hears it hasn't been computed won't.

The checklist to apply.

1. No second vertical axis anywhere on the slide.
2. Churn is visible as a negative component, never netted away.
3. The components sum to the total shown in the panel above.
4. Every retention figure carries its formula.
5. Runway appears next to growth, not three slides later.

Worked example 5: The market slide that failed its own test (US, $)

The setting. Yardpost needs one market slide in a twelve-slide seed deck. Here's what's available. There are 21,400 food distribution facilities in the addressable geography. To pay for itself, the product needs twelve or more dock doors, which is 38% of facilities. Of those, 55% run scheduled rather than first-come receiving. Contract values are $14,400 for small facilities under twenty doors (55% of the base), $27,600 for mid at twenty to forty-nine doors (33%), and $52,800 for large at fifty or more (12%). A published market study puts food and beverage distribution software at $18.5 billion.

The wrong version. A slide titled "A $18.5B Market" with three concentric circles labelled total, serviceable and obtainable, and the sentence "capturing just 1% represents a $185M opportunity". The word "just" is doing all the work, and a reader who has seen two hundred seed decks discounts the slide without reading the circles.

Machine 4 of 7, or 0.5714, failing title-is-the-message, one-message and declutter. Rubric 0.30, low because the rubric asks whether the claim is defensible, and this one isn't. Composite 0.4629.

The corrected version. One slide titled "4,473 facilities can buy this today at a blended $23,364, a $104m serviceable market", carrying a funnel of four countable steps and a tier table, with the top-down comparison stated in one line at the foot.

That gives a serviceable market of 4,472.6 × 23,364 = $104,497,826. At its foot, the slide carries one sentence a weaker deck would omit: the published top-down figure implies $185,000,000 at a 1% share, which is 1.770 times this build, so we show the build. That's a gap of $80,502,174, and the build is 56.49% of the claim.

Adjacent verticals get a separate labelled bar of their own. At an assumed multiplier of 2.6 for pharmaceutical and building-products distribution, the extended market is $271,694,349, and the multiplier is printed as an assumption. Nobody should mistake it for a finding.

Two scale lines close the slide. Current recurring revenue of $1,437,640 is 1.376% of the built market, and the eighteen-month milestone of $4,200,000 is 4.02% of it, or 1.55% of the extended figure. Those two lines turn an opportunity into a plan.

Machine 7 of 7, rubric 0.91, composite 0.9640.

Read it. The rebuilt slide claims a market roughly 56% the size of the original claim, and it's far more persuasive. Every filter is a number a reader can argue with, and arguing is engagement. What would break it is the 55% scheduled-receiving figure. It rests on 74 discovery conversations and it's the weakest input on the slide. At 40% the market falls to $75,998,419. Print the sensitivity or expect to be asked for it.

The checklist to apply.

1. The market is built from a countable unit, with the count sourced and dated.
2. Every filter is stated on the slide with its evidence.
3. Contract values are tiered and the blend is shown.
4. Any top-down figure is quoted as a claim, with the ratio to your build printed.
5. Adjacency is a separate bar with its multiplier stated.
6. Current revenue and the milestone are expressed as shares of the built market.

Worked example 6: The ask tied to a milestone (US, $)

The setting. Yardpost is raising $3,500,000. Recurring revenue is $1,437,640, growing at $341,040 of net new in the most recent quarter, with net revenue retention of 112.00%. Its plan is to reach $4,200,000 of recurring revenue in eighteen months, at a plan contract value of $30,000, hiring four account executives with a three-month ramp. Last quarter it closed 11 logos from 6 account-executive months of selling capacity. Existing cash is $1,180,000, and burn at month eighteen is planned at $180,000 a month.

The wrong version. A slide titled "The Ask" carrying "$3.5M Seed Round" in large type, a six-slice pie of use of funds, and three bullets: scale go-to-market, invest in product, expand into adjacent verticals. No milestone. No arithmetic. The pie's two smallest slices differ by two degrees and cannot be told apart.

Machine 5 of 7, or 0.7143, failing chart-choice and title-is-the-message. Rubric 0.50. Composite 0.6286. That's below the 0.80 pass, and it's the sort of near-miss that survives review because the slide looks finished.

The corrected version. One slide titled "$3.5m takes ARR from $1.44m to $4.2m in 18 months, which needs 84 new logos against 60 seller-months of capacity."

Use of funds goes in a sorted bar, largest first, with five categories.

People account for 86.29% of the raise, which the slide says outright.

The original slide had no milestone block at all. This one is four lines of arithmetic a reader can redo.

Required net new recurring revenue is 4,200,000 − 1,437,640 = $2,762,360. That's $153,464 a month against a demonstrated $113,680 a month in the most recent quarter, so the required rate is 1.350 times what has been shown. Decomposed, the existing base at 112.00% retention grows to 1,437,640 × 1.12^1.5 = $1,704,030 over eighteen months, contributing $266,390 by itself. So new logos must supply 4,200,000 − 1,704,030 = $2,495,970. At $30,000 of contract value that's 83.199 logos, so 84. Capacity is 4 × (18 − 3) = 60 account-executive months. So the plan requires 84 / 60 = 1.400 logos per seller-month against 1.8333 demonstrated in the last quarter, a cushion of 23.64%.

Two lines close it. If the fourth account-executive hire slips ninety days, the plan loses 3 × 1.400 = 4.20 logos, being $126,000 of recurring revenue or 4.56% of what's required. And on cash, the raise plus existing cash gives $4,680,000. Of that, $3,600,000 is planned spend over eighteen months, averaging $200,000 a month. That leaves $1,080,000 at the milestone, or 6.00 months of runway at the planned month-eighteen burn of $180,000. The implied burn multiple is 3,600,000 / 2,762,360 = 1.303, against 1.599 over the trailing twelve months.

Machine 7 of 7, rubric 0.93, composite 0.9720.

Read it. The slide isn't claiming the plan is safe. Six months of runway at the milestone is thin, and the slide says so. Leave a reader to compute it and they'll conclude you hadn't. What would break the plan is the plan contract value. At $24,000 instead of $30,000, the requirement rises to 104 logos and the required close rate to 1.733 per seller-month. That's 94.5% of the demonstrated rate, which leaves almost no cushion. The sensitivity belongs in the appendix, and the founder should know it exists before the meeting.

The checklist to apply.

1. The use-of-funds lines sum to the ask, shown.
2. Use of funds is a sorted bar, never a pie with more than three slices.
3. The milestone describes a state, never an activity, and it's dated.
4. Required growth is decomposed into base contribution and new-logo requirement.
5. The requirement is checked against a demonstrated rate, with the cushion stated.
6. Runway at the milestone is on the slide.

Verify

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

Nilgiri Leaf eight-quarter contribution, ₹ crore: 11.0416, 12.2148, 13.2334, 14.3664, 15.2721, 15.5041, 14.0315, 14.0976. Closing cash: 102.80, 99.20, 94.80, 89.60, 83.70, 77.30, 70.20, 62.40, from opening 105.90 less cumulative burn summing to 43.50.

Revenue growth 71.20/58.40 − 1 = 0.2191780821917808 (21.9178%). Contribution 58.40 × 0.2460 = 14.3664 and 71.20 × 0.1980 = 14.097600000000002; change −0.2688; −0.018710324089542207 (−1.8710%). Margin change −4.800000000000001 points. Contribution at the prior margin 71.20 × 0.2460 = 17.5152; handed back 3.4176; as a share of the 12.80 revenue rise 0.2670 (26.70%). Quarter on quarter revenue 71.20/66.50 − 1 = 7.0677%; margin −1.30 points.

Bridge in points +1.40, −3.10, −1.20, −2.50, +0.60, summing −4.80; on Q2 FY27 revenue +0.9968, −2.2072, −0.8544, −1.7800, +0.4272 ₹ crore. Check 24.60 + (−4.80) = 19.80. Two worst lines 5.60 points.

Channel shares FY26 15.0, 34.0, 23.0, 13.0, 9.0, 4.0, 2.0 and FY27 30.0, 24.0, 19.0, 12.0, 8.0, 5.0, 2.0, each summing 100.0. Channel revenue FY27 21.360, 17.088, 13.528, 8.544, 5.696, 3.560, 1.424, summing 71.2000. Growth +143.84%, −13.94%, +0.71%, +12.54%, +8.37%, +52.40%, +21.92%. Own-website revenue change −2.7680. Top two share 54.0%.

Average order value 71.20 × 10^7 / 960,000 = 741.6666666666666. Saving 37.00. Break-even 1.35 × 10^7 / 37 = 364,864.86486486485; 38.0068% of volume; fall 61.9932%. Quarterly saving 2.202; annual 8.808. Payback 14.00/2.202 = 6.357856 quarters, 19.07356948228883 months. Runway 62.40/7.80 = 8.0 quarters, 24.000000000000007 months. Cash after build 32.80; burn after 5.598; runway 7.859235441229012 quarters, 23.577706323687035 months; cost 0.4223 months.

Top-three revenue 44.1440. Volume change −4.80%. Revenue after 44.54661 and gain +0.402611. Contribution before 9.49096; variable cost before 34.65304; after 32.98969; contribution after 11.556928; gain 2.0659392000000043; annualised 8.263756800000017; as company margin +2.9016% points. Contribution per retained unit 12.139584; retained fraction 0.7818181818181817; break-even loss 21.818181818181838%; cushion 17.0182 points; implied elasticity −3.6363636363636402.

Acquisition spend by channel, ₹ crore: 1.8952, 1.6120, 2.5960, 1.7640, 0.7776, 0.6160, total 9.2608; customers 142,000; blended 652.169014084507. Ratio dearest to cheapest 2.8640776699029127. Contribution per order 146.85; per customer per year 499.29000000000...; per month 41.6075. Blended payback 15.674313863714644 months; marginal 910/41.6075 = 21.871056900799136; ceiling 14.0; breaches +1.6743 and +7.8711. Incremental customers 2.50 × 10^7 / 910 = 27,472.5. Reactivations 1.80 × 10^7 / 190 = 94,736.8; contribution per reactivation 279.01500000000004; payback 8.171603677221656 months; 52.1276% of blended; advantage 7.5027 months.

Dashboard: tiles 18 → 6, cut 12, 66.6667%; numbers 47 → 18, 61.7021%; dual axes 3 → 0. Cost of goods 71.20 × 0.415 = 29.548; inventory days 21.60/29.548 × 91 = 66.52226885068364; prior 58.00; change +8.5222688506836; cash held by the rise 8.5222688506836/91 × 29.548 = 2.767222... ₹ crore.

Yardpost ARR by month: 412,000, 463,600, 516,800, 587,600, 644,000, 727,800, 804,400, 892,600, 983,000, 1,096,600, 1,191,400, 1,317,800, 1,437,640; customers 19 to 51. New ARR 976,200; expansion 83,440; churn 34,000; net new 1,025,640. Multiple 3.4894174757281555; growth 248.9417%. Net revenue retention 1.1200000000000001 (112.0000%); gross 0.9174757281553399 (91.7476%); logo (51 − 36)/19 = 78.9474%. Churned average contract value 8,500; opening-base average 21,684.2105; revenue lost to churn 8.2524%. Ending contract value 28,189.019607843136; average new-logo 27,116.666666666668.

Quarterly net new 175,600, 216,800, 292,200, 341,040, summing 1,025,640; Q4/Q1 = 1.9421412300683372. Acquisition cost 784,000/36 = 21,777.777777777777; monthly gross profit per customer 1,762.5833333333335; payback 12.355601783997605 months. Burn multiple 1,640,000/1,025,640 = 1.599001599001599. Runway 1,180,000/152,000 = 7.763157894736842 months.

Market: top-down 18,500,000,000 × 0.01 = 185,000,000.0. Facilities 21,400; after doors 8,132.0; after receiving 4,472.6. Blended contract value 0.55 × 14,400 + 0.33 × 27,600 + 0.12 × 52,800 = 23,364.0. Serviceable market 104,497,826.4. Ratio 1.7703717519621058; gap 80,502,173.6; build share 56.4859%. Extended at 2.6 is 271,694,348.64000005. Current ARR share 1.3757%; milestone share 4.019222355805862% and 1.5459%. Sensitivity at a 40% receiving filter 75,998,419.2.

Ask 3,500,000; uses 1,480,000, 1,120,000, 420,000, 290,000, 190,000, summing 3,500,000; shares 42.2857%, 32.0000%, 12.0000%, 8.2857%, 5.4286%; people 86.2857%. Required net new 2,762,360; per month 153,464.44444444444; against Q4 rate 1.3498. Base grown 1,437,640 × 1.12^1.5 = 1,704,029.7889...; base contribution 266,389.7889; new-logo requirement 2,495,970.2111; logos 83.199007 so 84. Capacity 60 seller-months; required rate 1.4; demonstrated 11/6 = 1.8333333333333333; cushion 23.636363636363633%. Slip 4.2 logos, 126,000.0, 4.5613%. Cash after raise 4,680,000; plan spend 3,600,000; average monthly 200,000.0; cash at milestone 1,080,000.0; runway 6.0 months; implied burn multiple 1.3031.

Pack arithmetic: pages 68 → 14 + 41 = 55, cut 13; body share 25.4545%. Reading 14 × 190 / 240 = 11.083333333333334 minutes. Meeting 120 minutes; per decision 20.0. Pitch 12 slides in 12 minutes, 1.0 a slide, with 23 appendix slides.

Composites, at 0.6 machine plus 0.4 rubric, with the machine score being rules passed out of seven: decisions page 0.6 × 0.5714285714285714 + 0.4 × 0.40 = 0.5028571428571429 before and 0.968 after; dashboard 0.31142857142857144 and 0.96; margin bridge 0.5228571428571429 and 0.976; traction 0.5108571428571428 and 0.98; market 0.4628571428571429 and 0.964; ask 0.6285714285714286 and 0.972. Pass mark 0.80.

Practice set

Work each one before you read the solution. You'll need a calculator. Numeric answers within 2% of the stated figure count as correct unless a problem says otherwise. Where a problem asks for a sentence, the test is whether a director or an investor could disagree with it.

P1 (guided). The headline and the number under it (India). Nilgiri Leaf reports net revenue of ₹71.20 crore against ₹58.40 crore a year earlier, at a contribution margin of 19.80% against 24.60%. Compute revenue growth and the change in contribution in both rupees and percent, then write the page title a board pack should carry.

Solution. Revenue growth is 71.20/58.40 − 1 = 21.9178%. Contribution is 71.20 × 0.1980 = ₹14.0976 crore now against 58.40 × 0.2460 = ₹14.3664 crore a year ago, a fall of ₹0.2688 crore or −1.8710%. A title that carries the message: "Contribution fell ₹0.27 crore while revenue rose ₹12.80 crore, because contribution margin fell 4.80 points." A common wrong answer is "Revenue grew 21.9% year on year". It's true, and it's the headline the original pack used, but it directs attention away from the only fact that changes what the board should do.

P2 (guided). Two quarters, not four (India). Contribution margin was 22.70% in Q4 FY26 on revenue of ₹68.30 crore and 19.80% in Q2 FY27 on ₹71.20 crore. Compute the change in margin and in contribution over those two quarters, and say what it adds to the year-on-year comparison in P1.

Solution. Margin fell 19.80 − 22.70 = −2.90 points in two quarters. Contribution went from 68.30 × 0.2270 = ₹15.5041 crore to 71.20 × 0.1980 = ₹14.0976 crore, a fall of ₹1.4065 crore. Year on year, the comparison shows a 4.80-point fall over four quarters. This one shows 2.90 of those points arriving in the last two, so the decline is accelerating, and it hasn't flattened. It's the reason the standing performance page carries eight quarters as a line and never two periods in a table. A two-period comparison is compatible with a problem that has stopped, and the line isn't.

P3 (guided). Retention three ways (US). Yardpost began the year with $412,000 of recurring revenue across 19 customers, added 36 logos and $83,440 of expansion, and lost 4 customers carrying $34,000. Compute net revenue retention, gross revenue retention and logo retention, and say which one an investor reads first.

Solution. Net revenue retention is (412,000 + 83,440 − 34,000)/412,000 = 112.00%. Gross revenue retention is (412,000 − 34,000)/412,000 = 91.75%. Logo retention is (51 − 36)/19 = 78.95%, on the opening base of nineteen, so that it's comparable with the gross figure. Customers who joined during the year stay out of it. An investor reads gross first, because expansion can't rescue it, so it measures whether the product holds. A net figure above 100% on a weak gross figure describes a company growing inside a leaking bucket, and you have to show the two together for the pair to mean anything.

P4. Break-even at three fixed costs (India). The owned fulfilment centre saves ₹37 an order against the third party. Compute the break-even quarterly order volume at fixed costs of ₹1.00 crore, ₹1.35 crore and ₹1.80 crore a quarter, and say what changes in the decision row at each.

Solution. Break-even volumes are 1.00 × 10^7 / 37 = 270,270.3 orders, 1.35 × 10^7 / 37 = 364,864.9 and 1.80 × 10^7 / 37 = 486,486.5. Against current volume of 960,000 those are 28.15%, 38.01% and 50.68%, so the volume fall the decision survives goes from 71.85% to 61.99% to 49.32%. The "turns on" column changes at every level and the recommendation doesn't. That's the finding a board wants. The decision survives every plausible level of the fixed cost, so the row should say so, and it shouldn't present a single point estimate as though it were the only case examined.

P5. Runway at three burn levels (India). Cash is ₹62.40 crore. Compute runway in months at quarterly net burn of ₹7.80 crore, ₹6.80 crore and ₹5.80 crore, and state the sentence the cash tile should carry.

Solution. Runway is 62.40/7.80 × 3 = 24.00 months, 62.40/6.80 × 3 = 27.53 and 62.40/5.80 × 3 = 32.28. The tile sentence: "₹62.40 crore of cash, 24.00 months at the current ₹7.80 crore quarterly burn. The first ₹1.00 crore of quarterly burn removed buys 3.53 months and the next buys 4.75, because runway is convex in burn." Putting the sensitivity next to the level turns a monitoring number into a decision input. A board wants to know what a cost decision would buy, and the level alone can't tell it.

P6. The price rise across four elasticities (India). Top-three SKU revenue is ₹44.1440 crore at a 21.5% contribution margin. Compute the quarterly contribution gain from a 6% price rise at elasticities of −0.5, −0.8, −1.2 and −1.5, and find the elasticity at which the gain disappears.

Solution. Volume change is elasticity times 6%. Variable cost before is 44.1440 × 0.785 = ₹34.6530 crore and contribution before is ₹9.4910 crore. At each elasticity, revenue after is 44.1440 × 1.06 × (1 + change) and variable cost after is 34.6530 × (1 + change). The gains are +₹2.2845 crore at −0.5, +₹2.0659 crore at −0.8, +₹1.7746 crore at −1.2 and +₹1.5561 crore at −1.5. The gain reaches zero when volume falls 21.82%, which requires an elasticity of −0.2182/0.06 = −3.636. So the decision is insensitive to the elasticity assumption across any plausible value, and that's the strongest argument the row has.

P7. Payback by channel against a covenant (India). Contribution per customer per year is ₹499.29. Compute acquisition-cost payback in months for channels costing ₹412, ₹520, ₹1,180, ₹735, ₹648 and ₹880 per new customer, and say which pass a 14.0-month covenant ceiling.

Solution. Monthly contribution is 499.29/12 = ₹41.6075. Paybacks are 412/41.6075 = 9.90 months (own website), 12.50 (marketplaces), 28.36 (quick commerce), 17.67 (paid social), 15.57 (influencer and affiliate) and 21.15 (retail media). Two pass and four breach. A blended figure of 15.67 months hides that, so a blended metric belongs on a dashboard only alongside its decomposition. What you can act on is that quick commerce costs 2.864 times the cheapest channel and takes 28.36 months to repay. No board decision follows from the blend alone.

P8. Sizing sensitivity (US). Yardpost's build is 21,400 facilities, a dock-door filter, a scheduled-receiving filter and a blended contract value of $23,364. Compute the serviceable market at filter pairs of (0.38, 0.55), (0.30, 0.55), (0.38, 0.40) and (0.45, 0.65), and say which line to print on the slide.

Solution. For the base pair, 21,400 × 0.38 × 0.55 × 23,364 = $104,497,826. The other three pairs give × 0.30 × 0.55 = $82,498,284, × 0.38 × 0.40 = $75,998,419 and × 0.45 × 0.65 = $146,246,958. Across defensible filters the range is roughly $76m to $146m, a factor of 1.9. Print the base build as the headline and the receiving filter as the named sensitivity, because it rests on 74 discovery conversations and moves the answer most in the downward direction. At 40% the market falls 27.3% to $75,998,419. Print the range, too. It's what stops a reader from assuming the number is a fiction.

P9. Logos required at three contract values (US). New-logo recurring revenue required is $2,495,970. Compute the logos needed at contract values of $24,000, $30,000 and $36,000, and the close rate each implies against 60 seller-months.

Solution. 2,495,970/24,000 = 104.00 logos, /30,000 = 83.20, /36,000 = 69.33. Against 60 seller-months those are 1.733, 1.387 and 1.156 logos per seller-month, against a demonstrated 1.8333. At $24,000 you need 94.5% of the demonstrated rate, leaving a cushion of 5.5%, and that isn't a plan. So contract value is the sensitivity that matters, more than close rate, and a founder who has computed this can answer the question before it's asked.

P10. Burn multiple at three plan spends (US). Required net new recurring revenue is $2,762,360. Compute the implied burn multiple at planned spends of $3,000,000, $3,600,000 and $4,200,000, and say what the trailing figure of 1.599 adds.

Solution. Those spends give 3,000,000/2,762,360 = 1.086, 3,600,000/2,762,360 = 1.303 and 4,200,000/2,762,360 = 1.520. Over the trailing twelve months the multiple ran at 1.599, so the middle case is a claimed improvement of 0.296 and the top case is essentially no improvement at all. At seed stage an investor looks at the direction, and the level comes second. A plan whose burn multiple is flat against the trailing figure is buying growth at an unchanged price, the opposite of the claim a raise usually makes.

P11. The channel move, and the chart that carries it (India). Channel shares moved from 15.0, 34.0, 23.0, 13.0, 9.0, 4.0 and 2.0 percent to 30.0, 24.0, 19.0, 12.0, 8.0, 5.0 and 2.0. Name the largest gain and the largest loss, state the correct chart form, and say which channel's absolute revenue fell.

Solution. Quick commerce has the largest gain, at +15.0 points, and the own website the largest loss, at −10.0 points. You want a slope chart of share between the two periods, because the message is change between exactly two periods across a few categories. A pie fails twice over. Seven slices exceeds the three the grammar allows, and one pie can't express change at all. On revenue, applying shares to ₹58.40 crore and ₹71.20 crore, the own website is the only channel to fall in absolute terms, from ₹19.856 crore to ₹17.088 crore, down 13.94%. Marketplaces lost 4.0 points of share while growing 0.71%, and that distinction is what makes a share-only chart dangerous on its own.

P12. The dashboard cut and the hiring slip (India and US). An inherited dashboard carries 18 tiles and 47 printed numbers. Six tiles survive the decision test, with 18 numbers. Compute both reductions. Then, for Yardpost, compute what a 90-day slip in the fourth account-executive hire costs in logos, dollars and share of the requirement, at a required rate of 1.400 logos per seller-month and a $30,000 contract value.

Solution. Tiles fall 18 − 6 = 12, a reduction of 1 − 6/18 = 66.67%. Numbers fall from 47 to 18, a reduction of 61.70%. Slipping the hire removes three selling months, so 3 × 1.400 = 4.20 logos, which is 4.20 × 30,000 = $126,000 of recurring revenue, or 126,000/2,762,360 = 4.56% of what the milestone requires. Both computations do the same job in two different artefacts. They turn an editorial judgement and a hiring risk into a number a reader can hold. A risk with a computed cost gets managed. A risk in a bullet gets read and forgotten.

Applied mini-project

Build one of the two archetypes for the supplied brief, then write a half-page note on what would change if the same content had to go to the other audience. Budget about four hours for the build and forty minutes for the note.

Choose one of the two briefs below.

Brief A, board pack (India). Nilgiri Leaf's Q3 FY27 board meets in three weeks. Since the pack above, two things have happened. The quick-commerce platform has proposed moving from a 22% trade-spend rate to a 26% rate in exchange for guaranteed top-of-category placement. By the growth team's forecast, that would lift quick-commerce revenue 35%, and it would apply to a channel already at 30.0% of revenue. And the fulfilment centre approved last quarter has hit a permitting delay of one quarter, which pushes the saving out by three months. Cash, burn and margins are as computed above. Build the pack.

Brief B, seed pitch (US). Yardpost has a first meeting with a generalist seed fund in nine days. Since the deck above, the largest customer, at $52,800 of contract value, has given notice for renewal in four months over a missing integration, and engineering estimates the integration at eleven weeks. All other figures hold. Build the pitch.

Four artefacts are submitted, and each has a test it must pass.

1. The storyboard, one page, every page or slide as a one-sentence title in order, with nothing built. Test: read only the titles in order and the argument must follow, with no page whose title is a noun phrase.
2. The built body, in the Studio, on the matching template. Board: 12 to 16 pages. Pitch: 10 to 13 slides. Test: every page passes the seven-rule check at 7 of 7, and the composite of machine and committed rubric self-score is at or above 0.80 on every page.
3. The appendix index, one page listing every appendix item with the body page it supports. Test: every body page that makes a numeric claim names its appendix item, and no appendix item is orphaned.
4. The computation, a script or spreadsheet producing every number on every page. Test: re-running it reproduces every printed figure to the stated rounding, and any total shown as a sum ties exactly.

After the build comes the note: 250 to 350 words on the other audience. For Brief A, what changes if this quarter's content goes to a prospective investor. For Brief B, what changes if it goes to a board that already owns 40% of the company. Name three things that move, at least one of which is the placement of bad news, and state one thing that must not change and why.

The rubric runs to twenty points, and the pass mark is 16 with no zero on any row.

Any artefact whose test fails scores zero on its rows, however the pack reads. A beautiful board pack whose bridge doesn't tie will be caught by the one director who checks, and that's a worse outcome than an ugly pack that ties.

Reading & resources

Core (do these).

• Barbara Minto, The Pyramid Principle: the structural spine behind the decisions page and the title stack. Read the introduction structures and the section on ordering ideas. Its deductive and inductive material repays a second pass once you've built two packs. [Paid] [Intermediate]
• Cole Nussbaumer Knaflic, Storytelling with Data: the closest published treatment of the declutter and chart-choice grammar applied throughout the worked examples. Its companion exercise volume is the better buy if you want repetitions more than principles. [Paid] [Beginner]
• Gene Zelazny, Say It With Charts: the original message-to-form mapping, and the shortest route to the habit of choosing the chart from the sentence rather than from the data shape. [Paid] [Beginner]
• Stephen Few, Information Dashboard Design: Few argues that a dashboard is a monitoring instrument and not a display, and the dashboard section above is a compressed application of that argument. His treatment of what earns space is more rigorous than the decision test used here. [Paid] [Intermediate]

On boards and what they need.

• Ram Charan, Dennis Carey and Michael Useem, Boards That Lead: the clearest treatment of where a board should decide, where it should monitor and where it should stay out, which is the distinction the decisions-first structure encodes. [Paid] [Intermediate]
• Colin Carter and Jay Lorsch, Back to the Drawing Board: older and more sceptical, and useful for understanding why so many packs are written as updates. [Paid] [Advanced]
• The Companies Act 2013, sections 173 and 179, and the SEBI Listing Obligations and Disclosure Requirements Regulations: the statutory floor for Indian board meetings and for what must be placed before a board, including the minimum number of meetings a year and the maximum gap between two. Read the sections themselves and skip the summaries. Check the current text on the Ministry of Corporate Affairs and SEBI sites, since both are amended regularly (checked September 2026, confirm before relying on any threshold). [Free] [Advanced]
• The National Association of Corporate Directors' published guidance on board materials (United States): practitioner guidance on pre-read timing and pack length. Availability and access tier change from year to year, so check the current publication list and don't rely on a cached copy (checked September 2026). [Paid] [Intermediate]

On pitching and the ask.

• Brad Feld and Jason Mendelson, Venture Deals: not a deck book, yet the best preparation there is for the ask slide, because it explains what the reader on the other side is optimising and therefore what your milestone has to buy. [Paid] [Beginner]
• Nancy Duarte, Resonate: the strongest published treatment of narrative structure in presented material, and the right corrective if your pitch is data-led and lifeless. Use it for the shape of the argument, and look elsewhere for slide aesthetics. [Paid] [Intermediate]
• Alberto Cairo, The Truthful Art: how to stop a chart claiming more than the data supports, which is the failure a market slide is most prone to. [Paid] [Intermediate]

Reference and practice.

• Edward Tufte, The Visual Display of Quantitative Information: the source of the data-ink argument that the declutter rule compresses. Read it once for the argument and keep it for the plates. [Paid] [Advanced]
• Published seed and Series A decks released by the founders who raised on them: a large number of these are public, and reading twenty in an evening teaches the seven-question order faster than any book. Read them for structure and page order. Treat every market slide in them as an exercise in the check taught above, and never as a model. [Free] [Beginner]

Teach it back & journal

Teach it back. Find someone who has never sat on a board and never raised money, and give them ten minutes. Start with the two artefacts side by side and the single question that separates them: what is the reader's default answer before they open it? A director's default is yes, because they already own the outcome and are looking for what changed and what needs deciding. An investor's default is no, and every slide is being read for a reason to stop. Show how that one difference reorders everything downstream. Then take one number through the whole chain. Revenue up 21.92% and contribution down 1.87% is the fact. The bridge from 24.60% to 19.80%, in five steps that sum to −4.80 points, is the explanation. The price-rise decision, with its 21.82% break-even volume loss, is what the board does about it. If your listener can restate why the bridge has to tie before the decision row means anything, the idea has landed. If they can't, the gap is almost always that you presented the decomposition as a chart when it needed to be an argument. That's the same failure the original pack made.

For the journal, two prompts, twenty minutes.

First, find a deck you built in the last year for someone who was deciding something. Count the pages before the first decision appears, and count the pages that no decision consumed. Then write the decisions page that deck should have opened with, in the five-column form, and for each row write the number that would have flipped your own recommendation. If you can't name that number now, note whether you knew it then.

Second, write down the worst piece of news you've ever put in a document for a senior audience, and where you put it. Then write the version that would have gone in the body, with its size, its decomposition, the action already taken and the trigger that would make it worse. Note in one sentence what stopped you the first time, and be truthful about it. Usually the good news wasn't yet strong enough to carry it. That makes it a scheduling problem, and you can't write your way out of it. It's the thing to watch for in your own drafts from here on.