The 1–2 Hour Teardown

Phase 8 · Week 65 · ~20 hours · CROWN MODULE

Where it sits. The second module of the signature-skill phase, and the first place the whole program converges into a single repeatable act. M8.01 gave you the raw motor skills: skim a 10-K or Indian annual report in the right order in 45–60 minutes, run the Ctrl-F trick list ("related party," "pledge," "contingent," "going concern," "restated," "exceptional," "guarantee," "receivable"), and assemble the source stack in order. That stack is the company's own "what we do" (10-K Item 1 / MD&A + segment notes / the investor deck), the last one or two earnings-call transcripts, ten years of numbers on one screen (screener.in or Tijori for India, TIKR/Koyfin/stockanalysis.com for the US), and one competitor deck plus one bear note to borrow and verify the map. This module tells you what to do with all that once it is open: run the twelve-question teardown checklist, run the moat quick-test, and produce the eight-line output, which is the finished thing. You have been building toward this since week 4 (Section 5.2's company-of-the-week has trained ~57 reps of progressively harder reads); M8.02 makes the method explicit, complete, and fast enough to finish in one to two hours. M8.03–M8.04 take the same twelve questions and spend two to three days on them (the deep dive); M8.05 puts you under a stopwatch on companies you have never seen and certifies the skill. This is the crown gate for the teardown itself, at ≥90% rather than the usual 85%, because a teardown that is 80% right is worse than none: it gives you the confidence to act on the 20% that is wrong.

Suggested hour budget. Core content (the twelve questions, the moat quick-test, reversing the multiple, the eight-line output): 8h · the two full worked trails (Titan, Costco): 3h · additional worked examples: 1.5h · practice set (twelve teardowns, ramped): 4.5h · flashcards + mastery check: 3h. Do at least three of the practice teardowns against a running clock; speed is a graded outcome here, not a nicety.

App link. This module is the timed 1-hour teardown mode and the "analyze this business" case engine in the Rapid-Analysis branch. The app presents a company's source stack (deck + transcript excerpt + ten-year panel), starts a 60-minute clock, and walks the twelve-question chain, auto-grading the objective parts (profit-engine identification, the ROIC-vs-hurdle call, the reverse-multiple arithmetic to a tolerance band, moat-source classification against ground truth) and rubric-scoring the eight-line output against a model answer. Your per-question accuracy and your finish time both feed the mastery heatmap. The eight-line output writes straight into your permanent company file (Section 5.3).

Learning objectives

You will be able to:

1. State the teardown's purpose and its limits. Produce a "good-enough-to-have-an-opinion" read (explain the business to a smart friend, judge whether it is a good business, name the three things that matter, and say whether the price looks demanding or generous) without mistaking it for a to-the-decimal valuation; the output of a teardown is a provisional verdict plus the one variant question worth a deep dive, not a target price.
2. Run all twelve teardown questions in order, inside 60–90 minutes, spending your minutes where the value is (the money machine and the price), not drowning in the parts that don't move the verdict.
3. Find the real profit engine. That is the revenue line, segment, or hidden subscription that actually produces the economics, and it is frequently not the biggest revenue line. Then walk ₹1 or $1 of revenue down to free cash flow, naming exactly where the money leaks.
4. Run the moat quick-test. Classify the candidate advantage into Dorsey's five sources, confirm or reject it in a high, stable-or-rising ROIC across a cycle, and refuse the four false moats (great product, hot management, high share, good execution).
5. Judge the economics against a hurdle. Decompose growth into volume versus price, read the margin structure, compare ROCE/ROIC to the ~12% INR / ~8–9% USD cost-of-capital hurdle, and check cash conversion (CFO/EBITDA, FCF/PAT) to confirm the profit is real.
6. Reverse the multiple. Read what growth, margin, or return the current P/E, EV/EBITDA, or P/B is implying, using back-of-envelope inversions, and judge that implied story against base rates and the moat evidence (question 5 of the five-questions spine, answered fast).
7. Write the eight-line teardown output. Business in one line, revenue engine, unit economics, moat (source + strength), economics verdict, the three things that matter, the two things that could break it, and the price verdict plus the one question to answer before acting. Defend every line from what you read.
8. Execute the full method end-to-end on a novel company of either market, producing a defensible eight-line output and the variant question, on the clock.

Prerequisites & connections

Builds on. M8.01, the source stack and the 45–60-minute filing skim; this module assumes you can already find segment notes, the cash-flow statement, the related-party note, and the shareholding pattern fast, and reads them for verdicts rather than navigation. Everything in Phases 1–7 is now a reflex you call in seconds: the three-statement linkage (M1.05) to walk revenue to cash; NOPAT and invested capital built operating-vs-financing (M2.03) to get ROIC honest; the cash-conversion ratios CFO/EBITDA and FCF/PAT (M2.04); the quality-of-earnings Ctrl-F reflexes (M2.06–2.07) so you don't build a thesis on numbers you shouldn't trust; the reverse DCF and the "what's priced in" discipline (M3.05); the four-lens moat protocol compressed to a quick-test (M4.04–M4.05); the sector KPI panels and the right valuation multiple per business model (Phase 5); and the macro overlay (Phase 7) for the one line about where the cycle sits. The teardown is not new knowledge. It is old knowledge under a stopwatch, in a fixed order.

Feeds forward. M8.03–M8.04: the deep dive is this same twelve-question spine given two to three days, and where the teardown notes a question mark, the deep dive resolves it (builds the ten-year model, computes incremental ROIC, runs the full reverse DCF, does the scuttlebutt and the proxy read). The teardown's job is to decide whether a company is worth those three days, and to hand the deep dive its starting agenda: the one variant question. M8.05 certifies the timed teardown on never-seen companies. M8.06 teaches the six documents beyond the annual report and the 10-K that the ninety minutes keeps reaching for, so a quarterly result, an investor deck, an 8-K, an ownership filing and a rating rationale each arrive with a known skeleton rather than as unfamiliar paper. Competency C3 ("understand a business model cold") and the front half of C10 ("run the 1–2 hour teardown") are built here; from week 62 onward the company-of-the-week runs in this exact format, and from week 65 it feeds a watchlist and a simulator action (Section 5.2).

The one-sentence version of this module. A teardown is twelve questions, asked in a fixed order and answered from a pre-assembled source stack in about ninety minutes, that end in eight lines: what the business is, whether it is any good, the three things that decide its future, the two that could break it, and whether today's price is asking for more than the evidence supports, plus the single question you would research next if the answer matters.

Core teaching content

8.2.4 The moat quick-test: Q6 in five minutes

In a teardown you cannot run Phase 4's four moat lenses and its full protocol. You run the quick-test: Dorsey's five sources, confirmed by one number, screened against four false positives. Where the position is contested, note that a deep dive would need the full M4.05 protocol.

Step 1 — the number first. Pull ten years of ROIC/ROCE from the panel (M2.03's build, or the screener's version cross-checked). A moat is a claim about the past that must already be visible in returns: high, stable-or-rising ROIC across a full cycle, comfortably above the ~12% INR / ~8–9% USD hurdle, and not collapsing in the bad years. No persistent spread over the hurdle → no moat, whatever the story → skip to a steady-state, no-franchise valuation (M4.05 §4.6). If the spread is there, something is protecting it, and Step 2 names it.

Step 2 — which of the five sources? Classify the protection into exactly one primary source (occasionally two interlocked, which is the strongest case):

The confirming logic is always the same: the source explains why the ROIC spread in Step 1 has survived competition, and the two must agree. A brand story with commodity margins is not a brand moat (M4.05's Mercedes point); a "network" where user N+1 doesn't benefit from user N is scale mislabelled.

Step 3 — reject the four false moats. These erode, and paying a moat multiple for them is the most expensive teardown error:

• A great product. This year's superior gadget/drug/app; products are copied and leapfrogged (the moat is what stops the copy, not the product).
• Hot management. A celebrated CEO; talent is mobile and mortal, and "bet on the jockey" is not a barrier to entry.
• High market share. Share is an outcome, not a source; it can be won and lost (M4.05: share without captivity is the IndiGo pattern, real skill on a thin moat).
• Good execution / operational excellence. Running faster on a level field earns a temporary bonus, not a durable spread; on a no-barrier field, efficiency is the only game and the CAP is near zero (M4.05 §4.6).

Step 4 — the strength call. State the moat as wide / narrow / none, and its direction: widening, stable, or eroding (Q5's share direction and Q10's threats feed this). The deliverable for Q6 is one line: "[wide/narrow/no] moat from [source], [widening/stable/eroding] — confirmed by [X]% ROIC over [N] years." In a teardown that is enough; if the position is contested or the CAP is doing heavy lifting in the price, that is exactly the kind of thing the variant question flags for the deep dive's full four-lens protocol and CAP estimate.

8.2.6 The eight-line teardown output

Every teardown ends in the same eight lines. This is the artifact you keep, the thing that goes in the company file, the format the app grades, and the discipline that forces the ninety minutes to converge on a verdict. Each line is one or two sentences, no more. A teardown that can't fit a line isn't finished; it's still reading.

1. Business in one line — what it does and who pays, jargon-free. 2. Revenue engine — the real profit engine (often not the biggest line) and how a rupee/dollar becomes cash. 3. Unit economics — the smallest repeatable unit and whether it earns a good return on what it costs. 4. Moat — source (one of five) + strength (wide/narrow/none) + direction, confirmed by the ROIC number. 5. Economics verdict — good / average / poor, with the one-number reason (ROIC vs hurdle across the cycle). 6. The three things that matter — the two-to-three variables that decide the future. 7. The two things that could break it — the specific, company-killing risks. 8. Price verdict + the one question — demanding / fair / generous vs quality, plus the single variant question to answer before acting.

Read lines 5, 8 together and you have the whole teardown in two sentences: is it a good business, is the price asking too much of it, and if I had to check one thing before betting, what would it be? Everything else is the evidence for those two sentences. The worked trails in §8.2.7–8.2.8 produce exactly this artifact, twice.

8.2.7 Worked trail: Titan Company (India), the full ninety minutes

(All figures illustrative, modeled on ~FY24/FY25 scale. Pull the actual annual report from titancompany.in, the segment note, and the screener.in ten-year panel; the lesson is the method and the shape of the numbers, not the decimals. Source stack assumed already open per M8.01: the FY annual report + investor presentation, the latest two earnings-call transcripts, the ten-year panel, one competitor read (Kalyan Jewellers' investor deck), and one bear note on jewellery competition.)

0–10 min · Orient (Q1–Q2). Titan is a Tata-group consumer company that sells jewellery, watches, and eyewear through ~2,000+ owned and franchised stores. One line per revenue line (from the segment note): Jewellery (Tanishq, Mia, CaratLane) ~₹44,000 cr, ~88% of revenue; Watches & Wearables (Titan, Sonata, Fastrack, smart) ~₹3,900 cr, ~8%; Eyewear (Titan Eye+) ~₹850 cr, ~2%; Emerging (Taneira ethnic wear, Skinn fragrances, accessories) ~₹1,000 cr, ~2% and loss-making in investment mode. The profit engine is unmistakable and, unusually, is the biggest line: jewellery is ~88% of revenue and ~90%+ of segment EBIT. This is a jewellery company with a beloved watch heritage attached. But the sharper Q1 read is within jewellery. Gold is a low-margin pass-through (Titan buys or leases gold and sells it on at a modest markup), and the real profit sits in making charges and the studded (diamond) mix, because a solitaire ring carries a far richer margin than a plain gold chain. So the true engine is not "jewellery revenue" but "making charges + studded mix on a growing branded volume." Q2, customers: hundreds of millions of Indian households, buying an aspirational-but-trust-critical product (wedding and daily-wear jewellery) on a recurring-over-a-lifetime, occasion-driven basis; the purchase is high-value, infrequent, and dominated by one fear: am I being cheated on purity and price?

Hold that fear. It is the whole moat.

10–35 min · The money machine (Q3, Q4, Q7). Q3, ₹1 to cash: of ₹1 of jewellery revenue, gross margin is deceptively slim because gold is a pass-through, but on the value-add (making + studded) margins are healthy; blended EBIT margin ~12–13% in jewellery, ~11% consolidated → ~₹0.11 of EBIT per ₹1. The interesting part is below EBIT. Jewellery is working-capital heavy (you hold inventory of gold and finished pieces), which would normally leak cash into inventory as the business grows, except that Titan funds that working capital with two clever floats: gold-metal loans (it borrows physical gold from banks, which simultaneously hedges the gold-price risk and finances inventory at a low rate) and customer-advance schemes (the monthly purchase plan, where customers pre-pay in installments for a future purchase, an interest-free float that also pre-sells inventory). So the working-capital leak that sinks ordinary jewellers is plugged by other people's gold and customers' advances.

Q4, unit economics: the smallest repeatable unit is one Tanishq store. A mature store does high revenue on a modest fixed cost, payback on a new store is a few years, and the incremental store earns a high return on capital because the inventory is substantially funded by gold-on-lease rather than equity. That is why Titan can grow the store count fast without a cash crunch.

Q7, economics: growth decomposes into volume (grammage + new stores) and value (gold price + studded mix + making charges). The healthy part is the volume and mix; the gold-price part is pass-through inflation, not real growth, and you strip it out mentally. Margins: stable-to-rising as studded mix and scale improve. ROCE ~22–25% for a decade, roughly double the ~12% INR hurdle, and it held through gold-price shocks and COVID: a durable, wide spread. Balance sheet: optically ~1.5× net-debt/EBITDA, but a chunk of that "debt" is hedged gold-metal loans, so the economic leverage is lower than it looks (a teardown catch; read the borrowings note, don't take the ratio raw). Cash conversion: lumpier than an FMCG because growth ties cash in inventory, so FCF/PAT runs below 1.0 in high-growth years, funded by the floats. Over a cycle it is acceptable but not pristine, so flag it. Economics verdict: good, bordering on excellent, with high, durable ROCE well above the hurdle and the one honest asterisk that cash conversion is working-capital-gated.

35–55 min · The franchise (Q5, Q6, Q8). Q5, competition and share: the Indian jewellery market is ~₹5–6 lakh crore and still ~60–65% unorganized (family jewellers), so the dominant game is organized taking share from unorganized, a long tailwind. Tanishq is the largest branded jeweller and has been gaining share for years while holding margin, a genuine right to win. But the competition is intensifying within organized: Kalyan Jewellers (listed, expanding fast on a franchise model), Malabar (large, unlisted), and Reliance Jewels (Reliance Retail's deep pockets) are all chasing the same formalization. Direction: gaining share at stable-to-rising margin so far; watch whether that continues as funded rivals arrive.

Q6, moat quick-test: the number first, ROCE ~22–25% for a decade, stable through cycles → a real spread to explain. The source is intangibles/brand, specifically trust in a fear-driven category: in jewellery the customer cannot verify purity or fair making-charge at the counter, so the known, assured brand wins by default (uncertainty reduction + search costs, per M4.05's Tanishq case). Tanishq built this with the Karatmeter (in-store purity testing), transparent pricing, buyback assurance, and two decades of consistency. It is a genuine intangibles moat, reinforced by retail scale and the switching-friction of the advance schemes. Reject the false moats: not "a great product," not "hot management," not merely "high share"; the moat is the trust attribution, and it shows up as pricing power and share gains against cheaper unorganized rivals. Strength: wide, historically stable, but now contested by well-funded organized entrants, so the direction is stable-to-slightly-narrowing at the margin, a deep-dive flag.

Q8, capital allocation: a decade of cash went into store expansion and inventory at high incremental ROIC (the core compounding engine), a modest and growing dividend, the CaratLane majority buyout (~₹4,600 cr in 2023, a debated price for the online-jewellery arm, worth scrutinizing), and ongoing investment in the loss-making emerging businesses (Taneira and Skinn, small drains to be watched). Overall: capital reinvested mostly at high returns in the core, a compounder's allocation, with two watch-items (the CaratLane price, the emerging-business burn).

55–70 min · People & risk (Q9, Q10). Q9, management/incentives quick pass: a Tata-group company (Tata Sons + Tamil Nadu's TIDCO as promoters, ~52%), professional long-tenured management, high governance standards, and, critically for India, no promoter-pledge or related-party-abuse red flags (the Tata umbrella is about as clean as Indian promoter structures get). Transcripts read candid: management discusses studded-mix softness and competition without spin. One line: aligned and candid; the only debated capital decision is the CaratLane buyout price.

Q10, the three real risks: (1) gold-price and demand elasticity, since a sharp gold spike suppresses volumes and shifts mix away from high-margin studded toward plain gold, hitting margin twice; (2) funded organized competition, where Reliance and Kalyan competing making-charges down could compress the very value-add that is the profit engine; (3) regulatory/discretionary, meaning tighter cash-purchase/PMLA rules, import-duty swings, and the fact that jewellery is ultimately discretionary and wedding-cycle-sensitive. These are the ones that could actually dent the thesis, not the boilerplate.

70–85 min · The price (Q11, Q12). Q11, reverse the multiple: Titan traded at roughly a ₹3 lakh crore market cap on ~₹3,500 cr PAT → ~85× earnings (illustrative, FY2023-24 scale; not a current quote). Reverse it (multi-stage): require r = 11%, assume a de-rate over N = 10 years to a still-premium PE_term = 35: g = 1.11 × (85/35)^0.1 − 1 ≈ 1.11 × 1.0928 − 1 ≈ 21.3%. Cross-check (Gordon): at payout ≈ 27%, g_implied ≈ 0.11 − 0.27/85 ≈ 10.7% perpetual, right at India's long-run nominal GDP, i.e., the price assumes Titan grows with the whole economy, forever, at current payout. Both reads say the same thing: the price is asking for ~21% earnings growth for a decade or economy-matching growth into perpetuity, essentially a continuation of Titan's best decade, with no allowance for the organized-competition drag now arriving.

Q12, cheap or expensive vs quality: this is a genuinely great business, with a wide trust moat, ~22–25% ROCE and a long formalization runway, so a premium multiple is deserved. But ~85× is demanding even for this quality, because it grants little room for the competitive intensification Q5 and Q10 flagged. What the method shows on these figures: demanding, priced for the moat holding and growth continuing at near-peak rates, a reading of the method rather than a view on the stock today.

85–90 min · The eight-line output.

1. Business: India's largest branded jeweller (Tanishq), a Tata company that also sells watches and eyewear, selling trust-assured jewellery to hundreds of millions of households. 2. Revenue engine: jewellery is ~88% of revenue and ~90% of profit; within it, gold is a pass-through and the real margin is making charges + studded (diamond) mix — funded by gold-on-lease and customer advances, so growth doesn't drown in working capital. 3. Unit economics: one Tanishq store — modest fixed cost, inventory largely financed by leased gold, high incremental ROCE, few-year payback → attractive and scalable. 4. Moat: intangibles/brand (trust) — wide, historically stable, now contested — a purity-and-price assurance moat in a fear-driven category, confirmed by ~22–25% ROCE for a decade against cheaper rivals. 5. Economics verdict: good, bordering excellent — ROCE ~2× the ~12% hurdle, durable through cycles; one asterisk: cash conversion is working-capital-gated (FCF/PAT < 1 in growth years). 6. The three things that matter: organized-vs-unorganized share gains continuing; studded/premiumization mix holding up the margin; store-count runway. 7. The two things that could break it: funded organized competition (Reliance/Kalyan) competing making-charges and the trust premium down; a gold-price/demand shock hitting volume and mix together. 8. Price verdict + the one question: demanding — ~85× implies ~21% EPS growth for a decade even after a de-rate, i.e., its best decade continues with no competitive drag. Variant question for the deep dive: is Tanishq's growth still genuine premiumization and share-gain, or is it increasingly being bought with making-charge discounting against Reliance/Kalyan? (Resolve that — in the studded-mix trend, per-store throughput, and gross-margin bridge — before acting.)

That is a complete teardown: a great business identified, its engine and moat named and confirmed in the returns, its risks made specific, and a price verdict that refuses to say "expensive" without the quality on the scale, ending in the one question a deep dive exists to answer.

8.2.8 Worked trail: Costco Wholesale (US), the full ninety minutes

(All figures illustrative, modeled on ~FY24 scale. Pull the actual 10-K from EDGAR, the membership and renewal disclosures, and the stockanalysis.com ten-year panel. Source stack open: 10-K Item 1 + the membership/renewal disclosures, the latest earnings call, the ten-year panel, one competitor read (Walmart/Sam's Club), and one bear note on the valuation.)

0–10 min · Orient (Q1–Q2). Costco runs ~890+ no-frills membership warehouses selling a tightly curated ~3,800 SKUs in bulk at rock-bottom markups. One line per revenue line: merchandise sales ~$250bn, ~98% of revenue, at a ~11% gross margin (versus ~25–30% for an ordinary grocer, because Costco charges almost nothing over cost); membership fees ~$4.8bn, ~2% of revenue. Now the Q1 punchline, which is that the real profit engine is not the biggest line: operating income is ~$9.3bn, and the ~$4.8bn membership fee is roughly half of it, at essentially 100% flow-through (the fee has almost no incremental cost). The merchandising operation, after its ~9%-of-sales SG&A, runs close to break-even by design: Costco deliberately sells goods at near-cost and makes its profit on the annual membership subscription. This is the canonical "profit engine ≠ revenue headline." Q2, customers: ~76 million paid member households (~137 million cardholders), buying everyday essentials and treasure-hunt discretionary on a recurring annual-subscription basis; ~46% are higher-tier Executive members who drive ~73% of sales. Why they buy: the member has pre-paid for the right to low prices, so every trip amortizes the fee and reinforces the habit. A subscription bolted onto a grocery run.

10–35 min · The money machine (Q3, Q4, Q7). Q3, $1 to cash: of $1 of merchandise revenue, ~$0.11 is gross profit and ~$0.09 goes to SG&A, leaving merchandising operating profit near ~$0.02, which is thin, and then the membership fee is added almost entirely to the bottom line. But the cash story is the star. Costco runs a negative cash conversion cycle: it turns inventory ~12× a year (sells it in ~30 days) while paying suppliers on ~30–45-day terms, so the supplier finances the inventory and growth releases cash rather than consuming it. Net income converts to more than 100% cash (CFO > net income routinely).

Q4, unit economics: the smallest repeatable unit is one warehouse (or, better, one member). A member pays ~$65 (Gold Star) or ~$130 (Executive) a year at ~100% margin and renews at ~90%+, so the LTV of a member is a long, cheap, high-margin annuity against a near-zero CAC (members acquire themselves for the value). That is one of the best unit economics in retail. A new warehouse ramps to high volume on low fixed cost and pays back fast.

Q7, economics: growth = new warehouses (~25–30/yr) + comparable-sales (mostly volume/traffic) + periodic fee increases + membership growth, overwhelmingly volume, the healthy kind. Margins: stable and slowly rising with scale and Executive-mix. ROIC ~20%+, comfortably above the ~8–9% USD hurdle, and stable across cycles (warehouse clubs are recession-resilient, and members trade down into Costco). Balance sheet: net cash / minimal net debt, investment-grade, a fortress. Cash conversion: excellent (negative working capital). Economics verdict: excellent, a high-return, recession-resilient annuity with cash conversion above 100%.

35–55 min · The franchise (Q5, Q6, Q8). Q5, competition and share: Costco competes with Walmart/Sam's Club, BJ's, Amazon, and grocers, but the warehouse-club model is a different game. Costco gains share of members' wallets steadily and its renewal rate is stable-to-rising, the cleanest possible sign of a durable right to win.

Q6, moat quick-test: number first, ROIC ~20%+, stable for a decade → a real spread. The source is primarily cost advantages interlocked with switching/membership loyalty: the tiny SKU count means gigantic volume per item → best-in-world buying power and lowest cost → lowest price, and the membership fee both funds the profit and locks the behavior. That is "scale economies shared": Costco caps its markup and passes savings to members, which drives renewal and volume, which deepens the cost advantage, the flywheel from M4.01/M4.05. Confirm: gross margin structurally below rivals by choice while ROIC stays high, and ~90% renewal is the switching-cost proof. Reject false moats: it is not "good execution" alone (the cost edge is structural, not just operational) and not "high share" as the source (the source is scale-driven cost + the membership lock). Strength: wide and stable-to-widening (each new warehouse and member deepens the scale advantage).

Q8, capital allocation: a decade of cash went into new warehouses at high returns (the core compounding engine), a steadily growing regular dividend plus periodic large special dividends (e.g., a ~$15/share special, a disciplined return of excess cash rather than empire-building), and modest buybacks. Exemplary, Outsiders-style capital allocation. One line: reinvests in the high-return core and returns the rest sensibly.

55–70 min · People & risk (Q9, Q10). Q9, management/incentives quick pass: professional management with a famously durable, member-first and employee-first culture (high wages and low turnover, which lower cost and theft and reinforce the flywheel); founder philosophy intact; candid, plain-spoken calls; aligned. No governance flags. Q10, the three real risks: (1) valuation itself, since the stock is priced for perfection (see Q11), so even a great business can be a poor investment here; (2) any crack in the membership value proposition or renewal rate, because the fee is the profit and a sustained renewal decline would hit the engine directly; (3) consumer/competitive, meaning a severe discretionary downturn or an Amazon/Walmart assault on the value gap, though the model is unusually recession-resilient. The specific, engine-level risk is #2.

70–85 min · The price (Q11, Q12). Q11, reverse the multiple: Costco trades around a ~$370bn market cap on ~$7.4bn net income → ~50× earnings (illustrative, and richer than almost any other large retailer on Earth). Reverse it: require r = 9%, assume a de-rate over N = 10 years to a still-premium PE_term = 25: g = 1.09 × (50/25)^0.1 − 1 = 1.09 × 1.0718 − 1 ≈ 16.8%. Cross-check (Gordon): at payout ≈ 28%, g_implied ≈ 0.09 − 0.28/50 ≈ 8.4% perpetual, roughly twice US long-run nominal GDP, forever. Both reads: the price implies ~17% EPS growth for a decade (against a ~9–11% historical rate) or a permanently elevated multiple.

Q12, cheap or expensive vs quality: this is one of the finest businesses in retail, with an annuity profit engine, a structural cost moat, ~20% ROIC, a fortress balance sheet and exemplary capital allocation, so a premium multiple has a basis. The ~50× multiple prices an acceleration the operating history doesn't show, so the question to carry into a deep dive is whether that acceleration is coming: the classic "wonderful company, but is the multiple leaving me anything?"

85–90 min · The eight-line output.

1. Business: a membership warehouse club — ~890 no-frills stores selling ~3,800 bulk SKUs at near-cost to ~76 million member households. 2. Revenue engine: merchandise is ~98% of revenue at ~11% margin and near break-even after costs; the ~2% membership-fee line is ~half of operating profit at ~100% flow-through — the retail runs at cost so that the subscription pays. 3. Unit economics: one member — ~$65–130/year at ~100% margin, ~90%+ renewal, near-zero CAC → a long, cheap, high-margin annuity; one warehouse ramps fast on low fixed cost. 4. Moat: cost advantage (scale) interlocked with membership switching — wide and stable-to-widening; tiny SKU count → best buying power → lowest price → renewal → more scale (the flywheel), confirmed by ~20%+ stable ROIC and ~90% renewal. 5. Economics verdict: excellent — ~20%+ ROIC vs an ~8–9% hurdle, recession-resilient, negative working capital so cash conversion exceeds 100%. 6. The three things that matter: membership renewal rate and growth; warehouse-count runway (US and international); the periodic fee increase (high-margin, near-100% flow-through). 7. The two things that could break it: a sustained renewal-rate decline (it hits the profit engine directly); the valuation itself de-rating. 8. Price verdict + the one question: demanding — ~50× implies ~17% EPS growth for a decade or a permanent premium multiple, above the ~9–11% history. Variant question for the deep dive: how durable is the ~90% renewal rate and the fee-increase runway through a genuine recession — i.e., is the membership annuity as unbreakable as the multiple assumes? (Resolve renewal-rate behavior across the 2008/2020 stress windows before acting.)

Notice the deliberate mirror between the two trails. Both are wonderful businesses trading at demanding prices, but they teach opposite mechanics of the same skill: Titan hides its profit inside its biggest line (making charges + studded mix within jewellery) and funds working capital with customer advances and leased gold; Costco hides its profit outside its biggest line (the membership fee, not merchandise) and funds working capital with supplier float. Both moats are real but of different families: Titan's is trust/intangibles in a fear-driven category, Costco's is scale-cost interlocked with a subscription lock. And in both, the teardown's job ended identically: name the business and its engine, confirm the moat in the returns, make the risks specific, judge the price against the quality, and hand the deep dive one sharp question.

That is the pattern you are drilling until it is automatic.

Common mistakes & how experts think differently

1. Reaching for a valuation model in hour one. The beginner opens a spreadsheet and starts forecasting revenue; the expert knows a model built on forty minutes of skimming is false precision, and that the teardown's job is a verdict and a variant question, not a per-share number. If you catch yourself building a DCF during a teardown, stop. You have confused this module with M3.08. The price work here is reversing the existing multiple (Q11), never building a new one.

2. Describing the revenue mix and calling it the business. "It's a diversified consumer company with jewellery, watches, and eyewear" is a description, not an analysis. The expert's Q1 always ends on the profit engine, the one line that carries the economics, and frequently that engine is not the biggest revenue line (Costco's membership fee, ITC's cigarettes) or is hidden inside the biggest line (Titan's making-charges-and-studded-mix within jewellery). If you can't name the single line that keeps the returns high, you haven't done Q1.

3. Reading one year instead of a cycle. A single great year of margins and ROIC is a data point, not a verdict; the whole point of the ten-year panel (M8.01) is to see whether the returns persist through the bad years. The expert's economics verdict (Q5, the good/average/poor call) is a statement about durability across a cycle, which is exactly why a cyclical's peak year is a trap, not a triumph (see Worked Example 2).

4. Confusing a low multiple with cheap and a high multiple with expensive. The multiple means nothing without the quality on the other side of the scale. A 6× P/E on a price-taking cyclical at peak earnings is dear; a 50× P/E on a durable mid-teens compounder may be merely full. The expert never says "expensive" or "cheap" as a standalone word. Q12 is always price-relative-to-quality, and the reverse-multiple (Q11) is what makes "relative to quality" precise.

5. Accepting "sticky," "loyal," "trusted," or "leading" without a mechanism. These are conclusions dressed as evidence. The expert demands the mechanism: stickiness must cash out as a switching cost, a habit, a contract, or a network; a moat must show up as a high, persistent ROIC and a named source; leadership must survive the question "leader of what market?" A teardown full of adjectives is a teardown that read the investor deck and believed it.

6. Taking the cash-flow statement's word for granted, or never opening it. The P&L is where management tells its story; the cash-flow statement is where the story is audited. The expert always checks CFO/EBITDA and FCF/PAT (Q7's cash-conversion read) because a business can report rising profit while leaking every rupee into receivables, inventory, or capex, and the divergence is often the earliest red flag (M2.04, M2.07). If profit and cash disagree, believe the cash.

7. Listing generic risks instead of company-killing ones. "Competition, regulation, and macro conditions could affect the business" is true of every company and therefore informs nothing. The expert's Q10 names the two or three risks that could appear in this specific company's obituary: a key plant's USFDA import alert, an NBFC's ALM mismatch, a moat under a funded attack, each with a one-line mechanism. If your risk list would fit any company, you haven't found the real risks.

8. Being charmed by the CEO. A charismatic founder on a podcast is not evidence of alignment; the transcript's candour, the pay metrics, the ownership, and, in India, the promoter pledge percentage outrank the narrative every time. The expert reads the incentives and the related-party note before believing the story, because "show me the incentive and I'll show you the outcome" (Munger) is more predictive than any amount of vision.

9. Spending equal time on all twelve questions. The novice gives ten careful minutes each to twelve questions and runs out of clock before the price. The expert front-loads the money machine and the price (two-thirds of the value) and spends ninety seconds on the questions that rarely move the verdict. The budget in §8.2.2 is not a suggestion; disciplined time allocation is the difference between a finished teardown and an abandoned one.

10. Forgetting the variant question, and ending on a verdict instead of a question. A teardown that ends "good business, fair price" has stopped one step short. The expert's last act is naming the single thing that, if researched, would confirm or destroy the read, because that question is the entire bridge to the deep dive, and an analyst who can't name it is an analyst who doesn't yet know what they don't know. The variant question is the teardown's most valuable output, not its afterthought.

11. Letting the price you already know poison the analysis. If you glance at the multiple first, a cheap stock makes you generous about a weak business and an expensive one makes you dismissive of a great one. That is anchoring in its purest, most expensive form. The expert runs the entire quality assessment (Q1–Q10) before looking hard at the price (Q11–Q12); the order in §8.2.2 exists precisely to quarantine this bias.

Worked examples

Worked Example 1: Visa (US): the thirty-minute compressed teardown of a capital-light network

(Figures illustrative, ~FY24 scale; pull the 10-K. This shows the teardown compressed to its essentials on a very different business model from Costco: a capital-light toll-road, to prove the method travels.)

Orient (Q1–Q2). Visa operates the rails that move money between banks when you tap a card; it earns fees on payments volume (~$16 trillion moved a year), data-processing, and cross-border transactions. Visa takes no credit risk: the issuing bank lends, Visa just tolls the transaction. Revenue ~$36bn. Q1 engine: it's a toll on money movement, and nearly all incremental revenue drops to profit because the network is already built. Q2: two customer sets in a two-sided network (cardholders and merchants, intermediated by banks); the product is a must-have utility used millions of times a day, entirely recurring.

Money machine (Q3, Q4, Q7). Of $1 of revenue, ~$0.65 is operating profit, an operating margin ~65%, because the marginal transaction costs almost nothing to process. Capex is trivial; this is one of the most capital-light business models on Earth, so FCF ≈ net income and both are enormous (~$20bn net income, ~55% net margin). Unit economics (Q4): the "unit" is a transaction with near-zero marginal cost, so contribution margin is ~100% and every incremental dollar of volume is almost pure profit. Q7: growth = secular shift from cash to digital + volume + pricing → ~low-double-digit revenue growth, mostly volume (healthy); ROIC is effectively off the chart (little capital employed against huge profit), vastly above the ~8–9% USD hurdle. Economics verdict: exceptional.

Franchise (Q6). Moat quick-test: the number, sustained ~65% margins and enormous ROIC for years. The source is network effects (more cardholders → more merchants accept Visa → more valuable to the next cardholder), reinforced by scale and brand, in a duopoly with Mastercard. This is a textbook wide network-effect moat, stable-to-widening as digital payments grow. Not a false moat: the returns are structural, not executional.

Price (Q11–Q12). Visa trades around ~30× earnings (illustrative). Reverse it: r = 9%, de-rate over N = 10 to PE_term = 20: g = 1.09 × (30/20)^0.1 − 1 = 1.09 × 1.0413 − 1 ≈ 13.5%. Visa grows revenue ~10% and EPS ~mid-teens (buybacks help), so the implied ~13.5% is roughly in line with the business's actual trajectory, and the price is full but not stretched, unlike Costco/Titan. Q12 verdict: a superb business at a fair-to-full price, because the multiple isn't demanding a heroic acceleration, just a continuation. Risk (Q10): regulatory interchange caps, and disruption from account-to-account rails (India's UPI is the live example of exactly this threat, a real-time bank-to-bank system that bypasses card rails; watch whether it stays domestic or the model spreads). Variant question: does the rise of real-time A2A payments (UPI-style) and stablecoins erode the card networks' take rate over the next decade, or do Visa's value-added services replace the lost toll? Lesson of this example: a wide moat at a fair price can be a better setup than a wider moat at a demanding price, and Q11 is what tells them apart.

Worked Example 2: "Meridian Metals" (synthetic cyclical): the teardown that catches a value trap

(Fully synthetic, with clean numbers, to isolate the cyclical trap. This is the case where the raw multiple screams "cheap" and the teardown says "trap.")

Meridian Metals is a mid-cost aluminium producer. Price ₹600; trailing EPS ₹100 → trailing P/E = 6×. A screen flags it as "deep value." Run the teardown.

Money machine (Q7). Decompose growth: revenue is up because aluminium prices are near a cycle high. This is price, not volume, and price in a commodity is the cycle, not a durable trend (M5.09). Margins are at a cyclical peak. ROIC swings violently: ~30% at this peak, but it was −5% in the last trough three years ago. No persistent spread, so by the moat quick-test, no moat: Meridian is a price-taker sitting in the third quartile of the cost curve (survives a downturn but isn't first to). Balance-sheet landmine (the M8.01 Ctrl-F pays off): net debt has risen as the company funds a capacity expansion, which is debt-funded expansion at the top of the cycle, the classic capital-cycle error (M4.06, M5.09).

Price (Q11–Q12). Here is the trap, made explicit. The 6× P/E is on peak earnings. Normalize to mid-cycle: suppose mid-cycle EPS is ~₹40 (peak margins fade as the price mean-reverts). Then the mid-cycle P/E = 600/40 = 15×, and at the trough, EPS could be near zero or negative, so the trough P/E is meaningless or infinite. The rule from M5.09 crystallizes: for a cyclical, the lowest trailing P/E often marks the top, not a bargain, because the market correctly refuses to capitalize peak earnings. Q12 verdict: at 15× mid-cycle with no moat and debt rising into the peak, Meridian is not cheap. It is a value trap, and the "6×" was the market pricing earnings it knows won't last. Correct teardown output: poor-to-average business (no moat, price-taker); the apparent cheapness is peak-cycle illusion; value it on mid-cycle earnings and a mid-cycle multiple, never trailing. Variant question: where are we in the aluminium cost/price cycle, and what is Meridian's true through-cycle mid-point EPS? Lesson: the teardown's Q7 (volume vs price, ROIC across the cycle) and Q11 (normalize before you multiply) are what stop a cyclical's peak-earnings from masquerading as value.

Worked Example 3: "Sunrise Finance" (synthetic NBFC): swapping the lens, and the P/B reverse

(Synthetic, to show the sector-lens swap for a lender, where P/E lies and P/B rules, inside a teardown.)

Sunrise Finance is a fast-growing NBFC: AUM growing ~25%/yr, ROE ~20%, trading at P/B = 3.0× and ~15× earnings. A generalist calls the 15× P/E "reasonable for 25% growth." The teardown corrects them on two counts.

Lens swap (Q7). For a lender, P/E lies: a bank can flatter near-term EPS by under-provisioning for bad loans or over-levering, both of which raise the risk that the P/E ignores. So you swap to the sector lens (M5.01–M5.02): read asset quality (GNPA/NNPA, credit cost, the SMA buckets), capital (CRAR), funding (ALM mismatch, CP reliance, the IL&FS/DHFL failure mode), and returns (ROA, ROE). The teardown reads: 25% AUM growth is fast, so is it quality growth or a book being bought with loose underwriting that will show up as credit costs in two years?

The P/B reverse (Q11). From M3.06, justified P/B ≈ (ROE − g)/(r − g), so invert for the ROE the price assumes: implied ROE = g + P/B × (r − g). At r = 13% (cost of equity), g = 8% (sustainable long-run): implied ROE = 0.08 + 3.0 × (0.13 − 0.08) = 0.08 + 0.15 = 23%. Read: at 3× book, the market is assuming a ~23% sustainable ROE, but Sunrise earns ~20% and is at a size where growth (and thus ROE) is likely to slow. So the price is demanding: it needs the ROE to rise even as the base matures. Q12 verdict: a good-but-not-exceptional lender at a demanding price, with the added tail risk that P/E-based comfort is hiding balance-sheet fragility. Variant question: is the 25% AUM growth being underwritten at stable credit cost and matched funding (check ALM and slippage), or is it a spread-and-quality sacrifice that a downturn will expose? Lesson: for lenders, the teardown swaps P/E for P/B-vs-ROE and reads the balance sheet as the business. The reverse-P/B turns the multiple into a testable ROE claim, and the risk always lives in funding and asset quality.

Worked Example 4: "Griha Foods" (synthetic, India ₹): running the whole kit to a "price isn't the problem" verdict

(Fully synthetic, clean numbers, INR hurdle 12%. Every teardown above ended in "great business, demanding price"; this one runs the identical eight-line method on a kit and lands on the rarer green-ish verdict, to prove the method can say "fair," not only "expensive." All figures Python-verified.)

The kit (one screen, per M8.01), ₹ crore. Revenue 4,000 (Packaged Foods 3,600 ≈ 90%, Food-Services/B2B 400 ≈ 10%); gross profit 1,800; EBIT 560; D&A 140; interest 60; tax 25%; PAT 375; invested capital 2,100; net debt 350; CFO 595; capex 175. Market cap ₹9,000 cr (P/E 24×); payout 35%.

Orient (Q1–Q2). Griha sells branded kitchen staples (atta and packaged foods) to hundreds of millions of Indian households through a distributor network; a small B2B/food-services line rides the same factories. The profit engine is branded packaged foods: ~90% of revenue and, at a far richer margin than the near-commodity B2B line, essentially all the profit. Customers: millions, buying a must-have, habitual, recurring staple, non-discretionary and frequent, the healthy kind of demand.

Money machine (Q3, Q4, Q7). Walk ₹1: gross margin = 1,800/4,000 = 45%; EBIT margin = 560/4,000 = 14% → ~₹0.14 of EBIT; NOPAT = 560 × 0.75 = ₹420 cr. ROIC = 420/2,100 = 20.0%, a wide +8-point spread over the 12% hurdle. EBITDA = 560 + 140 = ₹700 cr; CFO/EBITDA = 595/700 = 85%; FCF = 595 − 175 = ₹420 cr; FCF/PAT = 420/375 = 112%, so the profit is fully cash-backed (negative working capital from distributor advances and fast staple turns releases cash). Leverage is light: net debt/EBITDA = 350/700 = 0.5×. Unit (one SKU through one distributor): high contribution margin, negligible incremental capital, cash-positive from day one. Economics verdict: good, a 20% ROIC near double the hurdle, durable and cash-backed.

Franchise & people (Q5, Q6, Q8, Q9, Q10). Moat: intangibles/brand (habit + trust in an adulteration-fearful category) plus regional distribution scale, confirmed by the 20% ROIC: narrow-to-wide, stable. Reject the false moats (not "a great product" or "high share"; the source is trust-attribution and shelf reach). Capital allocation (Q8): reinvested in the core at high incremental returns plus a steady 35%-payout dividend. Q9: founder-family promoter, no pledge, candid calls, so aligned. Q10 risks: (1) a funded national FMCG/D2C attacker competing the brand premium down; (2) an adulteration/food-safety scandal that breaks the trust; (3) a wheat-cost spike the brand can't fully pass through.

Price (Q11–Q12). Reverse the 24× P/E: require r = 11%, de-rate over N = 10 years to a still-premium PE_term = 18×: g = 1.11 × (24/18)^(1/10) − 1 = 1.11 × 1.0292 − 1 ≈ **14.2%**. Gordon cross-check at 35% payout: g ≈ 0.11 − 0.35/24 ≈ **9.5%** perpetual, below India's ~10–11% nominal GDP, so the price does not demand the firm outgrow the economy forever. Griha actually grows ~13–14% (mid-single-digit volume + price + new categories), so the implied ~14% is roughly what the business can deliver. Q12 verdict: a good business at a roughly fair price, the rare teardown where the multiple is not the problem.

The eight-line output.

1. Business: a branded kitchen-staples maker (atta + packaged foods) selling must-have, habitual groceries to hundreds of millions of Indian households. 2. Revenue engine: branded packaged foods (~90% of revenue, ~all profit); ₹1 → 45% gross → 14% EBIT → cash, with negative working capital so growth self-funds. 3. Unit economics: one SKU through one distributor — high contribution margin, negligible incremental capital, cash-positive immediately. 4. Moat: intangibles/brand (trust + habit) + regional distribution scale — narrow-to-wide, stable — confirmed by 20% ROIC vs a 12% hurdle. 5. Economics verdict: good — ROIC ~1.7× the hurdle, durable, FCF/PAT 112%, net debt/EBITDA 0.5×. 6. The three things that matter: volume (not just price) growth; share gains vs unorganised + new-geography expansion at core returns; premiumisation mix. 7. The two things that could break it: a funded national/D2C attacker competing the brand premium down; a food-safety scandal that snaps the trust. 8. Price verdict + the one question: roughly fair — 24× implies ~14% growth (≈ what it delivers) and below-GDP perpetual growth. Variant question: is the growth genuine volume + share gain, or increasingly price-led on flat volume (the FMCG warning sign)? Resolve the volume-vs-value split before acting.

Lesson: run the identical kit and the verdict can be "fair," not "expensive." Q12 is always price relative to quality, and when the reverse-multiple's implied growth matches what a genuinely good business can deliver, the multiple is doing no heavy lifting.

Worked Example 5: "Ridgeline Brands" (synthetic, US $): the leverage that flatters the ROE

(Fully synthetic, clean numbers, USD hurdle 9%. This kit teaches the most common quality misread in a teardown, a fat ROE that is mostly borrowed. The one-line rule: read ROIC, not ROE. All figures Python-verified.)

The kit (one screen), $m. Revenue 8,000; EBIT 960; D&A 240; interest 240; tax 25%; PAT 540; invested capital 6,000; net debt 3,600; book equity 2,400; CFO 900; capex 480. Market cap $9,720m (P/E 18×); payout 30%. The headline the sell-side loves: ROE 22.5%.

Orient (Q1–Q2). Ridgeline is a mature branded household-and-personal-care company: a portfolio of shelf staples sold to mass retailers and, through them, to tens of millions of households; recurring, low-ticket, mildly discretionary. Nothing wrong with the demand. The question the teardown must settle is how good the returns really are.

Money machine (Q3, Q4, Q7). EBIT margin = 960/8,000 = 12%; NOPAT = 960 × 0.75 = $720m. Now the number that decides everything: ROIC = 720/6,000 = 12.0%. Against a 9% hurdle that is a thin +3-point spread, not the +13 the 22.5% ROE seems to promise. Where did the fat ROE come from? Leverage: invested capital 6,000 = net debt 3,600 + equity 2,400, so net debt/EBITDA = 3,600/1,200 = 3.0×, and ROE = PAT 540 / equity 2,400 = 22.5%. The gap between 22.5% and the honest 12% ROIC is the borrowing. Cash conversion is soft: CFO/EBITDA = 900/1,200 = 75%; FCF = 900 − 480 = $420m; FCF/PAT = 420/540 = 78%, both below the ~80% healthy bar, and capex = 2.0× D&A, heavy reinvestment at only a 3-point spread. Economics verdict: average dressed as good, with a genuinely modest ROIC, mediocre cash conversion, and a balance sheet doing the flattering.

Franchise & people (Q5, Q6, Q8, Q9, Q10). Moat quick-test: the number first. A 12% ROIC only three points over the hurdle is a weak, easily-arbitraged spread, so narrow at best (some brand habit), possibly none; the "beloved brands" story outruns the returns. Q8: much of the "growth" is capital retained at a thin spread plus debt-funded M&A, value-neutral at best (M4.05's I × (ROIC − WACC) barely positive). Q9 nothing alarming. Q10 risks: (1) the 3.0× leverage itself, since a downturn in a mildly-discretionary category strains fixed interest; (2) private-label and D2C compressing the brand premium; (3) the reinvestment (2× D&A) failing to earn its keep.

Price (Q11–Q12). Reverse the 18× P/E: r = 9%, de-rate over N = 10 to PE_term = 13×: g = 1.09 × (18/13)^(1/10) − 1 ≈ **12.6%**. Gordon at 30% payout: g ≈ 0.09 − 0.30/18 ≈ **7.3%** perpetual, above US ~4–5% nominal GDP, so the price already asserts the firm outgrows the economy forever. A mature staples business growing ~5% cannot credibly deliver 12.6% EPS for a decade. Q12 verdict: demanding, and the quality is weaker than the ROE advertised. The multiple is pricing an acceleration onto a thin-spread, levered base.

The eight-line output.

1. Business: a mature branded household/personal-care portfolio sold through mass retail to tens of millions of households. 2. Revenue engine: shelf staples across the portfolio; $1 of revenue → ~12% EBIT → cash, but heavy capex (2× D&A) and interest on 3.0× leverage skim the free cash. 3. Unit economics: one brand/SKU — decent contribution margin, but incremental capital earns only ~12%, barely over the hurdle. 4. Moat: intangibles/brand at most narrow (and contested) — the 12% ROIC, only +3 over the hurdle, does not confirm a wide moat. 5. Economics verdict: average dressed as good — ROIC 12% (not the 22.5% ROE, which leverage flatters); CFO/EBITDA 75%, FCF/PAT 78%, net debt/EBITDA 3.0×. 6. The three things that matter: whether reinvestment (2× D&A) starts earning a real spread; brand-premium defence vs private label/D2C; deleveraging. 7. The two things that could break it: the 3.0× leverage in a discretionary downturn; a private-label/D2C assault on the price gap. 8. Price verdict + the one question: demanding — 18× implies ~12.6% EPS growth for a decade (and above-GDP perpetual growth) on a business growing ~5%. Variant question: strip the leverage — what is the unlevered, ROIC-based earning power, and is any of the reinvestment earning above the hurdle? Resolve that before mistaking a levered ROE for quality.

Lesson: read ROIC, not ROE. A 22.5% ROE on a 12% ROIC is a borrowing story, not a quality story; leverage amplifies the good years and the bad ones worse (M0.01's Cascade Tools), and a fat ROE paired with a high multiple can hide a thin-spread business.

Practice set

Problems 1–4 are guided (scaffolded), 5–9 independent, 10–12 exam-condition (timed as marked, closed book). Currency ₹ crore unless stated; INR hurdle 12%, USD hurdle 9%, unless stated. All companies synthetic unless named.

P1 (guided: find the real profit engine). "Bharat Diversified" reports three segments. Chemicals: revenue ₹6,000, EBIT ₹300, capital employed ₹4,000. Consumer Brands: revenue ₹2,000, EBIT ₹500, capital employed ₹1,250. Infrastructure/EPC: revenue ₹4,000, EBIT ₹200, capital employed ₹3,000. (a) Which is the real profit engine? (b) What is the Q1 trap, and what does it imply for Q8?

Solution. (a) Compute segment ROCE: Chemicals 300/4,000 = 7.5%; Consumer Brands 500/1,250 = 40%; EPC 200/3,000 = 6.7%. Consumer Brands is only ~17% of revenue but 50% of EBIT (500 of 1,000) and the only segment earning above the 12% hurdle. It is the engine and the sole value creator. (b) The trap is calling this "a chemicals company" because chemicals is the biggest revenue line; it is really a consumer-brands compounder wrapped in two value-destroying cyclicals (both earning below their cost of capital). For Q8 this is the decisive capital-allocation question: is management starving the 40%-ROCE jewel to feed the 7%-ROCE segments? If so, growth is destroying value (M4.05's ROIC < WACC); a sum-of-the-parts would value the brands richly and the rest near or below book.

P2 (guided: walk ₹1 to FCF, find the leak). A company reports: EBIT ₹1,500; tax rate 25%; D&A ₹300; change in receivables +₹400; change in inventory +₹500; change in payables +₹200; capex ₹600. (a) Compute free cash flow (NOPAT + D&A − ΔWC − capex). (b) What is FCF/NOPAT, and where is the biggest leak?

Solution. (a) NOPAT = 1,500 × (1 − 0.25) = ₹1,125. Net working-capital change = receivables 400 + inventory 500 − payables 200 = ₹700 outflow. FCF = 1,125 + 300 − 700 − 600 = ₹125. (b) FCF/NOPAT = 125/1,125 = ~11%, dreadful conversion. The biggest leak is working capital (₹700): growth is being swallowed into receivables and inventory faster than payables fund it. The teardown flag: this profit is not becoming cash. Before trusting the economics, ask whether the WC build is a one-off (a new plant stocking up) or a structural feature (a business that must fund its customers), and check whether DSO/DIO are rising (a possible M2.07 red flag) or merely scaling with revenue.

P3 (guided: moat quick-test, including a rejection). Classify each and give the verdict. Company X: gross margin ~65% stable for 10 years; ROCE ~30% stable; sells a branded packaged food; raises price ~5%/yr with no volume loss. Company Y: ROCE jumped to ~28% in the last 2 years (it was ~11% for the eight before); makes "the highest-rated product in its category"; celebrated founder-CEO; market share rising fast.

Solution. Company X: wide moat, intangibles/brand (habit). The number confirms it: ~30% ROCE persistent across a decade plus demonstrated pricing power (raising price without losing volume is the cleanest pricing-power proof). Source and number agree → real moat. Company Y: reject, no confirmed moat (yet). The setup is three of the four false moats stacked (a great product, hot management, and rising share), and the ROCE has been high for only 2 years, which is not "across a cycle." The recent returns could be a product cycle that competitors will attack; there is no durable source and no persistence. Correct teardown treatment: grade it "unproven — watch whether ROCE persists 3–5 more years and name the barrier that would make it stick," and do not pay a moat multiple for it.

P4 (guided: reverse the P/E, multi-stage). A stock trades at a P/E of 40. You require a 10% return and assume the multiple de-rates over 8 years to a still-premium terminal P/E of 22. What earnings growth is the price implying? Interpret.

Solution. g = (1 + r) × (PE_now / PE_term)^(1/N) − 1 = 1.10 × (40/22)^(1/8) − 1. Compute (40/22) = 1.818; 1.818^(1/8): ln 1.818 = 0.598, ÷ 8 = 0.0747, e^0.0747 = 1.0776. So g = 1.10 × 1.0776 − 1 = 1.185 − 1 ≈ **18.5%**. Interpretation: even after de-rating to a still-rich 22×, the price needs ~18.5% EPS growth every year for eight years to deliver your 10%. That is a demanding hurdle: sustainable high-teens growth for the better part of a decade is a rare event (Mauboussin's base rates). Verdict: the multiple is doing heavy lifting; unless Q1–Q10 show a wide moat with a long runway, the price is demanding.

P5 (independent: reverse the P/B for a lender). An NBFC trades at P/B 2.5×. Take cost of equity r = 14% and sustainable long-run growth g = 10%. (a) What ROE is the price implying? (b) If the NBFC currently earns 16% ROE and is maturing, is the price demanding or generous, and why is P/E the wrong lens here?

Solution. (a) justified P/B = (ROE − g)/(r − g) → invert: implied ROE = g + P/B × (r − g) = 0.10 + 2.5 × (0.14 − 0.10) = 0.10 + 0.10 = **20%**. (b) The price assumes a ~20% sustainable ROE, but the NBFC earns 16% and is slowing, so the price needs the ROE to rise as the base matures, which is the opposite of the likely path: demanding. P/E is the wrong lens because a lender can inflate near-term EPS by under-provisioning or over-levering, both of which raise the very risk P/E ignores; P/B-vs-ROE, read alongside asset quality (GNPA, credit cost) and capital (CRAR) and funding (ALM), is the correct teardown lens for any bank/NBFC (M5.01–M5.02).

P6 (independent: cash conversion diagnosis). A company reports EBITDA ₹2,000; CFO ₹1,200; PAT ₹1,000; FCF ₹700. (a) Compute CFO/EBITDA and FCF/PAT. (b) What do they say, and what would you check next?

Solution. (a) CFO/EBITDA = 1,200/2,000 = 60%; FCF/PAT = 700/1,000 = 70%. (b) Both are below the ~80% healthy thresholds: a meaningful slice of EBITDA is not turning into operating cash, and profit is not fully converting to free cash. That divergence is a yellow flag: check whether working capital is bleeding (rising DSO/DIO, per M2.07's payables-stretch and receivables checks), whether capex ≫ depreciation (heavy reinvestment, possibly value-creating, possibly not; tie it to Q8's incremental ROIC), and whether any revenue is being booked ahead of cash (contract assets ballooning). A business whose profit persistently outruns its cash is telling you the profit is partly on paper until proven otherwise.

P7 (independent: unit economics of a subscription). A streaming service: ARPU ₹200/month; gross contribution margin 40% per subscriber; CAC ₹1,500 per new subscriber; average subscriber life 30 months. (a) Monthly contribution per sub. (b) CAC payback in months. (c) LTV and LTV/CAC. (d) Verdict.

Solution. (a) Monthly contribution = ₹200 × 40% = ₹80. (b) CAC payback = 1,500/80 = 18.75 ≈ 19 months (well beyond the "great" bar of <12 months). (c) LTV = 80 × 30 = ₹2,400; LTV/CAC = 2,400/1,500 = 1.6× (below the healthy 3× bar). (d) The unit, one subscriber, is marginally profitable but with a long payback and thin LTV/CAC; growth funded this way burns cash and creates little value unless churn falls (extending life) or CAC falls (improving payback). Teardown verdict: weak unit economics; the two things that matter are churn and CAC efficiency, and the variant question is whether either is improving with scale or deteriorating as easy customers run out.

P8 (independent: produce the eight-line output). "Sattva Foods" (synthetic): branded spices and ready-to-cook; revenue ₹5,000, gross margin 45%, EBIT margin 15%, ROCE ~35% for 12 years; negative working capital (distributor advances + fast inventory turns); ~55% share of branded spices in its core states and gaining; founder-family promoter 60%, no pledge, candid calls; trades at 55× earnings. Write the eight lines (including the variant question).

Solution (model eight-line output). 1. Business: a branded-spices and ready-to-cook leader selling everyday kitchen staples to Indian households through a deep distributor network. 2. Revenue engine: branded packaged spices — high gross margin (45%) on a low-cost, high-frequency staple; negative working capital means growth releases cash (distributors and turns fund it). 3. Unit economics: one SKU through one distributor — high contribution margin, near-zero incremental capital, cash-generative from day one. 4. Moat: intangibles/brand (habit + trust in an adulteration-fearful category) plus regional distribution scale — wide, stable-to-widening, confirmed by ~35% ROCE for 12 years. 5. Economics verdict: excellent — ROCE ~3× the 12% hurdle, durable, cash-generative (negative WC). 6. The three things that matter: volume growth (not just price); share gains in core states + expansion to new geographies at similar returns; premiumization/RTC mix. 7. The two things that could break it: a national FMCG or D2C attacker competing the brand premium down; a food-safety/adulteration scandal that breaks the trust. 8. Price verdict + variant question: demanding — 55× on a great business prices continued high-teens growth; variant question: is the growth genuine volume + share gain, or increasingly price-led on flat volume (the FMCG warning sign)? Resolve the volume-vs-value split before acting.

P9 (independent: spot the trap). "Deccan Sugar" trades at a trailing P/E of 8× on record EPS of ₹50 (price ₹400); realizations are at a multi-year high; ROIC is 25% now but was 4% in the last trough; net debt has risen to fund a new mill. Is it cheap? Show the reasoning.

Solution. No. It is a value trap. The 8× is on peak-cycle earnings. Growth here is price (sugar realizations), not volume, and price in a commodity is the cycle. Normalize: if mid-cycle EPS is ~half of peak (say ₹25), the mid-cycle P/E = 400/25 = 16×, ordinary rather than cheap, and at the trough EPS could be near zero, making the multiple meaningless. There is no moat (a price-taker with ROIC swinging 4%→25% has no persistent spread), and net debt is rising to add capacity at the top of the cycle, the textbook capital-cycle error (M4.06/M5.09). The rule crystallizes: for a cyclical, the lowest trailing P/E often marks the top. Correct teardown output: average-to-poor business, no moat; apparent cheapness is peak-earnings illusion; value on mid-cycle EPS and a mid-cycle multiple. Variant question: what is true mid-cycle EPS, and where is the sugar cycle now?

P10 (exam, 20 min: full mini-teardown to eight lines). "LedgerCloud" (synthetic): cloud accounting for SMBs; ARR ₹1,200 growing 30%/yr; net revenue retention 115%; gross margin 80%; Rule of 40 = 30% growth + 10% FCF margin = 40; CAC payback ~15 months; SMBs run their books and payroll on it (deep embedding); ROIC high and rising; trades at EV/Sales 12× (~60× earnings). Produce the eight-line output + variant question.

Solution (model). 1. Business: subscription cloud-accounting software that SMBs run their bookkeeping, invoicing, and payroll on. 2. Revenue engine: recurring SaaS subscriptions — 80% gross margin, NRR 115% (the installed base grows revenue on its own as customers add seats/modules), so a rupee of ARR compounds. 3. Unit economics: one SMB customer — high gross margin, CAC payback ~15 months (acceptable, not elite), and NRR >100% means negative churn on the base → strong lifetime value once past payback. 4. Moat: switching costs (a business's books and payroll are deeply embedded; ripping them out risks the whole operation) plus emerging scale — narrow-to-wide and widening, consistent with high, rising ROIC. 5. Economics verdict: good, improving — Rule of 40 = 40 (healthy), NRR 115%, high ROIC; a quality SaaS profile. 6. The three things that matter: NRR staying above ~110%; CAC payback shortening as the brand scales; retention/logo churn among SMBs (a volatile segment). 7. The two things that could break it: SMB churn spiking in a downturn (small businesses fail); a platform giant (an ERP or a payments company) bundling accounting for free (counter-positioning). 8. Price verdict + variant question: demanding — 12× sales / ~60× earnings prices durable 25%+ growth and NRR holding; variant question: is the 115% NRR durable, or is it flattered by a temporary up-sell cycle that normalizes toward 100% as the base matures? Resolve NRR cohort durability before acting.

P11 (exam, 12 min: reverse EV/EBIT to an implied CAP). A US company trades at EV/EBIT = 14×; tax rate 25%. Its NOPAT is $500m, WACC 9%, incremental ROIC 18%, reinvestment 40% of NOPAT. (a) Convert to EV and EV/NOPAT. (b) Using the M&M skeleton, compute the market-implied competitive-advantage period. (c) One-line verdict.

Solution. (a) EV/NOPAT = EV/EBIT ÷ (1 − tax) = 14/0.75 = 18.67×; EV = 18.67 × 500 = $9,333m. (b) Steady-state value = NOPAT/WACC = 500/0.09 = $5,556m. Reinvestment I = 0.40 × 500 = $200m; value per CAP year = I × (ROIC − WACC)/(WACC × (1+WACC)) = 200 × (0.18 − 0.09)/(0.09 × 1.09) = 18/0.0981 = $183.5m/yr. Implied CAP = (EV − steady state)/value-per-year = (9,333 − 5,556)/183.5 = 3,777/183.5 ≈ 20.6 years. (c) The price is granting ~21 years of above-hurdle returns, a wide-moat assumption. It is justified only if the moat evidence (Q6) supports two decades of durability, which places the company in a small minority, so the multiple is demanding unless the moat is elite (M4.05's base rates).

P12 (exam, 8 min: name the one variant question). A teardown of a fast-growing D2C skincare brand concludes: "good claimed unit economics, ~40% revenue growth, moat unproven (2-year-old brand), trading at 70× earnings." Which of these is the correct single variant question to hand the deep dive, and why are the others wrong? (a) What is the precise WACC to two decimals? (b) Is the 40% growth being bought with rising CAC and falling contribution margin, or is it efficient, retained, repeat-driven growth? (c) What is the CEO's exact salary? (d) What will next quarter's revenue be?

Solution. (b) is correct. For a young, high-multiple D2C brand, the entire investment case turns on whether the growth is value-creating. Efficient, cohort-retained, repeat-driven growth points toward a nascent moat and justifies the runway; growth bought with escalating CAC and discounting is a cash bonfire with no moat, and the 70× is then indefensible. It is the single question whose answer most changes the verdict. (a) is wrong: precision on WACC is deep-dive hygiene, not the variant perception; a teardown doesn't hinge on the second decimal. (c) is wrong: pay is a Q9 governance detail, not the thing that decides the thesis here. (d) is wrong: next quarter's revenue is short-termism; a teardown reasons over the multi-year moat and unit economics, not the next print. The variant question is always the one thing that most moves the verdict and is researchable, here the quality of the growth.

P13 (guided, ROIC vs ROE: which number is the business?). A company reports EBIT ₹1,000; tax rate 25%; D&A ₹200; interest ₹160; invested capital ₹5,000; net debt ₹2,000; book equity ₹3,000. Hurdle 12%. (a) Compute NOPAT, ROIC, EBITDA, and net debt/EBITDA. (b) Compute PAT and ROE. (c) Which number tells you the business's quality, and what explains the gap?

Solution. (a) NOPAT = 1,000 × 0.75 = ₹750 cr; ROIC = 750/5,000 = 15% (+3 over the 12% hurdle). EBITDA = 1,000 + 200 = ₹1,200 cr; net debt/EBITDA = 2,000/1,200 = 1.67×. (b) PBT = 1,000 − 160 = ₹840 cr; PAT = 840 × 0.75 = ₹630 cr; ROE = 630/3,000 = 21%. (c) ROIC (15%) is the business-quality number, the return on all the capital in the machine, before the financing choice. The 21% ROE is six points higher only because ₹2,000 cr of the ₹5,000 cr of capital is borrowed; leverage lifts the return on a thinner equity base while adding fixed-interest fragility. In a teardown you applaud the 15%, not the 21%, and you read net debt/EBITDA (1.67×, moderate) as the price of that lift. Mistaking ROE for quality is exactly how a levered, average business passes for a wonderful one (Worked Example 5).

P14 (guided: reverse EV/EBIT to an implied CAP). A US company trades at EV/EBIT = 12×; tax rate 25%; NOPAT $600m; WACC 8%; incremental ROIC 16%; reinvestment 50% of NOPAT. (a) Convert to EV/NOPAT and EV. (b) Compute the market-implied competitive-advantage period. (c) One-line verdict.

Solution. (a) EV/NOPAT = EV/EBIT ÷ (1 − tax) = 12/0.75 = 16×; EV = 16 × 600 = $9,600m. (b) Steady-state (no-franchise) value = NOPAT/WACC = 600/0.08 = $7,500m. Reinvestment I = 0.50 × 600 = $300m; value per CAP-year = I × (ROIC − WACC)/(WACC × (1+WACC)) = 300 × (0.16 − 0.08)/(0.08 × 1.08) = 24/0.0864 = $277.8m/yr. Implied CAP = (EV − steady state)/value-per-year = (9,600 − 7,500)/277.8 = 2,100/277.8 ≈ 7.6 years. (c) The price grants only ~7.6 years of above-hurdle returns, a modest, narrow-moat assumption; the multiple is not demanding, and if Q6 shows a moat likely to last longer than seven years, the stock may even be conservative. Contrast P11's ~21-year implied CAP: identical machinery, opposite conclusion. The reverse-multiple turns "12× — is that expensive?" into a duration you can test against the moat evidence.

P15 (independent: the good/average/poor economics verdict). "Konark Retail" (synthetic): revenue ₹5,000; gross profit ₹2,000; EBIT ₹800; D&A ₹200; interest ₹80; tax 25%; invested capital ₹2,500; net debt ₹700; CFO ₹830; capex ₹300. Hurdle 12%. Compute the five Q7 reads and give the verdict.

Solution. Gross margin = 2,000/5,000 = 40%; EBIT margin = 800/5,000 = 16%; NOPAT = 800 × 0.75 = ₹600 cr → ROIC = 600/2,500 = 24% (+12 over the hurdle). EBITDA = 800 + 200 = ₹1,000 cr → CFO/EBITDA = 830/1,000 = 83%. PAT = (800 − 80) × 0.75 = ₹540 cr; FCF = 830 − 300 = ₹530 cr → FCF/PAT = 530/540 = 98%. Net debt/EBITDA = 700/1,000 = 0.7×. All five reads are strong: healthy margins, a 24% ROIC at double the hurdle, cash conversion above ~80%, light leverage. Economics verdict: good, bordering excellent, with the one caveat you must state aloud: this holds only if the ROIC has persisted across a cycle, not merely this year. A single strong year is a data point, not a verdict (Mistake #3).

P16 (independent: walk ₹1 to cash, find the leak). A company reports EBIT ₹2,000; tax 25%; D&A ₹400; change in receivables +₹600; change in inventory +₹700; change in payables +₹300; capex ₹800. (a) Compute FCF (NOPAT + D&A − ΔWC − capex). (b) FCF/NOPAT, and where is the biggest leak?

Solution. (a) NOPAT = 2,000 × 0.75 = ₹1,500 cr. Net working-capital change = 600 + 700 − 300 = ₹1,000 cr outflow. FCF = 1,500 + 400 − 1,000 − 800 = ₹100 cr. (b) FCF/NOPAT = 100/1,500 = ~7%, dreadful conversion. The biggest leak is working capital (₹1,000 cr): receivables and inventory swallow cash faster than payables fund it, a larger drain even than the ₹800 cr of capex. The teardown flag: this "profit" is not becoming cash. Before trusting the economics, establish whether the WC build is a one-off (a new plant stocking up) or structural (a business that must finance its customers), and whether DSO/DIO are rising (an M2.07 red flag) or merely scaling with revenue. When profit and cash disagree, believe the cash.

P17 (independent: reverse the P/B, the generous case). An NBFC trades at P/B 1.2×. Take cost of equity r = 14% and sustainable growth g = 8%. (a) What ROE is the price implying? (b) The lender actually earns a stable 18% ROE. Is the price demanding or generous, and what is the catch?

Solution. (a) implied ROE = g + P/B × (r − g) = 0.08 + 1.2 × (0.14 − 0.08) = 0.08 + 0.072 = **15.2%**. (b) The price assumes only a ~15.2% sustainable ROE, but the NBFC earns 18%, stably, so the market is asking for less than the business delivers: generous (the rare cheap lender). The catch, and the reason you never stop at the multiple: a lender's 18% ROE can be manufactured by under-provisioning or over-levering, so the "cheapness" is real only if the asset quality is genuinely clean. Verify GNPA/NNPA, credit cost, the SMA buckets, CRAR, and the ALM/funding profile (M5.01–M5.02) before believing the 18%. P/E would have hidden all of it; the reverse-P/B turns the price into a testable ROE claim. Variant question: is the 18% ROE clean and durable, or flattered by thin provisioning a credit cycle will expose?

P18 (exam, 15 min: full eight-line output from a kit). "Helix Medical" (synthetic, US $): sells diagnostic instruments ($400m revenue at ~5% gross margin, placed near cost to seed the installed base) and recurring test cartridges/consumables ($1,600m at ~70% gross margin) usable only on Helix instruments. Total revenue $2,000m; EBIT $500m; D&A $80m; net cash (interest income +$20m); tax 25%; PAT $390m; invested capital $1,500m; CFO $470m; capex $80m. Market cap $15,600m (P/E 40×); reinvests all profit (no dividend). Hurdle 9%. Produce the reverse-multiple arithmetic and the eight-line output with a variant question.

Solution (model). Blended gross profit = 400 × 5% + 1,600 × 70% = 20 + 1,120 = $1,140m (57% GM); NOPAT = 500 × 0.75 = $375m → ROIC = 375/1,500 = 25% (+16 over the 9% hurdle). EBITDA = 500 + 80 = $580m → CFO/EBITDA = 470/580 = 81%; FCF = 470 − 80 = $390m → FCF/PAT = 390/390 = 100%. Reverse the 40× P/E: r = 9%, de-rate over N = 10 to PE_term = 22×: g = 1.09 × (40/22)^(1/10) − 1 ≈ **15.7%** (no Gordon cross-check: zero payout). The eight lines: 1. Business: a diagnostics-instrument maker whose money is really a consumables annuity. 2. Revenue engine: not the biggest headline — instruments ($400m) are sold near cost to place the box; the $1,600m of ~70%-margin consumables pulled through the installed base carry the economics (razor-and-blades). 3. Unit economics: one placed instrument → years of recurring high-margin cartridge pull-through at near-zero incremental capital → a long, high-margin annuity. 4. Moat: switching costs (validated clinical workflows + regulatory lock to the instrument's own consumables) — narrow-to-wide, stable — confirmed by 25% ROIC. 5. Economics verdict: excellent — 25% ROIC vs a 9% hurdle, CFO/EBITDA 81%, FCF/PAT 100%, net cash. 6. The three things that matter: consumable pull-through per instrument; installed-base growth; consumable gross margin. 7. The two things that could break it: an open-platform/third-party reagent rival breaking the consumable lock; reimbursement pressure on test pricing. 8. Price verdict + variant question: demanding — 40× implies ~15.7% EPS growth for a decade; variant question: is the consumable pull-through per instrument durable and protected, or will open-platform competition and reagent-rental models erode the blade margin? Resolve the razor-blade lock before paying 40×.

Applied mini-project: the timed teardown on a real company

Task. Run ONE full teardown, on the clock (75 minutes hard cap), on a real company you have never studied, and produce the eight-line output plus the variant question, then self-score against the rubric. Pick from this list (chosen to span models and both markets; deliberately not Titan or Costco, which are worked above):

• India: Pidilite (adhesives: brand + distribution), DMart/Avenue Supermarts (value retail: cost moat), Nestlé India (FMCG: brand), HDFC Bank (bank: use the P/B lens), or Divi's Laboratories (API/pharma: process moat).
• US: Chipotle (restaurant unit economics), Moody's (ratings duopoly: network/switching), Copart (salvage-auction network), Deere (cyclical industrial: normalize), or Adobe (software: switching costs).

Exactly what to pull (per M8.01's source stack):

• India: the latest annual report + investor presentation from the company site; the last one or two earnings-call transcripts (screener.in links them); the screener.in ten-year panel (revenue, margins, ROCE, cash flow, debt, promoter holding/pledge); the segment note; one competitor deck; and for share data, the relevant DRHP/CRISIL citations.
• US: the latest 10-K from EDGAR (Item 1 Business, Item 1A Risk Factors, the segment note, MD&A); the last earnings-call transcript; the stockanalysis.com or TIKR ten-year panel; the DEF 14A proxy for the Q9 pay/ownership glance; one competitor's 10-K competition section.

Deliverable. The eight-line output, the reverse-multiple arithmetic shown (Q11), and the one variant question, filed into your permanent company file (Section 5.3). Time yourself and record the finish time; if you ran over 75 minutes, note which question ate the clock and drill that block.

Scoring rubric (self-score honestly against the model discipline; pass ≥17/22, and note the deep-dive gate later demands the timed version at ≥90%).

Case Lab: run case CL-8.1 (The Ninety-Minute Verdict, a synthetic, never-seen company with a full source stack) in the app's Cases tab for this module. The app starts the 60-minute clock, presents a deck + transcript excerpt + ten-year panel you have not seen, and auto-grades your profit-engine call, your ROIC-vs-hurdle verdict, and your reverse-multiple arithmetic (to a tolerance band) against the key before rubric-scoring your eight-line output and variant question.

Reading & resources

• Mauboussin & Rappaport, Expectations Investing (revised 2021). The discipline behind the whole teardown: read the price as an implied expectation, find the variant perception, and reverse the multiple. This is the Phase 8 reading spine; chapters on the expectations infrastructure and the "expectations mismatch" are the theory of Q11–Q12. [Paid] [Intermediate]
• Pat Dorsey, The Five Rules for Successful Stock Investing. The fastest, cleanest catalogue of the five moat sources and the false ones (Q6); the "hitchhiker's guide" chapters are effectively a teardown manual for the moat and economics questions. [Paid] [Beginner]
• Aswath Damodaran, Narrative and Numbers. The habit of turning every story into one number and every number back into a story; the antidote to teardown-by-adjective. His free "narrative and numbers" sessions on YouTube demonstrate the fast version. [Paid book / Free lectures] [Intermediate]
• Mauboussin & Callahan, Measuring the Moat and The Base Rate Book (Morgan Stanley / Credit Suisse, free PDFs). The moat audit (the deep-dive version of Q6) and the mean-reversion evidence your reverse-multiple implied growth must answer to. [Free] [Intermediate/Advanced]
• Business Breakdowns and Acquired (podcasts). The single best free source of model teardowns; the Costco, Visa, and Moody's episodes are hour-long versions of exactly this module's method, and listening to how the hosts find the profit engine trains your Q1 ear. [Free] [Beginner]
• The program's own artifacts (use these live during every teardown): artifacts/playbook-1-2-hour-teardown.md, the one-page sheet this module produces, print it and keep it beside you; artifacts/sector-kpi-cheatsheets.md, the right KPIs and the right multiple per business model (Q7, Q11); artifacts/valuation-workflow.md, whose steps 1, 6, 7 are the fast price sanity-check; artifacts/checklist-red-flags-quality-of-earnings.md, to run the Ctrl-F reflexes; artifacts/checklist-mental-models-and-moat.md, the moat lenses behind the quick-test.
• Damodaran's data library (pages.stern.nyu.edu, updated each January). Industry margins, ROIC, and multiples to benchmark your Q7 economics and Q11 reverse against sector norms. [Free]
• Data stack: screener.in, Tijori, Trendlyne (India, ten-year panels + concall links); stockanalysis.com, TIKR, Koyfin (US); SEC EDGAR full-text search and BSE/NSE filings + MCA21 for primaries. All the input for a teardown is here, free or near-free; the skill is the reading, not the data. [Free/Freemium]
• Going deeper: Michael Mauboussin & Dan Callahan, Return on Invested Capital (Counterpoint Global / Morgan Stanley, free PDF). The rigorous version of Q7: how to build ROIC honestly, the operating-versus-financing split, and precisely why a leverage-flattered ROE is not a quality signal (Worked Example 5). Read it before your first live teardown of a levered business. [Free] [Intermediate]
• Going deeper: Bruce Greenwald & Judd Kahn, Competition Demystified. The barriers-to-entry lens and earnings-power (steady-state) value that formalise the moat quick-test's rule: no persistent spread over the hurdle → no franchise value → value the business at steady state. The clearest deep-dive companion to Q6. [Paid] [Intermediate]

The Modern Analyst's Addendum

Everything above teaches this skill from first principles, by hand. That is how you learn it, and the mastery check still tests it that way. This addendum shows how a working analyst amplifies the same skill today. It adds; it never replaces. (R1/R10)

AI-Augment this skill

{
  "skill": "The twelve-question teardown run inside 60-90 minutes: the profit-engine read, the moat quick-test, the three reverse-multiple inversions, the good/average/poor economics verdict, the eight-line output and the variant question",
  "use": "This is the module where handing the work to a chat assistant is most tempting and most destructive, and the reason is structural rather than moral: the eight-line output is a FORMAT, and a language model is superb at formats. A substituted teardown comes back fluent, correctly shaped, plausibly worded, and completely hollow — and the failure is invisible precisely because the artifact looks right. So fix the line first. LEGITIMATE, all clerical or adversarial: (1) ASSEMBLE the panel — Layer 3's ten years of revenue, margins, ROIC, CFO/EBITDA, FCF/PAT, net debt and share count arrive from the XBRL endpoints as tagged facts rather than forty minutes of retyping, and 8.2.2's Block A gets its numbers before its narrative, which is the anchoring defence anyway. (2) ASSEMBLE the peers — the same concept pulled across a peer set in one request, so Q5's share-and-margin comparison starts from a table instead of six browser tabs. (3) ARITHMETIC you specify: the three inversions of 8.2.5 are one-liners, and having them coded means you run all three every time instead of the one you remember. (4) STEELMAN, and only AFTER your eight lines exist: hand it your own written verdict and ask it to argue the opposite from the same numbers, naming what in the panel would support the other case. That is 8.2.11's anchoring defence weaponised, and it works because you are grading the argument, not receiving it. BANNED, every one of them a graded judgement: Q1's profit engine (the entire Costco and Titan lesson is that the engine is not the headline — noticing that IS the teardown); Q6's moat call (Step 1 is a number in the ROIC series, and a source asserted without that number is exactly the false-moat error the module spends a section rejecting); Q12's demanding/fair/generous verdict; and above all the VARIANT QUESTION, which 8.2's mistake #10 names as the teardown's most valuable output. A variant question you did not write is a deep-dive agenda you do not understand.",
  "tools": ["The data.sec.gov XBRL endpoints and screener.in / stockanalysis.com exports — Layer 3 and the peer set as data rather than as typing", "A notebook with the three inversions coded — reverse-P/E multi-stage, the Gordon cross-check, and the P/B-to-implied-ROE swap for lenders", "Chat assistants — Claude, ChatGPT, used only as a post-hoc steelman against an output you have already written, and to translate document furniture between markets", "Document/PDF AI with retrieval over the deck, the transcript and the filing — with a page citation on every claim, because reconciling the deck against the notes is the one place its recall genuinely saves you time"],
  "prompt": "Here is the eight-line teardown output I wrote for <COMPANY>, and here is the ten-year panel I based it on: <PASTE BOTH>. Argue the OPPOSITE verdict as strongly as the same numbers allow. Specifically: which rows of my own panel most undercut my economics verdict; what a well-informed bear would say my moat line mistakes for a moat; and what would have to be true for my price verdict to be exactly backwards. Do not tell me I am probably right, do not soften, and do not introduce any figure that is not in the panel I gave you. I will decide what survives.",
  "verify": "Two verifications, and the second is the one that catches a substituted teardown. First the arithmetic: prove any coded inversion against this module's own worked figures before it touches a company — 50x de-rating to 25x over ten years at r = 9% must give 16.8%; the Gordon cross-check at r = 11%, 40% payout and a 25x P/E must give 9.4%; the P/B inversion at 3.0x, r = 13% and g = 8% must give an implied ROE of 23%. Second, and harder: audit the OUTPUT for the tells of a teardown nobody ran. A real one has a parking lot — a thread it abandoned mid-sentence when the block timer rang. Its moat line carries a number and a year count ('confirmed by ~22-25% ROCE over a decade'), not an adjective. Its price verdict names the required return it used, because 8.2.5's inversions are meaningless without one. And its variant question is specific enough that you could name the three exhibits that would settle it — 'is Tanishq's growth being bought with making-charge discounting' points at the studded-mix trend, per-store throughput and the gross-margin bridge, while 'is the moat durable' points at nothing. An output that fails those four tests is a document, not a teardown, and the distinction is invisible to anyone reading it — including, six months later, you.",
  "diy": "The gate is unaided and it is timed: sixty minutes, a company you have never seen, no tool in the room. You run the twelve questions in order from memory, find the profit engine yourself, confirm or reject the moat in the ROIC series, do the reverse-multiple arithmetic on paper with your own required return, and write the eight lines and the variant question. Nothing in this callout is admissible there, and M8.05 certifies exactly that room."
}

Modern Data Analysis

By hand first. You ran both worked trails end to end (§8.2.7–8.2.8), found the profit engine hiding inside Titan's biggest line and outside Costco's, computed a 22.5% ROE down to an honest 12.0% ROIC in Ridgeline (WE5), and inverted five different multiples with your own required return. Keep all of it, and the inversions especially — a reverse-multiple you can only do in a spreadsheet is a reverse-multiple you cannot do in a meeting.

Today's workflow. The teardown is a time-boxed process, which makes it the one module in this program where the honest question is not "what can the tools do" but "how many of the sixty minutes are they even allowed to touch." Take §8.2.2's budget as given and split each block into two kinds of minute. Clerical minutes are locating a number, retyping it, computing a ratio you already know the formula for, running a fixed search list. Judgement minutes are deciding what the number means, which of five moat sources it is, whether the implied growth is plausible, what the variant question is. The classification is a stated model, not a stopwatch reading, and the Quant lens below shows the conclusion barely moves when you argue with the split.

What the machine-readable half buys is then specific and bounded. Block A's panel arrives as a fetch, which is convenient and also does the anchoring work §8.2.11 demands, because the numbers land before the deck does. Block D's tripwire scan becomes a query across a decade of filings rather than a scan of one. Block E's three inversions become a function you call three times instead of the one you remember under pressure. Blocks C and F barely move at all, and Block F, writing the eight lines, is essentially incompressible, which §8.2.6 already told you when it said a teardown that cannot fit a line is not finished, it is still reading.

The India–US asymmetry lands squarely on this budget. For a US name Block A's assembly is genuinely a fetch; for an Indian name the same panel comes off screener.in's export button or out of a PDF, and the tripwire sweep is a search inside one downloaded annual report rather than a full-text query across a decade of them. Neither of those is slow enough to matter at teardown scale, since the aggregator export is thirty seconds. But the deep-dive version of the same gap is real, and M8.03 is where it bites.

Tools & sources (IN + US). US: data.sec.gov/api/xbrl/companyfacts/CIK##########.json for the ten-year panel in one request and .../frames/us-gaap/<Tag>/USD/<period>.json for the same concept across a peer set, which is Q5's benchmark without assembling peers by hand; data.sec.gov/submissions/CIK##########.json for the filing history behind Q9's auditor-change check; EDGAR full-text search for the tripwire terms across a decade; stockanalysis.com, TIKR and Koyfin free tiers for the panel when you want it pre-built; the DEF 14A for Q9's pay metrics. The SEC requires a declared User-Agent identifying you and publishes a fair-access rate policy; read the current policy rather than a number quoted anywhere. India: screener.in's ten-year data sheet, its Excel export and its concalls tab (the transcripts are the highest-yield unstructured source in the market); Tijori and Trendlyne; BSE and NSE announcement feeds and filing archives, scrapeable but PDF-payload; SEBI's LODR regime for the two disclosures Q9 lives on, namely Reg 31's shareholding pattern with the pledge percentage and Reg 23's half-yearly related-party disclosures; MCA21 for the unlisted subsidiaries when a consolidation stops making sense. Run any of it in your notebook.

# R3 — recomputed in-session. Where the hour goes, and what the filter is worth.
BLOCKS = {"A orient": 10, "B economics": 14, "C moat": 11,      # 8.2.2 / M8.05 4.5 budget
          "D stewardship+tripwire": 10, "E price": 8, "F write": 7}          # = 60 min
CLERICAL = {"A orient": .60, "B economics": .50, "C moat": .35,  # STATED model, not a stopwatch
            "D stewardship+tripwire": .60, "E price": .40, "F write": .05}
T = sum(BLOCKS.values())
c = sum(BLOCKS[b] * CLERICAL[b] for b in BLOCKS) / T
print(f"clerical {c*T:.1f} of {T} min = {c:.1%}; judgement {T*(1-c):.1f} min")   # 26.4 / 60 = 44.0%
for k in (2, 4, 10, float("inf")):
    t = T * ((1 - c) + (c / k if k != float("inf") else 0))
    print(f"  compress clerical {k}x -> {t:.1f} min, speed-up {T/t:.2f}x")
    #  2x -> 46.8 (1.28x) | 4x -> 40.2 (1.49x) | 10x -> 36.2 (1.66x) | free -> 33.6 (1.79x)
for cc in (.30, .40, .50, .60):                       # the ceiling barely cares where you draw the line
    print(f"  if clerical were {cc:.0%}: floor {T*(1-cc):.0f} min, ceiling {1/(1-cc):.2f}x")

# The teardown as a screening test. ASSUMED: 1 company in 20 clears your bar; the 60-minute
# filter is 80% sensitive and 80% specific.
prev, sens, spec = 0.05, 0.80, 0.80
tp, fp = prev * sens, (1 - prev) * (1 - spec)
fn, tn = prev * (1 - sens), (1 - prev) * spec
print(f"PPV {tp/(tp+fp):.1%}  NPV {tn/(tn+fn):.2%}  pass rate {tp+fp:.1%}")   # 17.4% / 98.70% / 23.0%
print(f"specificity needed for PPV=50%: {1-(prev*sens)/(1-prev):.1%}")        # 95.8%
tp2, fp2 = tp * 0.90, fp * 0.10                       # deep dive: 90% sensitive, 90% specific
print(f"two-stage PPV {tp2/(tp2+fp2):.1%}, cumulative sensitivity {tp2/prev:.1%}")  # 65.5% / 72.0%
# throughput per 100 candidates, at 1 teardown-hour and 3 analyst-days per deep dive, 8h days
no_screen = (100 * prev) / (100 * 3)
screened  = (100 * tp) / (100 / 8 + 100 * (tp + fp) * 3)
print(f"finds/day: unscreened {no_screen:.4f} vs screened {screened:.4f} -> {screened/no_screen:.2f}x")

Verify. Prove the inversion engine against this module's own worked figures before it goes near a company: the multi-stage read must give 16.8% for a 50× de-rating to 25× over ten years at r = 9%, 21.3% for Titan's 85× to 35× at r = 11%, 13.5% for Visa's 30× to 20×, 14.2% for Griha's 24× to 18× at r = 11%, and 12.6% for Ridgeline's 18× to 13×; the Gordon cross-check must give 9.4% at r = 11%, 40% payout and 25×, 10.7% for Titan, 8.4% for Costco, 9.5% for Griha and 7.3% for Ridgeline; the P/B inversion must give an implied ROE of 23% at 3.0×, r = 13%, g = 8%. The kits must reproduce Griha's ROIC of 20.0%, CFO/EBITDA 85%, FCF/PAT 112% and net debt/EBITDA 0.5×, and Ridgeline's ROIC of 12.0% against a ROE of 22.5% (the whole point of that example) with CFO/EBITDA 75%, FCF/PAT 77.8% and leverage 3.0×. Then verify the thing no engine checks: that the required return you fed it is yours. Every one of those figures moves point-for-point with r, which is why M8.01's WE6 spends four steps on a sell-side note whose "3% priced in" and your "5% priced in" differed only because the note silently used 7% where you used 9%. A coded inversion makes it trivially easy to inherit someone else's hurdle and never notice.

Quantitative lens

The teardown is a time-boxed filter, so its two honest quantitative questions are where do the sixty minutes go and what is a filter of this quality actually worth. Both are computable, both come from the block above on the inputs stated there, and both are properties of that arithmetic rather than measurements of anyone's practice.

The compressible part of the hour is bounded, and so is the speed-up. On the stated split, 26.4 of the 60 minutes are clerical, or 44%, and 33.6 are judgement. Now compress the clerical part and watch what happens: halve it and the teardown takes 46.8 minutes (a 1.28× speed-up); quarter it and 40.2; cut it tenfold and 36.2. Make it free, a perfect fetch, zero seconds, no typing at all, and the teardown still takes 33.6 minutes, a ceiling of 1.79×. This is Amdahl's law wearing a stopwatch, and the conclusion survives arguing with the split: at a 30% clerical share the ceiling is 1.43×, at 50% it is 2.00×, at a wildly generous 60% it is 2.50×. No plausible accounting produces a ten-minute teardown, because the thing that takes the time is deciding what the numbers mean. The practical instruction is the opposite of the one people reach for: do not try to make the whole hour faster. Buy back the clerical minutes in Blocks A and D, where they concentrate, and spend them on Block C's moat evidence and Block E's price work, the two blocks §8.2.2's budget already starves and M8.05's failure modes 2 and 3 name as the first casualties under pressure.

And the filter is a superb rejector and a poor selector, which is exactly why the funnel is shaped the way it is. Suppose one company in twenty in your screening universe genuinely clears your bar, and a sixty-minute teardown is 80% sensitive and 80% specific, which is generous for an hour's work. Then of every hundred companies you tear down, 23 pass and only 4 of those 23 deserve to: a positive predictive value of 17.4%. Most of what survives your filter is a false positive, and no amount of care inside the hour fixes it, because the arithmetic is driven by the base rate. To get the pass pile to be mostly real at 80% sensitivity you would need 95.8% specificity, a one-in-twenty-four false-positive rate on companies you have looked at for sixty minutes, which is not a thing an hour can deliver.

Now read the other diagonal, which is the one the module is actually built on. The negative predictive value is 98.70%: of everything the teardown rejects, fewer than one in seventy was a company you should have studied. The teardown is not a picking tool. It is a rejecting tool, and §8.2.1 says so in words: "the filter that makes the deep dive affordable… avoiding the twenty businesses not worth studying is worth more than perfecting the study of the one that is." Here is the throughput that justifies it. Deep-dive all hundred candidates at three analyst-days each and you spend 300 days to find 5, at 0.0167 finds per day. Screen first (a hundred teardown-hours, then three days on each of the 23 survivors) and you spend 81.5 days to find 4, at 0.0491 finds per day. That is 2.94× the throughput for the cost of one missed company, and the missed company is the honest price: stage-one misses are unrecoverable, so the two-stage funnel's cumulative sensitivity is only 72.0% even though its final positive predictive value climbs to 65.5%.

An analyst who cannot bear to reject anything has not avoided that cost. They have paid it in the twenty-nine companies they never got to.

Honest limits, and the first is the important one. The 5% base rate, the 80/80 filter and the clerical split are all assumed: they are the inputs that make the arithmetic run, not measurements of your universe or your hour. Change the prevalence and the story changes shape: at a 20% base rate the same filter's positive predictive value rises to 50%, and at 1% it falls to 3.9%, which means the same teardown skill is worth wildly different things depending on where you fish. That is a real finding and it argues for screening discipline upstream of the teardown, not for a better teardown. Sensitivity and specificity are also not independent of the analyst: your own bar moves them, and the module's mistake #4, calling a low multiple cheap, is precisely a specificity failure you can drill out. The clerical/judgement split is a classification of §8.2.2's budget under a stated criterion, and a learner whose plumbing is not yet automatic has a much higher clerical share than 44% and therefore a much higher ceiling; the ceiling shrinking as you improve is not a defect of the model, it is what automaticity is. Finally, none of this says a teardown that passes a company is right about it. It says what fraction of a pass pile, under stated assumptions, would survive contact with three more days, which is a sharp sentence about a process, and not a statement about any company.

Do it in code: fetch the panel before you open the deck, code the three inversions and run all three every time, log the required return you used beside every implied-growth figure, and keep a tally of your own pass rate. If you are passing far more than a fifth of what you tear down, your specificity is the thing to fix, not your speed. See G1/QM1.01, G2/DA1.02 and G3/CN5.01.

Where this goes next: galaxy cross-links

• G1 · QM1.01 Probability, Distributions, Sampling & Estimation. The screening arithmetic above: base rates, positive and negative predictive value, and why a filter's worth depends more on the prevalence in your universe than on the care you took inside the hour.
• G3 · CN5.01 Business-Model Teardown & Unit Economics: The Advisory Lens. The same twelve questions asked for a client rather than a portfolio, where Q4's smallest repeatable unit and Q3's revenue-to-cash walk become the deliverable rather than an input to a price verdict.
• G2 · DA1.02 pandas I: Data Wrangling. Turning a companyfacts payload and a peer frames query into the one-screen ten-year panel Block A reads, including the concept-name and re-segmentation breaks that make a naive ten-year join quietly wrong.
• G2 · CS2.05 APIs, HTTP, JSON & Web Scraping. How the endpoints behind Block A actually work, and the difference between a documented API and an exchange portal you are scraping, which is the whole US–India asymmetry in one lesson.
• G5 · AI0.07 The AI-Augmented Analyst — Applying AI Across Your Work. The general form of the line this module has to draw hardest: which parts of a repeatable analytical process amplify, which substitute, and why a well-formatted output is zero evidence that the process ran.

Teach it back & journal

Feynman prompt. Take one company you use every week (the coffee chain you buy from, the phone in your pocket, the bank that holds your salary) and write a one-page teardown a smart 15-year-old could follow: what it sells and who pays; where the money actually comes from (is it the obvious line, or something hidden, the way a warehouse club really earns from the membership card, not the groceries?); whether it's a good business or an ordinary one, and how you'd know from the outside; and finally, the one thing you'd want to find out before deciding it was worth owning. If your page needs the words "moat," "ROIC," or "multiple," you haven't made it simple enough — but the ideas behind those words must all be there.

Journal prompt. Three entries. (1) Run a full timed teardown on a company you've never studied (your mini-project company) and record your finish time and the eight-line output; then note which of the twelve questions was hardest and slowest. That is your deliberate-practice target for the next month of company-of-the-week reps. (2) Look back at a company you formed a quick opinion on months ago (from your earlier weekly reads): re-run the moat quick-test and the reverse-multiple on it now, and write whether your faster, more disciplined self reaches a different verdict, and which specific question changed it. (3) Write your single most common teardown failure mode from this module's practice (falling for a false moat? anchoring on the raw multiple? forgetting the variant question?) and the one-line rule you'll pre-commit to that stops it.

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