Dynamics + Capstone

Phase 4 · Week 41 · ~20 hours

Where it sits. The last module of the strategy phase, and the one that puts the previous five in motion. The first three taught you to dissect a business model and its industry; the next two gave you four lenses for judging whether a moat exists today and how long the ROIC−WACC spread might last (the CAP). But every verdict you have rendered so far is a photograph, and moats are movies. Two great forces write the next reel: the capital cycle (Marathon/Chancellor: high returns summon capacity; supply, not demand, is where cycles are made) and disruption (Christensen: some entrants win precisely because the incumbent's economics make retreat rational). Alongside them sits the erosion dashboard: five numeric signals that a moat is dying, visible in filings years before the story admits it. Then the phase ends the only way it can, with the Phase 4 capstone: two complete moat dossiers, one Indian and one US company, four lenses plus dynamics, proven in a 10-year numbers table, CAP estimated, direction called, graded against a rubric. This is the M5 milestone in the master map, and the moment competency C4 ("assess competitive advantage — and prove it in the numbers") is certified. Phase 5 then takes your finished moat machinery sector by sector, starting with banks.

Suggested hour budget. Core content: 6h · worked examples (including the model dossier): 2h · practice set: 2h · capstone dossiers: 8h · flashcards + mastery check: 2h.

App link. The Industry & Business Model branch completes here: the cycle-stage classifier (industry data in, cycle position out), the erosion-dashboard drill (six-year numbers in, five signals scored, verdict out), the disruption triage deck (entrant described, sustaining/low-end/new-market/not-disruption out; "not disruption" is the most common right answer), and the capstone submission flow (both dossiers self-graded against the rubric, then compared to the model dossier). Your dossiers also seed the moat section of the Mental-Model & Moat Checklist, the permanent artifact this phase leaves behind.

Learning objectives

By the end you can:

1. Narrate and detect the capital cycle: state the six-stage anatomy (high returns → capital floods in → capacity arrives → returns collapse → capital flees → returns recover), explain why the cycle is made on the supply side, and identify an industry's current stage from public data.
2. Read cycle position in the numbers: compute and interpret capex/depreciation ratios, industry capacity growth vs demand growth, capacity utilization, order books, and issuance clusters, and name the exact India and US sources for each gauge.
3. Apply the capital-cycle investing rule (be afraid of sectors where capital is flooding in, hunt in sectors starved of capital), including its demand-side caveat and its interaction with moats (the cycle is weather for no-barrier industries, climate for moated ones).
4. Classify innovation correctly: distinguish sustaining innovation from low-end and new-market disruption, explain the mechanism that makes incumbent retreat rational (asymmetric motivation, the margin-mix ratchet, RPV), and run the four-question test that filters real disruption from the ~90% of claimed "disruption" that isn't.
5. Audit any moat for disruption vulnerability: map which moat sources survive architecture shifts and which are architecture-specific, and write a named-threat vulnerability assessment rather than a vague "tech risk" line.
6. Run the erosion dashboard: compute and interpret the five numeric erosion signals: gross-margin slippage, rising churn/discounting, share loss to a structurally different rival, ROIC glide, and rising reinvestment with falling ROIIC, each with its false-positive check.
7. Render a direction verdict (widening / stable / eroding / broken) that names the mechanism, cites the numbers, and pre-commits falsifiable tripwires ("what would change my mind").
8. Produce an institutional-grade moat dossier: the full four-lens analysis + dynamics extension + 10-year numbers-proof + CAP estimate + direction verdict, at the standard of the model answer below, twice, on companies you have never formally studied.

Prerequisites & connections

Builds on. From M2.03 and M2.05 you bring ROIC and ROCE built by hand, the ROIC−WACC spread, and the mean-reversion base rates (Mauboussin's transition matrices). Here you get the machinery behind that mean reversion, which is less a statistical mist than capital arriving in trucks. From M3.05 comes the reverse DCF, where you'll read market-implied CAPs and ask whether the dashboard supports them. M3.07 gave you ROIIC (ΔNOPAT ÷ ΔInvested Capital, 3-to-5-year windows) and the five uses of capital. Erosion signal 5 is ROIIC weaponized, and the capital cycle is M3.07's capital-allocation report card written at industry scale. M3.10 taught mid-cycle normalization for cyclicals, and the capital cycle tells you where in the cycle you're normalizing from. M4.03 gave you five forces and profit pools. The capital cycle is the threat-of-entry force given a clock, and disruption is the substitute force given a mechanism. M4.04–M4.05 supplied the four moat lenses, share-stability test, and CAP. Everything here assumes them fluently, and the capstone grades them.

Feeds forward. Phase 5 leans on all of this. Its commodities module (M5.09) is applied capital-cycle analysis, SaaS (M5.04) applies the erosion dashboard to NRR, and autos (M5.08) runs the EV transition as a live disruption audit. Phase 6 files both forces in the mental-model latticework. Munger's line there is "show me the incentive", and the capital cycle is incentives at industry scale. Phase 8's teardown question 6 asks for moat and direction in ten minutes, and the deep dive's Day 2 industry map ends with a cycle-position call. Phase 9's sell rules ("thesis broken") fire on the tripwires you learn to write here.

The short version. A moat verdict without a direction is half a verdict. Run the capital cycle and you learn whether money is about to storm the industry's wall. Disruption theory asks whether a different business model is about to bypass it. And in gross margins, churn, share, ROIC and ROIIC, the erosion dashboard shows whether the storming has already begun. Then the capstone makes you prove you can run the whole machine twice.

Core teaching content

4.1 Moats are movies, not photographs

Of firms in the top quintile of ROIC, only about half are still in the top two quintiles a decade later. That's Mauboussin's persistence data from M2.03, and it's the uncomfortable fact behind everything that follows. Think about what it means for the moat work you've just finished. A perfectly executed static moat analysis, with the right source confirmed in ten years of ROIC and a stable share table, describes the past decade. The price you'd pay today is a claim about the next decade. And the base rates say that for roughly half of apparently great businesses, the next decade breaks the pattern.

So the question that separates a competent moat analyst from a dangerous one is not "is there a moat?" but "which way is it moving, and what specifically would move it?" Two forces do almost all the moving:

1. The capital cycle attacks the moat frontally. High returns attract capital. Capital builds capacity, and capacity destroys returns. It's the mechanism behind mean reversion in most industries most of the time. In capital-cycle language, a moat is whatever makes this mechanism fail, the wall that keeps the flood out. Watching the cycle tells you when the wall is being tested, and with how much water.
2. Disruption attacks the moat from below or beside. An entrant arrives whose business model makes the incumbent's rational response a retreat. It's rarer than the business press suggests, and far more dangerous, because instead of testing the wall it makes the wall irrelevant.

Both forces leave fingerprints in the financial statements long before they finish the job. So there's a third tool, a dashboard of five numeric signals, each with a computation and a false-positive check. Together they turn "I have a bad feeling about this moat" into evidence.

One orientation note before we start. These two forces relate to your moat verdicts in opposite ways. Think of the capital cycle as what happens to industries without barriers. A confirmed moat amounts to a claim of capital-cycle immunity, and the cycle analysis becomes a stress test of that claim. Disruption preys on strong incumbents instead. The better the incumbent's margins and the more captive its best customers, the more rational its fatal retreat.

So your widest-moat holdings are your most disruption-exposed ones. Hold both ideas at once, because the capstone requires it.

4.3 Reading cycle position in the numbers

You need gauges, not vibes. Here are six, in rough order of usefulness, each with its computation, its threshold pattern, and where to find the data for India and the US.

Gauge 1: Capex/Depreciation (the workhorse). Capex/D&A = capital expenditure ÷ depreciation & amortization. Compute it per company from the cash-flow statement, then aggregate across the industry's top players. Capex is purchases of PP&E, and you take D&A from the CFO section or the PP&E note. Read it with three rules of thumb. Around 1.0 to 1.3× is replacement mode, and a touch above 1 is normal because inflation makes replacement cost exceed historical-cost depreciation. Sustained above 2×, it's the capacity-building phase of stage 2 to 3, and supply is coming. Below 0.8× is starvation or harvest, stage 5, where tomorrow's tightness is being made today. Always compute it for the industry. Sum the top 4 to 6 players' capex and D&A instead of averaging their ratios, and always look at 3+ years, because single years are lumpy. Three caveats apply. Acquisitions add capacity without touching the capex line, so use capex plus acquisitions for the expansion read. Asset-light industries need a different numerator: fleet orders for airlines, content spend for streaming, R&D for software. And CWIP (capital work in progress, on the balance sheet) tells you how much of the wave is still inbound.

Gauge 2: Capacity growth vs demand growth. This is the cycle in one subtraction: industry capacity CAGR − demand CAGR over a 3 to 5 year window, counting the announced pipeline. If it's positive by more than ~2 points a year for several years, utilization must fall and pricing will break. For India, start with company investor presentations. Every cement/steel/chemicals deck has a capacity table and an expansion slide, but it's self-reported, so cross-check it against competitors. CRISIL/ICRA sector notes are often summarized free in press releases. Industry bodies help too: the Cement Manufacturers' Association, the Joint Plant Committee for steel and SIAM for autos. And the RBI OBICUS survey gives a free quarterly capacity-utilization reading for Indian manufacturing. For the US and global industries, use 10-K Item 2 (Properties) and investor decks. EIA covers energy. Baker Hughes publishes its rig count weekly and free, and it works as a real-time supply gauge for oil. Then there's the World Steel Association for steel, WSTS/TrendForce summaries for semiconductors, and Clarksons order-book data for shipping, whose headline numbers circulate free in the trade press.

Gauge 3: Capacity utilization. This is the stock variable the flow gauges drive. When it falls through ~75% in a fixed-cost industry, pricing power is gone. When it rises through ~85%, pricing power is returning and the next capex wave is being drafted. Sources are company presentations, OBICUS for the India aggregate, and Fed G.17 for US industrial capacity utilization, free on FRED.

Gauge 4: Order books and pipelines (the pre-announced future). In industries with long build times, the future supply schedule is literally published. Take the shipping order book as a % of the existing fleet, where >30 to 40% means a wall of steel is coming and <10% means starvation. Hotel rooms under construction (as a % of stock) and announced fab capacity work the same way, and so does cement CWIP. It's the purest expression of "supply is knowable."

Gauge 5: Issuance and entry clusters (the financing tell). Count the sector's IPOs, QIPs, and debt raises. A cluster of eager issuance at high valuations marks stage 2, with capital flooding in. Issuance to repair balance sheets at depressed prices marks stage 5. That's capitulation, and often the buy signal. In India, look at Prime Database summaries in the financial press and the NSE/BSE new-listings pages. In the US, count S-1s on EDGAR by sector and watch the sector's share of high-yield issuance. Then add the softer tells Chancellor lists. Watch for the sector's share of total market cap ballooning and analyst-coverage headcount growing. A "supercycle" narrative appears, with new-era vocabulary ("this time"). You'll get the full bubble anatomy later on.

Gauge 6: Asset prices vs replacement cost (Tobin's q, per unit). When enterprise value per unit of capacity trades far above replacement cost, the market is paying entrants to build. That's stage 2. Far below, it's paying acquirers to consolidate instead of build, which is stage 5: capacity changes hands at 50 cents on the rupee, and no rational actor adds new supply. Cement analysts price EV per tonne against ~US$100 to 120/tonne greenfield replacement cost (illustrative, so check the current build costs). Shipping analysts compare second-hand ship prices to newbuild parity.

Then assemble the six into a one-line cycle-position call: "Industry X is in stage [n]: capex/D&A [x], capacity pipeline [y]% vs demand [z]%, utilization [u]% and [rising/falling], issuance [eager/absent/forced], assets at [premium/discount] to replacement." Your capstone dossier needs that line, with sources.

4.4 Worked example: India: the cement capacity wave, FY07 → FY25

All figures here are illustrative, reconstructed at ~FY07 to FY25 order of magnitude from CMA data, company presentations, and rating-agency reports. Pull the actuals from UltraTech/Ambuja/Shree investor decks and CRISIL/ICRA notes when you rerun this.

Stage 1 (≈FY05–FY08): high returns. India's construction boom ran demand growth at ~9 to 10% a year into a capacity base of roughly 160 million tonnes (MT). Utilization pushed past 90%, cement prices rose double digits, and leading players printed ROCEs of 25 to 35%. Cement stocks re-rated to peak multiples on peak earnings. That's the cyclical trap from the normalization work, out in the wild.

Stage 2 (≈FY06–FY09): capital floods in. Every incumbent announced expansions. New entrants piled in, from real-estate groups to diversified industrialists, and bank credit to infrastructure was abundant. Announced pipeline came to roughly 100+ MT, around 60% of the installed base, and almost all of it was approved inside a three-year window of peak pricing. Industry capex/D&A ran well above 2.5× (illustrative).

Stage 3 (≈FY09–FY13): capacity arrives together. Capacity climbed from ~160 MT to ~330 to 350 MT by FY13 to FY15, a supply CAGR of ~11% a year. Demand, meanwhile, decelerated to ~5 to 7% a year as the investment cycle cooled and the GFC bit. The arithmetic was merciless. Utilization fell from >90% toward ~70%.

Stage 4 (≈FY12–FY16): returns collapse. With a quarter of the industry's kilns effectively idle, producers started undercutting each other region by region, and realizations stagnated for years while fuel and freight costs rose. Sector ROCE compressed to single digits or the low teens, below the ~12% INR hurdle. Leveraged expanders, who had built with debt at peak land and equipment prices, slid toward distress.

Stage 5 (≈FY14–FY20): capital flees. Industry capex/D&A dropped toward 1× and below (illustrative), expansion announcements vanished, and the weak sold. You should recognize the consolidation roll-call. Jaiprakash Associates sold ~21 MT to UltraTech (~₹16,000 crore, closed 2017). Binani Cement went through the IBC to UltraTech (~₹7,900 crore, 2018), and Century Textiles' cement business followed in 2019. Other stressed assets went to Nirma/Nuvoco and others. Notice the tell from Gauge 6. These deals priced capacity at meaningful discounts to greenfield replacement cost, so rational buyers bought old capacity instead of building new. That's the signature of stage 5.

Stage 6 (≈FY18–FY22): returns recover. Demand grew into the glut and utilization climbed back through the mid-70s. By now the top 5 controlled a decisively larger share, and these consolidated majors rediscovered pricing discipline. Leaders' ROCEs recovered into the mid-teens and beyond. If you'd bought the starved industry at replacement-cost discounts in stage 5, you were paid handsomely.

And then the reel restarts. From FY22, with returns healthy again, the industry launched the next wave. First the Adani group entered by buying Ambuja/ACC (~US$10.5bn, 2022) and announced a march toward ~140 MT. UltraTech countered toward ~200 MT, and the industry now points toward ~700 MT by ~FY28 (all illustrative). Can housing and infrastructure demand grow fast enough to absorb it? That's the question a cement analyst in 2026 has to answer with Gauges 1 to 6, and any straight answer starts with "supply is coming; I can already count it."

The meta-lesson. Nothing in this story required forecasting demand brilliantly. Every turn was visible years ahead in announced capacity, capex/D&A, and issuance behavior. That's the capital cycle's gift. The future of supply is published.

4.5 Worked example: global: the dry-bulk shipping supercycle, 2003 → 2021

Shipping is the capital cycle with the lid off. It's a pure commodity service, because a tonne-mile is a tonne-mile. Barriers are near zero, since in a boom banks will mortgage a ship to anyone. Supply lags 2 to 3 years from order to delivery, and with a 25-year asset life it exits at glacial speed. Chancellor's collection treats it as the canonical specimen. Dates and levels below are real headline history, but verify the details against Clarksons/Baltic Exchange data.

Stages 1–2 (2003–2008). China's industrialization exploded seaborne iron-ore and coal demand. Bookmark the Baltic Dry Index (BDI), a daily, free, public freight-rate composite. It rose from under ~1,000 in 2002 to a peak of ~11,800 in May 2008. Capesize vessels that cost ~$25k/day to run earned ~$200k+/day at the top. Owners ordered ships as fast as yards could take deposits, and by 2008 the dry-bulk order book reached roughly 70 to 80% of the entire existing fleet. That was the future supply schedule, published in advance, screaming at anyone who read it. Shipping IPOs clustered while banks lent against peak vessel values.

Stages 3–4 (2008–2016). Demand cracked with the GFC, but the ordered ships kept arriving for years. That lag is physics, because you can't cancel a half-built hull economically. Into flat demand, the fleet grew at high-single to double digits a year. In seven months the BDI collapsed ~94%, from ~11,800 in May 2008 to ~660 that December. It rallied briefly with China's stimulus, then ground down for seven more years as deliveries outran scrapping, and bottomed at an all-time low of ~290 in February 2016. Rates spent long stretches below operating cost, and owners with peak-priced, debt-financed fleets failed in waves.

Stages 5–6 (2016–2021). Ordering stopped almost entirely and scrapping accelerated. By 2020 to 21 the order book had shrunk to ~6 to 8% of the fleet, a multi-decade low. Then demand merely normalized with post-COVID restocking, and against a starved supply pipeline the BDI printed >5,600 in October 2021, with Capesize rates back near ~$80k/day. Investors who bought quality owners at discounts to steel-scrap value in 2016 to 2020, when consensus called the industry uninvestable, collected the recovery.

The transferable arithmetic. You'll drill this in the practice set. If the order book is X% of the fleet, deliverable over ~3 years, gross supply growth is ≈ (1+X)^(1/3) − 1 per year. Subtract scrapping, which runs ~1 to 3% a year and higher when rates are below opex and ships hit 20+ years. Then compare with demand growth. In 2008 that computation said ~+17% a year of gross supply against low-single-digit demand, a published catastrophe. In 2020 it said ~+2% against recovering demand, a published boom. Nobody needed to predict China or COVID. They needed to read the order book.

Two compressed variations to file alongside (both label-checked, approximate). First, US shale oil ran the whole cycle twice in a decade. E&Ps outspent operating cash flow through 2010 to 2015, funded by a high-yield bond flood, and oil rigs peaked at ~1,600 in late 2014. Then came the crash. Rigs fell to ~170 by mid-2020 and hundreds of E&Ps went bankrupt over 2015 to 2020. By October 2020 the energy sector was ~2 to 3% of the S&P 500, down from ~13 to 16% in 2008, which is the starvation marker. After that the industry converted to a capital-returns regime, reinvesting ~40 to 50% of cash flow and posting record FCF and buybacks, and it was the index's best sector in 2021 to 22.

Second, DRAM memory shows stage 5's happiest ending. In the mid-1990s there were ~15 to 20 manufacturers. Brutal cycles bankrupted or consolidated the field (Qimonda in 2009, Elpida in 2012, bought by Micron) down to three players controlling ~95% by ~2013. After that, trough returns sat structurally higher. It was still cyclical, but now it was an oligopoly's cycle. Capital fleeing can permanently improve an industry's economics when it exits through consolidation. You're watching Greenwald's barriers being built out of wreckage.

4.6 Investing with the capital cycle

Here are four rules, each with its caveat. They're the "so what" your dossier's cycle section has to land on.

Rule 1: Fear the flood. Capex/D&A runs hot, issuance clusters, the pipeline exceeds demand growth, and a supercycle narrative is in the air. When you see that, expected returns are being destroyed in plain sight, whatever the current P&L says. Peak-cycle P/E looks cheapest right at the top, because earnings are at their peak and the multiple is compressed. That's the cyclical trap again. So in stage 2 to 3 industries, either underwrite the coming supply explicitly or stand aside.

Rule 2: Hunt in the famine. Your highest-probability capital-cycle purchase is an industry at stage 5. Capex/D&A is below 1, consolidation is underway, assets are changing hands below replacement cost, and coverage has been abandoned. Buy it through the best balance sheet and the lowest-cost operator, because they survive to see stage 6. The trough kills the weak first, and that's how the trough ends. Its caveat is the demand test. Starved supply only creates pricing power if demand persists. Newsprint and film photography were starved of capital all the way to zero, so the recovery leg requires a product that's still needed at scale. Ask Christensen's question, which comes next, before you trust Chancellor's.

Rule 3: Respect the moat interaction. Run the cycle analysis even on moated businesses, but grade it differently. For a no-barrier industry the cycle sets the return. For a moated business it sets the stress test. A confirmed moat with a capacity wave inbound, like paints from 2024, keeps its verdict but earns a shorter or provisional CAP and a tripwire list. And a no-moat cyclical at stage 5 can be a better investment than a wide-moat business at a market-implied 25-year CAP. Price completes the argument, as always. That was Phase 3's lesson, and it's never suspended.

Rule 4: Grade management as cycle actors (the capital-allocation report card, reprised). Countercyclicality is the rarest capital-allocation skill. It means buying capacity cheap at stage 5, as UltraTech did through the IBC and Tata Steel did buying Bhushan Steel in 2018. It means refusing to build at stage 2, and returning cash when assets are dear. Reward it. A CEO who breaks ground at the peak because everyone else is doing it shows you the agency problem the cycle feeds on, and that's a governance red flag as real as any related-party note.

4.8 The false-positive problem: most "disruption" isn't

Now the corrective. "Disruption" has been debased into a synonym for "new competitor," and analysts who fall for it sell wonderful moats to buy stories. Christensen himself spent his last decade policing the term, in "What Is Disruptive Innovation?" (HBR, December 2015). Read it. It's the best twenty minutes of vaccination available. By his own test, Uber was not disruptive to taxis. It entered the mainstream market with a better product, which is sustaining innovation from a new player. And Tesla entered high, not low: a $100k sports car aimed at the most demanding customers is a classic sustaining trajectory. Whether those companies succeeded is a different question from whether the mechanism was disruption. And it's the mechanism that tells you whether the incumbent's moat protects it.

The four-question test. Run every claimed disruptor through all four questions. Real disruption answers yes to all of them. Your capstone's vulnerability section needs this test on at least one named threat.

1. Foothold: Is the entrant's product worse on mainstream performance dimensions but better on price/simplicity/accessibility, serving overserved customers or non-consumers? If it's simply better, as Uber was against taxis, it's a sustaining attack. An incumbent will fight that, because there's no margin-mix reason to flee, and incumbents usually win those fights. Then the only question is whether the incumbent's moat holds, and that's a job for the moat lenses, not disruption theory.
2. Trajectory: Is the entrant's offering improving on the mainstream dimensions fast enough to intersect what mainstream customers need? Cheap with no improvement path is just the low end forever, which makes it a segment and not a disruption. So name the technology curve that carries them upmarket, the way thin-slab casting carried the minimills.
3. Business-model economics: Is the entrant profitable, or structurally able to be, at its price point, on a cost structure that's actually different? VC-subsidized losses are stage-2 capital-cycle behavior wearing a disruption costume, and they confer no cost advantage. Your test is what happens when the funding cycle turns. India 2015 to 2024 is a museum of this, full of hyperfunded "disruptors" whose unit economics never closed. Capital-cycle theory and disruption theory filter them together, because real disruptors have better unit economics at the low price.
4. Asymmetry: Does an effective incumbent response require the incumbent to damage its own economics, cannibalizing its margin mix, breaking its channel, obsoleting its installed base? If the incumbent can respond without self-harm, it will. Banks co-opted mobile payments features, airlines matched LCC fares with fare classes, and Microsoft absorbed the browser. No asymmetry means no dilemma and no disruption, just competition.

The survivor case: Visa and Mastercard vs two decades of "disruption." Since roughly 2010 the payment networks have been declared doomed on a schedule: fintech wallets, Square and Stripe, crypto (twice), BNPL, big-tech pay buttons, real-time payment rails. Meanwhile Visa's net revenue roughly quadrupled, from ~$8bn in FY2010 to ~$36bn in FY2024, with operating margins in the 60s (illustrative, modeled on filed scale, so pull the 10-K). Its ROIC stayed in moat-hall-of-fame territory. Run the four questions and the failures are diagnostic. Most claimed disruptors failed Q1 in reverse. They were better-experience layers (Apple Pay, most wallets) that ride the networks' rails. A tokenized Visa transaction is still a Visa transaction, so the "disruptor" was actually a volume-growth channel. Square and Stripe failed to be threats for the same reason, seen from the merchant side. By making card acceptance easier, they expanded the network they were supposed to kill. They're customers wearing revolutionary t-shirts.

Crypto failed Q2. It had no trajectory toward mainstream payment jobs, because volatility, cost, and UX never intersected the "pay for coffee" dimension, through mid-2026 at least. Stablecoin rails remain the live watch item, and we flag that openly. BNPL failed Q3. Its growth was funded by credit losses, it had no structural cost advantage, and when rates rose its economics showed. And the networks systematically removed Q4 asymmetry by co-opting everything: tokenization, tap-to-pay, Visa Direct for real-time, acquisitions across open banking. None of it required cannibalizing their core economics, so they did it. Verdict: a two-sided network moat that twenty years of attackers strengthened. It rests on billions of credentials × ~130m acceptance points, plus the rules/fraud/chargeback infrastructure that is the actual product.

But finish the case, because the one attacker that made a real dent is the one the test flags: India's UPI. NPCI launched UPI in 2016, and a mandate made it free to merchants, with zero MDR on UPI from January 2020. By 2025 it was handling ~18 to 19 billion transactions a month (approximate; check at npci.org.in), and domestic debit-card usage collapsed as person-to-merchant payments moved to QR codes. Why did this succeed where the VC-funded wave failed? A state-sponsored public rail breaks Question 3's logic in the other direction. It doesn't need profit, so "unsustainable economics" isn't a bug. And regulation creates Question 4 asymmetry, because the networks can't match a zero-MDR rail without destroying their revenue model. Be precise about what moved, though. UPI displaced debit, which is a payments job. Credit cards, a lending job bundled with payments, kept growing, with ~110m cards in force and spends compounding at double digits (check the RBI data). Where the job differed, the moat held. It broke where the job was identical and the attacker didn't need returns. Brazil's Pix (2020) is the same experiment, replicating.

Three lessons here are permanent. (1) Often the deadliest "disruptor" of a profitable network is a regulator or public infrastructure, because it's exempt from the economics that filter private attackers. That's an India-specific alertness US playbooks won't teach you. (2) Moats are job-specific, so analyze the moat per job-to-be-done, not per company. (3) Surviving twenty attackers is evidence about the moat, and still no proof against the twenty-first.

4.9 Is this moat disruption-vulnerable? The audit

Different moat sources fail differently when the architecture shifts. Here's the pattern, distilled. Use it as the grid for your dossier's vulnerability paragraph:

Write out three audit questions, in prose, for every dossier. (a) What is the job each customer segment hires this company for, and does any current entrant do that job on a structurally different architecture? (b) Which of this company's moat sources are bolted to the current architecture (channel, technology, regulation), and which are architecture-portable?

(c) If the four-question test flags a real threat, what's the incumbent's rational response path? It can fight, if there's no margin-mix penalty. It can co-opt, absorbing the threat without self-harm. Or it can retreat, and the ratchet clicks. A moat with a "fight" or "co-opt" answer keeps its CAP. A "retreat" answer means the CAP you estimated with the moat lenses has an expiry mechanism, and your valuation must say so.

4.10 The erosion dashboard: five signals in the numbers

Both forces, the flood and the bypass, announce themselves in the statements before they finish. Your capstone scores five signals. Each comes with a metric, the pattern that counts as a hit, and a false-positive check that keeps you from fooling yourself. Run it on 6 to 10 years of data and score each signal 0 (clear), 1 (watch) or 2 (confirmed). Two or more confirmed signals with a nameable mechanism make an eroding verdict.

Signal 1: Gross-margin slippage. Track gross margin annually over 6 to 10 years. For moat purposes, use realization per unit against cost per unit where it's disclosed. A hit is a cumulative decline of ~150 to 300bps+ over 2 to 3 years that input costs don't explain. Gross margin is the home of pricing power, and pricing power is the moat's first casualty. Rule out three false positives. First comes a raw-material cycle, so deflate against the RM index, since an FMCG margin dip in a palm-oil spike is weather. Next is a deliberate mix shift into a lower-margin but value-creating line, which disclosed segment growth will show. Last is accounting reclassification, such as freight moved into COGS. What's left, the moat-relevant residue, is like-for-like price minus cost.

Watch the harvest trap too, which runs the other way. A rising gross margin on falling volume looks like strength, but it's the incumbent monetizing captivity on the way down. Film photography's margins rose for years, and so did the integrated mills' after each retreat. So pair margin with volume, always.

Signal 2: Rising churn / rising discounting. Track whatever line shows retention and price integrity in this business model. For SaaS that's NRR/GRR and logo churn. For FMCG it's trade discounts and scheme spends, so read the revenue-disaggregation and "customer incentives" notes, where a widening gross-to-net revenue gap is discounting in disguise. Autos show it in per-unit discounts and retail in promo depth. Anything contractual shows it in renewal rates. A hit is a sustained worsening, such as NRR sliding from 120% toward 100% or contra-revenue growing faster than revenue. It means customers are leaving or being paid to stay, and those are the same fact at different prices. Rule out one-time channel resets and COVID-type demand shocks. A deliberate strategic repricing is also a false positive, but management should say so before the numbers show it.

Signal 3: Share loss to a structurally different rival. Use your share table from the Greenwald lens, extended with one qualifier per rival: clone or different? Losing 1 to 2 points a year to a clone, with the same cost structure and the same channel, is competitive noise, since moats permit skirmishes at the border. Losing the same points to a structurally different rival is a different category of fact. That rival might have a lower cost architecture (EAF vs blast furnace), a different channel (D2C vs general trade, or exchange vs dealer network), or different economics (zero-MDR rail vs interchange). It counts differently because the rival's advantage compounds with scale and the incumbent's response options are asymmetric, as the disruption mechanism showed. Weight it several times heavier. For false positives, verify the rival's structural claim in its numbers. Does it have better unit economics, or just VC subsidies? That's Question 3 again. Then check whether the loss is in the incumbent's defended market or in a segment it rationally ceded. One is a moat eroding and the other is a moat being consciously narrowed, per Greenwald's niche discipline.

Signal 4: ROIC glide. Track 3-year rolling ROIC (and ROCE), compared peak-to-peak across the cycle against your cost-of-capital hurdle. A hit is each cyclical peak coming in lower than the last, say 28%, then 21%, then 15% against a ~12% hurdle. That's your CAP shortening in real time, and the spread halving is the market's mean-reversion base rate arriving on schedule. Three false positives can fool you. Denominator effects come first. A large capex program parked in CWIP, or recently commissioned assets not yet ramped, can depress ROIC for entirely healthy reasons. Check the CWIP and capacity-ramp disclosures and compute ROIC ex-CWIP as a cross-check. Acquisitions can also load goodwill into invested capital. Compute ROIC with and without it. And a one-off loss year can mislead you, which is what the rolling, peak-to-peak design handles.

Signal 5: Rising reinvestment with falling ROIIC. Use the tool pair from the capital-allocation work. Take the reinvestment rate (capex + acquisitions + ΔWC, net of depreciation, over NOPAT) and set it beside ROIIC = ΔNOPAT ÷ ΔInvested Capital in rolling 3 to 5-year windows. A hit here is the deadliest quiet signal in the set: the company is spending more and more to earn less and less on each new rupee. It comes in two variants, and both are terminal for the old CAP. Defensive capex pours reinvestment into the core just to hold position. Intel did this through the late 2010s. Its capex rose and its incremental returns collapsed, and it lost process leadership to TSMC anyway. Diworsification pushes cash into adjacent empires because the core can't absorb it, which is the conglomerate tell you've already learned to grade. False positives come in three kinds. Long-gestation projects need the capital lagged 2 to 3 years before you compute the window. A deliberate, disclosed land-grab can be fine if cohort-level evidence shows mature vintages earning well, so demand the cohort data. Accounting noise in NOPAT is the third, so normalize one-offs first.

Flows before stocks. One habit upgrades the whole dashboard. Wherever you can, compute each signal on the incremental cohort and leave the installed base aside: new-customer share, win rates on new contracts, NRR of the newest cohort, share of this year's market growth. Erosion always shows in the flows years before the stocks. Your installed base is the past voting, and the flow is the future voting. Kodak's installed film base was loyal to the end, but new photographers simply never arrived. A moat whose stock metrics are pristine and whose flow metrics are cracking is eroding, full stop. That explains why erosion feels "slow, then sudden". Stocks conceal what flows reveal, until the stock becomes last year's flows.

The direction-verdict grammar. Your dossier ends its dynamics section on one of four calls, and each has an evidence requirement:

• WIDENING: spread stable/rising, share stable/rising, ROIIC ≥ headline ROIC (the marginal rupee earns at least as well as the average one), current threats fail the four-question test, and the capital cycle is quiet or the company is the consolidator. It's rare, so demand all five.
• STABLE: spread high and flat across a full cycle, share table quiet, dashboard 0–1 hits with no mechanism, no stage-2 flood inbound.
• ERODING: ≥2 dashboard signals confirmed and a nameable mechanism, such as a capital-cycle flood, a four-question-passing disruptor, or captivity decay you can describe. Without a mechanism there's no erosion verdict. Two soft years is a slump, not a direction.
• BROKEN: spread at or below WACC ex-cycle, or share in structural motion to a different-architecture rival, so the share-stability test is now failing in real time. Your analysis reverts to Greenwald's no-barrier rules and to stage-based cyclical valuation.

Every verdict carries tripwires: 2 to 4 pre-committed, numeric, dated falsifiers. Think "gross margin −150bps two years running", "volume share below 48%", "NRR below 105%" or "capex/D&A above 2× industry-wide". Write them now, while you're calm. Phase 6 will show you how good your brain is at renegotiating evidence later, and Phase 9's sell rules run on the tripwires you set here.

4.11 The Phase 4 capstone: two moat dossiers

Everything in the phase converges here. You'll produce two complete moat dossiers, one on an Indian company and one on a US company, chosen from the candidate lists below. Each proves or rejects a moat in the numbers, grades its direction, and ends in a CAP you'd defend to a portfolio manager. This is the master map's M5 milestone and the certification of competency C4. Budget ~8 hours in total, roughly 4 per dossier. You've been rehearsing the pieces for five weeks, so the capstone is assembly instead of research from zero.

The required dossier structure (eight sections, ~2,000–2,500 words + tables each; the model answer in Worked Example 3 shows the standard):

1. Business & market definition (~half page). Say what the company does and give the revenue engine in three sentences. Then, critically, define the relevant market for share analysis and defend it in one paragraph. Market definition rigs every downstream test, so state your source (CRISIL/industry body/company presentations).
2. The numbers-proof table (10 years, one screen, and the dossier's spine). Rows: revenue; gross margin; EBIT margin; ROIC and/or ROCE vs the hurdle; ROIIC in two or three rolling windows; FCF conversion (FCF/PAT or FCF/NOPAT); reinvestment rate; capex/D&A; share count. For sources, use screener.in/Tijori (India) and TIKR/stockanalysis.com plus the 10-K (US), the toolkit you set up at the start. Every later claim must point back to a row here. Mark estimates and any hand-adjustments, using the returns-on-capital conventions you already know.
3. Lens 1, the five sources. Name the source or sources, with the mechanism named in cash-flow terms, and clear the false-moat checklist explicitly. Don't write "strong brand". Write "habit-based captivity visible as 20 years of price increases ≥ inflation with share stable; table 2 row 3".
4. Lens 2, Greenwald. Give the barriers-to-entry verdict with the share-stability table computed. That means shares of the relevant market at three snapshots ~4 to 5 years apart and the average absolute movement per firm per year. Apply the threshold, then run the incumbent-profitability cross-check.
5. Lens 3, Helmer. Test all seven powers. State Benefit + Barrier for any power you claim, and explicitly reject the rest with a sentence each. Then place the power on Helmer's clock, from origination through takeoff to stability.
6. Lens 4, Mauboussin & the CAP. Set your CAP estimate beside the market-implied CAP from a reverse-DCF-lite, and interpret the gap. Give your estimate as a range, with the reasoning: base rates, source durability, reinvestment runway.
7. Dynamics. (a) Your one-line cycle-position call with its gauges and sources. (b) Your disruption audit, meaning the job-to-be-done, the four-question test run on the most credible named threat, and the vulnerability grid row(s) that apply. (c) Your erosion dashboard, with all five signals scored 0/1/2 and the numbers shown.
8. Verdict (~half page). It's the part a PM would actually read. Write it last, and place it first if you like memo style. State the moat (none / narrow / wide) and its source or sources. Give the direction (widening/stable/eroding/broken) with the mechanism, and the CAP in years, as a range. Name the single most important thing to watch, and set 2 to 4 numeric, dated tripwires that would change your mind.

Candidate list, India (choose one). These were chosen for three things. Each has 10 years of accessible data, and where the listing is younger you can extend with DRHP or annual-report history, with a minimum of 8 years of statements. Each can be analyzed without Phase 5 sector mechanics, so there are no banks/NBFCs/insurers yet. And together they give a deliberate spread of likely verdicts. Some have wide moats, some narrow, some eroding, at least one probably none. Your milestone says prove or reject, and a well-evidenced "no moat" or "eroding" dossier scores as highly as a wide-moat one:

Candidate list, US (choose one):

Rules. (1) Pick one from each list. You may substitute a company off-list only if you write one paragraph justifying data availability and analytical fit, and it isn't a bank/NBFC/insurer. (2) You may not pick a company fully worked in this phase: Asian Paints, DMart, Costco, Coca-Cola, Visa, TCS, IndiGo, UltraTech/cement majors, Nucor, or Colgate-Palmolive India (the model dossier). Brief appearances elsewhere in the phase (CDSL, IEX, Info Edge, Pidilite, Titan) are fine, because the dossier demands original numbers work either way. (3) Primary sources are mandatory. Skim the latest annual report or 10-K the way you learned to at the start and read one investor presentation. Pull screener/TIKR 10-year data too, and note every number's source in the dossier. (4) Time-box it to ~4 hours each. Phase 8's deep dive gets three days, but this is the moat slice only, and part of the skill being certified is finishing.

Grading. Self-grade both dossiers against the rubric in the Applied mini-project section, then against the model dossier (Worked Example 3). Your pass bar is ≥24/30 per dossier, with no zero on the four critical criteria (numbers-proof table; share-stability computation; erosion dashboard; falsifiable verdict). One revision cycle is expected: grade, fix, regrade. When both dossiers pass, the M5 milestone is met. File both in your knowledge system and copy the verdict blocks into the moat section of your Mental-Model & Moat Checklist, the phase artifact you'll carry into every future analysis.

Common mistakes & how experts think differently

Mistake 1: Analyzing demand and ignoring supply. Ask an analyst about any industry and the reflex is a demand forecast ("India's cement demand will grow 8% because infrastructure"). Fine. But that's half the equation, and the harder-to-know half. Experts invert the effort. Demand is a range you humbly bracket, while supply is a schedule you count item by item: announced capacity, order books, capex/D&A. If your industry thesis has five paragraphs on demand and one sentence on supply, you've written the consensus.

Mistake 2: Trusting the trailing P/E at the cycle top. At stage 2 to 3, earnings are at their peak and multiples compress, so the screen says "cheap" just when future supply guarantees earnings will halve. You already know the rule for cyclicals, and now it has its mechanism attached: value them on mid-cycle economics and cycle position, never on trailing earnings. In a capex-flooded sector, the lowest trailing P/E is frequently a sell signal.

Mistake 3: Confusing the cycle with erosion (in both directions). A moated business having a cyclical margin year (an input spike, a demand air-pocket) is weather. Share is stable, the relative cost position is intact and the dashboard is clean. Calling it erosion makes you sell compounders at the bottom. Worse is the reverse error: blaming "the cycle" while a structurally different rival takes the flow metrics, which is Signal 3. Relative data always tells them apart: share tables, relative margins, win rates. Cycles hit everyone. Erosion hits you.

Mistake 4: Calling every entrant "disruption." Most entrants are clones, and beating clones is what moats are for. That's what your moat verdicts certified. A four-question test is a mechanism, and the word "disruption" on its own isn't. Experts use the term only when all four answers are yes. Otherwise they write the more useful sentence: "a well-funded clone is attacking; the moat should hold; here is the tripwire if it doesn't."

Mistake 5: Getting the speed wrong on both ends. Novices assume disruption is fast. It took minimills thirty years to reach sheet steel, and streaming twelve to pass cable. They also assume erosion is linear, when it's slow-then-sudden, because installed-base stocks conceal collapsing flows until the stock has become the dead flows. So experts watch flow metrics for direction and expect long runways. For the same reason, they refuse to pay for the terminal years of a threatened CAP even when near-term earnings look untouched.

Mistake 6: Reading rising capex as automatically bullish. "Reinvestment runway" is only a virtue at high ROIIC. Defensive capex, spending more to stand still, produces the identical cash-flow-statement line with the opposite meaning. Experts never cite reinvestment without its return. A rising reinvestment rate paired with ROIIC falling through WACC is Signal 5, and it has marked the top of more admired franchises than any headline story above.

Mistake 7: Grading direction by the latest narrative. After a good quarter the moat is "widening", and after a bad press cycle it's "eroding." That's recency masquerading as analysis. Phase 6 will name the bias properly, and the habits here starve it of oxygen. Experts pre-commit to a written dashboard, numeric tripwires with dates, and a rule that direction verdicts change only when the dashboard changes. Your capstone rubric gives a full criterion to that habit.

Worked examples

Worked example 1: Reading a full capital cycle from company numbers (synthetic, then mapped to real steel)

Bharat Ispat Ltd is a synthetic steelmaker, with all numbers invented for clean arithmetic. Here it is with its industry, over ten years:

Step 1: compute the supply/demand race. Capacity CAGR from FY1 to FY6 is (146/100)^(1/5) − 1 = 7.9%/yr. Over the same window, demand CAGR is (106/88)^(1/5) − 1 = 3.8%/yr. Supply outran demand by ~4 points a year for five years, so utilization had to fall from 88% to 73%. You didn't need a forecast for that. Subtraction was enough.

Step 2: date the stages from the gauges. Stage 1 runs FY1 to FY3, with ROCE climbing from 22 to 28% and the price rising. Stage 2 runs FY2 to FY4, when capex/D&A jumps to 2.8 to 3.2×. That's the flood, committed at peak prices. Stages 3 to 4 run FY4 to FY7. Capacity arrives late and together, the price falls 30% peak-to-trough, and ROCE collapses to 3%, far below a ~12% hurdle. Stage 5 runs FY6 to FY9, as capex/D&A falls to 0.6 to 0.8×. It's starvation, and in the real world it's when the leveraged expanders change hands. Stage 6 runs FY8 to FY10. Demand grows into the stagnant capacity, utilization recovers from 73 to 84%, and ROCE climbs back through the hurdle. Now look at FY10's capex/D&A, back at 1.4× and rising. The next reel is loading.

Step 3: the investor's timeline. FY3 was the worst purchase, with record ROCE, record optimism, and the screen's "cheapest" trailing P/E (on peak E). FY7 to FY8 was the best window. Returns were below the hurdle and capitulation was everywhere, with assets available below replacement cost. Your way in was the strongest balance sheet in the sector, because stage 5 kills the weak before it pays the survivors.

Step 4: map it to the real thing (approximate history; check against JPC/World Steel data and the filings). India's steel capacity ran from ~80 MT (FY11) to ~122 MT (FY16) on debt-funded expansion. It ran straight into decelerating domestic demand and China's 2015 export flood, ~112 MT of Chinese exports that year. Domestic flat-product prices fell ~30%. By June 2017, five of the RBI's twelve large-default IBC referrals were steelmakers: Bhushan Steel with ~₹56,000 crore of debt, Essar Steel, Bhushan Power, Electrosteel, Monnet. Stage 5 then ran its textbook course. Tata Steel bought Bhushan Steel through the IBC (2018, ~₹35,000 crore), and ArcelorMittal took Essar (2019, ~₹42,000 crore). Capacity changed hands at deep discounts to greenfield cost. When demand normalized in FY21 to FY22, the consolidated survivors printed the best spreads in a decade. Years before it paid or punished anyone, every stage showed up in Gauges 1 to 6.

Worked example 2: Running the erosion dashboard end-to-end (synthetic SaaS)

MeterWorks is synthetic, an enterprise usage-analytics software company that sells through a sales team. Here are six years of data, with WACC at ~9% (USD):

For context, Plainly is a product-led (self-serve, freemium) rival with ~1/5th MeterWorks' sales cost per dollar of new ARR. It won overserved smaller customers first and is now moving upmarket.

Start with Signal 1, gross margin. It went from 76% to 68%, −800bps over five years. Hosting costs were flat (that's given), so this is price and mix, not input costs. Score 2.

Signal 2, churn and discounting, looks just as bad. NRR fell from 122% to 96%, so the installed base has gone from expanding 22% a year to shrinking. Below 100% is the leaky-bucket line, which Phase 5 will formalize. Meanwhile S&M/revenue rose from 36% to 49% to buy less growth. That's paying customers to stay. Score 2.

On Signal 3, share to a structurally different rival, installed-base share (the stock metric) still looks respectable. The flow metric gives it away: new-logo win rate fell from 60% to 38% against a rival with a structurally lower go-to-market cost. Is that structural, or a subsidy? Plainly's PLG motion is a different cost architecture, so the structural claim checks out. Score 2.

Signal 4, the ROIC glide, is simple to read. ROIC fell from 24% to 10% against a 9% WACC, so the spread has gone from 15 points to ~1. Score 2.

Signal 5, ROIIC, closes the case. Take the early window first, FY1 to FY4: ΔNOPAT ÷ ΔIC = (24−18) ÷ (142−75) = 6/67 = 9.0%. That's already only ~WACC. In the late window, FY3 to FY6, you get (22−23) ÷ (195−118) = −1/77 ≈ −1.3%, negative incremental returns while the reinvestment rate rose. MeterWorks is spending more to shrink. Score 2.

Dashboard: 10/10. The mechanism is nameable too, which the verdict requires, so run the four-question test on Plainly. It has a foothold in overserved SMBs through lower price and more simplicity, so that's a yes. Its trajectory is improving upmarket as enterprise features ship, which is another yes. It's profitable at its price on a structurally cheaper GTM, so yes again. And on asymmetry, for MeterWorks to match freemium self-serve would cannibalize its sales-led enterprise pricing and break its channel. Yes. Real low-end disruption; verdict: ERODING, arguably BROKEN within two years. Set the tripwires now. NRR < 95% or win rate < 30% confirms BROKEN, and ROIC < WACC for two consecutive years confirms it in returns.

Valuation follows from the CAP rules: CAP ≈ 0 to 3 years, and steady-state-minus value. Any multiple paying for duration here is paying for the past. Notice when the verdict became available. It was FY3 to FY4, with growth still +11%, ROIC still 17% and the headline optics fine, because the flows (win rate, NRR trend, GM slope) had already voted. That's what the dashboard is for.

Worked example 3: THE MODEL DOSSIER: Colgate-Palmolive (India)

This is the standard your two capstone dossiers are graded against, the full required structure, executed. All figures are illustrative, modeled on ~FY15 to FY25 filing scale from screener.in, annual reports, and press-reported Nielsen-style share data. Pull the actuals before reusing any number. Colgate India is deliberately not on your candidate list.

1. Business & market definition. Colgate-Palmolive (India) Ltd sells oral care: toothpaste (~80%+ of revenue), toothbrushes and mouthwash, plus a small personal-care line. It sells through India's general trade, reaching on the order of 6 million outlets directly and indirectly (check the current figure). It also sells through modern trade and, increasingly, e-commerce/quick-commerce. Its revenue engine is low-ticket (₹10 to ₹120), high-frequency, habit-driven repeat purchase, the classic "razor-that-is-its-own-blade" repeat model. Relevant market: Indian toothpaste (value share), per Nielsen-style panels as quoted in company presentations and the trade press. Toothpaste is the choice because ~four-fifths of profit comes from it and all serious competition happens there. A "total oral care" definition would flatter stability by mixing in low-competition brushes, so always state choices like this one.

2. Numbers-proof (10 years, illustrative; rebuild from screener.in before relying on any cell):

Read it this way. A ~55 to 95%+ ROCE sustained for a decade against a ~12% hurdle is a spread so wide that this table alone nearly settles the existence question. But look at two quieter rows that will drive the verdict: volume growth hugging 0 to 4%, and payout ~90%. Almost nothing is reinvested, and the growth identity g ≈ reinvestment × ROIIC caps organic value growth near category growth.

3. Lens 1: five sources. Intangibles (brand-as-habit). Colgate has ~80 years in India, and brushing-with-Colgate is bought on reflex, evidenced not by fame but by behavior. Decades of price increases at or above CPI with volume share holding around half the market are your pricing-power evidence. And its marketing defends habit, with school programs seeding the next captive cohort, instead of arguing features. Cost advantages via scale (shared with Lens 3's scale economies). A ~50% share means A&P and a 6-million-outlet distribution system are carried at roughly half the per-unit cost of the nearest rival, by the fixed-cost arithmetic of the Greenwald lens. False-moat checklist cleared. Rivals execute well and still don't close the gap, so it isn't execution. Share is the evidence and habit is the mechanism, so it isn't share-as-moat. Nor is it regulation. Sources: brand/habit + scale-based cost advantage.

4. Lens 2: Greenwald. Share-stability table, Indian toothpaste value share (illustrative; reconstruct from Nielsen citations):

Average absolute movement is ≈ 0.3 to 0.4 points per firm per year, inside Greenwald's stable band (<~0.5). One nuance gets flagged: the leader's drift is one-directional, −6 points over the decade, and essentially all of it went to the ayurvedic/naturals flank. Incumbent profitability ran far above the hurdle throughout. So the cross-check passes and the barriers are confirmed (captivity + scale).

Patanjali's episode (FY16 to FY19) is a natural experiment worth a paragraph. An entrant with free celebrity media, a differentiated naturals story, and its own store network took ~5 to 6 points of category share at peak. Then it stalled and faded. Colgate counter-launched (Vedshakti in 2016, plus a naturals range), and the entrant had to fight for the same general-trade shelf it had bypassed. An attacker converged into a clone, and the wall held. That's what a tested barrier looks like in data.

5. Lens 3: Helmer. Branding: YES. Its benefit is a price premium and reflex purchase. Its barrier is hysteresis, since eight decades of accumulated trust can't be replicated by spend, and Patanjali's fade is the exhibit. Stage: stability. Scale economies: YES. Here the benefit is the lowest per-unit A&P/distribution cost, and the barrier is that a challenger must near-halve its unit economics to match reach. Stage: stability. Switching costs: NO, as Helmer defines them. There's no contractual or data lock-in, and habit is housed under branding. Network economies: NO. Counter-positioning: NO, because Colgate is the incumbent (the threat version shows up in the dynamics section). Cornered resource: NO. Process power: NO. That's two powers, both stability-stage, consistent with Lens 1, as it should be. If the lenses disagree, you've made an error somewhere.

6. Lens 4: CAP. On the spread, the dashboard supports a CAP of 15 to 20 years, on habit-grade captivity plus scale, tested by a live entry and held. Base rates demand humility beyond that. But keep the CAP's two dimensions separate. The spread's duration is long, while the reinvestment into it is thin. Payout is ~90% and category volume grows at ~low-single-digit rates. Per-capita consumption has a real but slow runway, roughly 200g/year vs ~500g+ in developed markets, a decades-long grind. For the market-implied read, run a reverse-DCF-lite. At the ~40 to 50× earnings this franchise has typically commanded (as of mid-2026), the price needs both the long CAP and healthy perpetual growth. Per the guardrails, your dossier's job ends with that observation. The moat verdict is emphatically wide, while the price embeds close to the full optimistic case. That's a valuation judgment about premium multiples, not a moat judgment.

7. Dynamics. (a) Cycle position. There's no commodity-style capacity cycle here. Capex/D&A is ~1.0 and the business is asset-light, with negative working capital. Instead, the capital cycle flows through competitive entry. First came the FY16 to FY19 Patanjali flood, which has receded. Then came the 2019 to 2022 D2C oral-care funding wave, receding since the 2022 funding winter (count the survivors, not the launches). Quick-commerce shelf competition is inbound. In one line: "Stage: quiet; last flood repelled; next wave is channel-borne, not capacity-borne."

(b) Disruption audit. Colgate's job is "trusted daily oral-hygiene reflex." Run the four-question test on the most credible structural threat, digitally-native brands riding e-commerce/quick-commerce rails, a channel that neutralizes distribution density. Foothold: partially, in premium/naturals niches, not the overserved low end. Trajectory: weak so far. It's niche positioning with no cost-side improvement curve. Profitable at its price point: largely no. That's a Q3 fail, since most D2C oral care has never cleared contribution-margin hurdles at scale (check the current cohort). Asymmetry: no. Colgate lists on the same rails without harming its economics, and it has co-opted every channel shift since the kirana. Verdict: no live disruption; the real exposure is the vulnerability grid's distribution row. Habit is architecture-portable, but the 6-million-outlet advantage matters less in a warehouse-app world.

(c) Erosion dashboard. On S1, gross margin is rising (from ~64 to 70%), but volume is near-flat, so run the harvest check. Pricing-led growth with stable share and sustained A&P is premiumization, not harvest. Score 1, and watch the pair. On S2, A&P is steady at ~12 to 13% of sales with no margin-propping cuts, and trade terms are stable, so 0. On S3, share loss runs −0.5 to 0.6 pt/yr to structurally similar rivals. Dabur runs the same GT model and Patanjali converged into it. Call it drift, not bypass. It scores 1. S4 shows no ROIC glide, since ROCE is rising, so 0. On S5, reinvestment is tiny by design and what little is deployed earns well. Its constraint is runway, not returns, so 0. Add the capital-allocation footnote: a 90% payout at 95% ROCE is sound capital allocation, not decay. Total: 0 confirmed, 2 watch, so no erosion mechanism is nameable.

8. Verdict. Moat: WIDE. Its sources are habit-grade brand captivity interlocked with scale economies in distribution and A&P, the consumer edition of the "king combination" from the moat work. Direction: STABLE, on a dashboard of 0/2/0/0/0 with both watch-items explained, and one live entry repelled inside the window. CAP: 15 to 20 years on the spread. Value growth is still capped near category growth by a ~90% payout, so this is a wide moat around a slow pool. The one thing to watch is channel mix. Suppose e-commerce + quick-commerce share of category sales (single digits today, rising) crosses ~25% while Colgate's online share stays materially below its offline share. That would mean the scale leg of the moat is being bypassed even while the habit leg holds. Tripwires (dated, numeric): (i) volume share below 48% for two consecutive years; (ii) gross margin −150bps+ over two years ex input inflation; (iii) A&P cut below 10% of sales while EPS is defended; (iv) online category share > 25% with Colgate's online share more than 5 points below its offline share. Any two of these, and you downgrade direction to ERODING and re-open the CAP.

Grade this dossier yourself against the rubric below before writing your own two. It's built to score 28 to 30 out of 30, and seeing why is half the training.

Worked example 4: India: a capacity wave hits a moat, and what the ROIC fade does to value (synthetic)

RangSetu Paints (synthetic) earns a 26% ROCE against a 12% INR hurdle, a 14-point spread, on ₹5,000 crore of invested capital. In FY0 a diversified industrial group commits ~₹10,000 crore to a greenfield entry. It's the paint story from the capital-cycle section, playing out: returns high enough for long enough summon a wave big enough to test the wall. Nobody's asking you to predict the winner. Your job is to price the risk that the spread mean-reverts. Use the classic mean-reversion pattern from the returns work, where the spread decays toward the hurdle and the ROIC level follows. Here it decays at 25% a year, so spread_t = 14 × 0.75^t points and ROIC_t = 12% + spread_t. Hold invested capital flat at ₹5,000 crore to isolate the fade, and discount economic profit (spread × ₹5,000 cr) at the 12% hurdle.

Step 1: read the fade as a CAP. The spread crosses below ~2 points, the practical "moat effectively gone" line, in year ln(2/14) ÷ ln(0.75) = 6.8. So CAP ≈ 7 years (FY6 spread 2.5 pts, FY7 1.9). Notice that this is a number. It comes from one mean-reversion assumption you can argue about out loud, and that beats a vibe.

Step 2: value the excess-return strip two ways. Suppose the 14-point spread persisted forever, which is the bull's silent assumption. Then the excess-return value is a flat ₹700 cr/yr perpetuity: 700 ÷ 0.12 = ₹5,833 cr. If it fades at 25% a year from today instead, as in the table, the strip is a declining geometric perpetuity. Its ten visible PV rows already sum to ₹1,393 cr, and the infinite tail brings it to ₹1,419 cr.

Step 3: the punchline. 1 − 1,419 ÷ 5,833 = 75.7%. More than three-quarters of the "moat value" was duration, a bet that the spread lasts, and a capacity wave is an attack on duration itself. The steady-state business (₹5,000 cr earning its 12% hurdle) is untouched. What the wave threatens is the strip, and the strip is what a premium multiple pays for. So here's the arithmetic behind Rule 3 of capital-cycle investing. A confirmed moat facing an inbound flood keeps its verdict but earns a shorter, provisional CAP, and the valuation must show the fade instead of assuming it away. Pre-commit a tripwire now. If industry gross margin falls 200bps within two years of the entrant's first full-capacity year, or RangSetu's volume share stays below its FY0 level for two straight years, the CAP reopens.

Worked example 5: US: a share-loss trajectory, and why the flow votes before the stock (synthetic)

Meridian (synthetic) is a US field-service-software incumbent with $4.0bn of revenue, 40% of a $10.0bn category growing 8% a year. A product-led, self-serve rival with a structurally cheaper go-to-market takes almost all the category's growth. Meridian lives off renewals and a trickle of new logos, and grows just 2% a year. Watch the two share numbers diverge. Its stock is installed-base share, and its flow is share of this year's growth. It's the flows-before-stocks distinction, made numeric.

Step 1: the flow voted in year 1. After year 1, Meridian's stock share is still a comfortable-looking 37.8%. But its share of the year's growth is 10.0%, because it captured only $0.08bn of the category's $0.80bn expansion. Installed base is the past, and it conceals what the flow, the future, already shows. Across Y0 to Y6 the rival takes 91.4% of all category growth.

Step 2: date the tripwire. Stock share slides below 30% only in Year 6 (30.1% in Y5, 28.4% in Y6), five years after the flow collapsed. If you police stock share ("alert if share < 30%"), you're warned in Year 6. If you police the flow ("alert if share of new growth < ~15%"), you're warned in Year 1. A five-year lead like that is the professional payoff of the flows-before-stocks habit.

Step 3: what it does to a price. Suppose the market prices Meridian as if it holds its category pace, with revenue compounding at 8%. Your share-loss underwrite (2%) puts Year-10 revenue at 4.0 × 1.02^10 = $4.88bn, against the priced-in 4.0 × 1.08^10 = $8.64bn. The market is paying for 77% more revenue than your trajectory delivers. A share-loss trajectory can sound like a soft "narrative risk", but it's a specific, compounding haircut you can put a number on: Signal 3, turned into a valuation.

Practice set

Work in order, because the set ramps from guided computation to timed integration. Write your answer before reading the solution, every time.

Problem 1 (guided: the workhorse gauge). Here are aggregate figures for the top five listed players in two Indian industries (synthetic numbers). Industry A (specialty chemicals). Combined capex ₹9,600 crore, combined D&A ₹3,200 crore, CWIP up 80% year over year, four QIPs and two IPOs in the sector in 18 months. Industry B (paper). Combined capex ₹2,100 crore, combined D&A ₹2,800 crore, no greenfield announcement in six years, two mills closed last year. Compute capex/D&A for each, name each industry's cycle stage, and state the expected direction of returns over the next 3 to 5 years.

Solution. For Industry A, 9,600 ÷ 3,200 = 3.0×. That's deep capacity-building territory (>2×). Swelling CWIP confirms more supply is already inbound, and eager issuance is the stage-2 financing tell. Call it stage 2 to 3: supply is coming. Unless demand growth is extraordinary, spreads compress from here, so be afraid, whatever the current P&L shows. For Industry B, 2,100 ÷ 2,800 = 0.75×. It isn't even replacing depreciation. With closures and zero announcements, this is stage 5 starvation. Expect tightening utilization and recovering returns, subject to the demand test. Writing/printing paper is in secular decline while packaging grows, so you must segment before buying the "starved" story. Put the two side by side. A's returns look great and are about to be attacked by A's own capex. B's returns look poor and are being repaired by B's own austerity.

Problem 2 (guided: anatomy and staging). (a) Put these six events in cycle order: capital flees · returns collapse · high returns · returns recover · capacity arrives · capital floods in. (b) An industry shows utilization 68% and falling, capex/D&A 0.7×, two bankruptcies this year, assets trading at ~45% of replacement cost, forced rights issues to repair balance sheets. Which stage, and what is the classic investor error here?

Solution. (a) High returns → capital floods in → capacity arrives → returns collapse → capital flees → returns recover. (b) Stage 5 (capital fleeing: sub-replacement asset prices, capitulation issuance, exits). Investors' classic error here is extrapolating the despair: demanding to see good earnings before buying, when at stage 5 they arrive only after the consolidation has repriced the assets. A better move is to verify the demand test, then underwrite the strongest balance sheet / lowest-cost survivor at a discount to replacement cost. Accept that the P&L will look terrible when you buy. That's why the price is available.

Problem 3 (numeric: the published future). A shipping segment has an order book equal to 45% of the current fleet, deliverable evenly over three years. Scrapping runs ~2% of the fleet per year. Demand (tonne-miles) is growing ~3%/year. (a) Compute gross and net annual supply growth. (b) What happens to utilization and rates? (c) What order-book percentage would have kept supply and demand balanced?

Solution. (a) Gross supply CAGR = (1.45)^(1/3) − 1 = 13.2%/yr, and net of scrapping it's ≈ 13.2 − 2 = 11.2%/yr. (b) Supply outruns demand by ~8 points a year for three years, so utilization collapses ~20+ points cumulatively. In a commodity service, rates fall toward the cash operating cost of the marginal vessel. They can stay below it for long stretches, too, since ships exit only by scrapping. (c) Balanced net supply is 3%, which means gross supply of 5%/yr and a three-year order book of (1.05)^3 − 1 ≈ 15.8% of the fleet. Anything much above ~16% in this configuration is a published overshoot. Notice what you never needed: a demand forecast better than "about 3%." The order book did the work.

Problem 4 (classification drill). Classify each as sustaining innovation, low-end disruption, new-market disruption, or not disruption at all, with one line of reasoning each. (a) A fintech offering a slicker app and richer rewards on premium credit cards to affluent urban customers. (b) A ₹299 at-home diagnostic panel targeting households that have never used a lab. (c) A discount broker at ₹0 brokerage with bare-bones execution and no advice, drawing self-directed traders from full-service firms. (d) An EV maker entering with a ₹45-lakh luxury SUV. (e) A new paint company with the same dealer-distribution model and a 10% lower price. (f) Single-use shampoo sachets priced at ₹1 to 2, sold through kiranas to first-time users.

Solution. (a) Sustaining. It's a better product for the same mainstream customers. Incumbents fight these and usually win, because there's no margin-mix reason to flee. (b) New-market. It competes with non-consumption and stays invisible on incumbent labs' share dashboards while it scales. (c) Low-end. Overserved customers are offered good-enough at structurally lower cost, and the full-service incumbent's rational retreat is upmarket to advisory/HNI, which is what happened in Indian broking. (d) Sustaining, as a high-end entry. That's Christensen's own Tesla point. It could be a great business, but its mechanism is head-on, so the incumbent's response is fight, not flight. (e) Not disruption. It's a clone with a discount, and this is what moats are for. Moat machinery applies, and the four-question test never triggers. (f) New-market. It created consumers the bottle never reached, and the historical FMCG sachet wave is the canonical Indian instance.

Problem 5 (short answer: the four-question test on a real transition). India's lead-acid battery duopoly (Exide, Amara Raja; use public materials) faces lithium-ion. Run the four-question test and the vulnerability grid, and state what kind of threat this actually is. (Model answer follows; write yours first.)

Solution. On Q1, the foothold, lithium doesn't enter cheaper-and-worse at the low end. It enters better and dearer, through new vehicles (EVs) and demanding stationary uses, so the classic low-end template fails. But the new-market door is open, because EV two/three-wheelers are substantially new vehicle categories. On Q2, trajectory, the answer is strongly yes. Cell costs have fallen for a decade (check the current $/kWh) and energy density keeps improving, so the curve points at the mainstream. On Q3, economics, it's yes at cell scale. But that scale is global, held by giga-scale specialists, and not national. Q4, asymmetry, is partial. Incumbents' profitable lead-acid replacement annuity (the stock) doesn't stop them investing in lithium, and both have announced cell plans. But their cost position in cells starts far behind the specialists, and every rupee of lithium capex earns less than their legacy ROCE for years. That's a soft version of the margin-mix ratchet.

Classification: not textbook low-end disruption, but an architecture substitution arriving via new-market vehicles, on a long fuse. Replacement-battery stock decays over 10 to 20 years even after new-vehicle flows flip. On the vulnerability grid, the duopoly's cost advantage and recycling/distribution network are architecture-specific (lead), so the moat doesn't port. Brand and dealer relationships port partially. Watch flows, not stocks: EV share of new two/three-wheeler sales and the telecom/UPS tender chemistry mix. Don't watch replacement-market revenue, which will look healthy right up to the cliff. For the legacy moat, the direction verdict is STABLE near-term cash and ERODING terminal value. A DCF must reflect that shape by shortening the CAP and fattening the near years.

Problem 6 (numeric: run the dashboard). "Trishul Tiles" (synthetic branded-tiles maker; input costs flat throughout; WACC/hurdle 12%):

Score all five signals (0/1/2 each), compute ROIIC for FY1 to FY4 and FY3 to FY6, name the mechanism, and give the direction verdict with one tripwire.

Solution. S1 scores 2: gross margin fell 800bps over five years with flat inputs. S2 scores 2, as dealer schemes rose from 3% to 7% of sales with price integrity leaking. S3 scores 2 because share fell 4 points to a rival with a structurally different architecture, factory-direct D2C with a ~25% landed-cost advantage. It's verified structural, not a subsidy. S4 scores 2, with ROIC gliding from 19% to 11%, now below the 12% hurdle. S5 scores 2 as well. ROIIC for FY1 to FY4 is (86−80) ÷ (615−420) = 6/195 = 3.1%, and for FY3 to FY6 it's (84−86) ÷ (780−540) = −2/240 = −0.8%. That's heavy reinvestment (IC +86% over the period) at near-zero-to-negative incremental returns. Your dashboard reads 10/10 with a nameable mechanism, the structural cost bypass of the dealer channel. Verdict: ERODING, effectively BROKEN, because the spread has crossed below the hurdle. Revert to Greenwald's no-barrier rules and value it as operations, with CAP ≈ 0. Management's channel-defending capex is Signal-5 defensive spending and should be challenged. For the remaining longs, set a tripwire: TileKart share >12% or ROIC <10% for a second year means the thesis is dead.

Problem 7 (discrimination: same share loss, different verdicts). Two market leaders each lose 3 share points over four years, from 44% to 41%. Case X: points go to a rival running the identical business model at similar gross margins. The leader's win rate on new contracts is stable, and the industry is mid-downcycle. Case Y: points go to a rival with a structurally lower-cost model. Here the leader's share of new customer signings has fallen from 40% to 22%, and industry demand is fine. Grade Signal 3 for each and explain the different verdicts.

Solution. Case X scores 0 to 1. A clone taking points in a downcycle is competitive noise. Moats permit border skirmishes, and cycle effects hit reported share through mix, so check segment exposure. Watch it, but don't issue a verdict. Case Y scores 2. Structurally different economics plus collapsing flow share (new signings), while stock share still looks respectable, is the erosion signature. The installed base is voting for the past and the flows for the future. So the headline share loss is identical and the meanings are opposite. It's the reason the dashboard qualifies Signal 3 by rival architecture and insists on flow metrics. A share table alone is a photograph, and the dynamics toolkit turns it into a movie.

Problem 8 (numeric: the quiet killer). A company reports NOPAT (₹ cr) 200, 210, 230, 260, 268, 272 over FY1 to FY6, and invested capital 1,000, 1,150, 1,350, 1,600, 1,900, 2,250. WACC 10%. Compute ROIIC for FY1 to FY4 and FY3 to FY6, describe the reinvestment trend, and state what Signal 5 says, including the one disclosure you would demand before finalizing.

Solution. For FY1 to FY4, ΔNOPAT ÷ ΔIC = (260−200) ÷ (1,600−1,000) = 60/600 = 10.0%. That's right at WACC, so the marginal rupee already creates nothing. For FY3 to FY6, (272−230) ÷ (2,250−1,350) = 42/900 = 4.7%, far below WACC, while the reinvestment pace accelerated. IC grew ~₹250 to 350 cr/yr late against ~₹150 to 200 early. Signal 5 is confirmed (2): the company is spending more to earn less per rupee. Headline ROIC (272/2,250 ≈ 12%) still exceeds WACC, so the average is coasting on the past while the margin destroys value. It's the trap the capital-allocation work warned you about. Demand one disclosure, a segment/vintage split of the new capital. Same-business capex means defensive spending, a moat under attack. New-adjacency capital means diworsification, a different disease with the same P&L symptom. And a long-gestation project (check CWIP and ramp timelines) is the false positive that would soften the score.

Problem 9 (timed: 10 minutes, integration). Sector A: capex/D&A 2.7×, fourteen IPOs in 18 months, capacity pipeline +12%/yr vs demand +6%/yr, a "decade-long supercycle" narrative, trailing P/E 11× on record margins. Sector B: capex/D&A 0.65×, zero IPOs in four years, nine players consolidated to three, assets trading at ~60% of replacement cost, trailing P/E 30× on depressed earnings. Which sector does capital-cycle logic favor, what would you buy within it, and what single piece of evidence could overturn the call?

Solution. B. Sector A is at stage 2 to 3. Published supply (+12% vs +6%) guarantees spread compression, and the 11× trailing P/E is peak-E optics (Mistake 2). Its IPO cluster and supercycle vocabulary are the financing and narrative tells. Sector B is at stage 5 to 6, with starvation, consolidation to an oligopoly, and sub-replacement asset values. Its 30× trailing P/E is depressed-E optics that says nothing, since mid-cycle earnings would re-rate it. Within B, buy the lowest-cost operator with the strongest balance sheet, because survivorship is the mechanism of the recovery. What could overturn the call? For B, a failed demand test: structural decline in the product itself, since starvation never pays if demand is dying. For A, verified demand acceleration well above +12%/yr from primary data. That's rare, but the humility clause requires stating it, because capital-cycle logic is a strong prior, not a certainty.

Problem 10 (timed: 10 minutes, direction meets price). A quality platform's market price implies a ~20-year CAP. Your reverse-DCF-lite says, roughly, that steady-state value explains ~60% of the price and the excess-return strip ~40%. On your dashboard, Signal 3 is confirmed (new-cohort share going to a four-question-passing PLG rival), Signal 1 is at watch, others clear. Say you re-underwrite the CAP at ~10 years, roughly halving the excess-return strip. What's the approximate repricing, and what's the general lesson about which valuation lever direction verdicts move?

Solution. Halving the excess-return strip cuts ~40% × ½ = ~20% off the value, holding all else constant. It happens before any change to near-term forecasts, which may still look perfect, because Signal 3 shows up in the flows and the stock P&L lags. The lesson generalizes. Direction verdicts work on the duration lever, the CAP, and the CAP is where premium prices hide their biggest, least-examined bet. Moat existence work tells you whether excess returns are real. Direction work, the dashboard above, tells you how many years of them to pay for. A duration cut hits value geometrically sooner than it hits reported earnings. So if you run the dashboard on flows, you reprice before the market's stock metrics force you to. That asymmetry is the professional payoff of the whole phase.

Problem 11 (guided: the ROIC glide, peak-to-peak). Vitrix Ceramics (synthetic) posts cyclical-peak ROCEs of 28%, then 20%, then 16% across three successive up-cycles; the INR hurdle is 12%. (a) Compute the spread at each peak. (b) Score Signal 4 and say what the pattern implies for the CAP. (c) Name the one false positive to clear before finalizing.

Solution. (a) Spreads = 28−12, 20−12, 16−12 = 16, 8, 4 points, so each peak's spread is half the last. (b) Each cyclical peak lower than the one before is the textbook Signal-4 glide, and it scores 2. Extrapolate the halving and the next peak's spread is ~2 points, the one after ~1. The excess-return CAP is being consumed in real time, the mean-reversion base rate arriving on schedule. (c) Clear the denominator false positive. Confirm the latest, lowest peak isn't depressed by a large capex program parked in CWIP or by freshly commissioned assets still ramping, and recompute ROCE ex-CWIP (and ex-goodwill) before downgrading. That's Signal 4's own check. If the ex-CWIP number is gliding too, the fade is real.

Problem 12 (guided: the supply/demand race sets utilization). A synthetic building-materials industry has installed-plus-announced capacity growing from 200 MT to 320 MT over four years. Demand grows 6%/year from a starting 85% utilization. (a) Compute the capacity CAGR. (b) Find the ending utilization. (c) Work out what capacity would have held utilization flat, and how large the overshoot is. (d) Write the one-line stage call.

Solution. (a) (320 ÷ 200)^(1/4) − 1 = 12.5%/yr. (b) Starting demand = 85% × 200 = 170 MT. After four years it's 170 × 1.06^4 = 214.6 MT, so utilization = 214.6 ÷ 320 = 67.1%, an ~18-point collapse. (c) To hold 85%, capacity should have tracked demand's 6%: 200 × 1.06^4 = 252.5 MT. That means the industry built ~67.5 MT too much, a published glut. (d) "Stage 2 to 4: supply CAGR 12.5% vs demand 6%, utilization falling from 85% to 67%, pricing power about to break." Notice what you never forecast. Demand was fine (+6%), and the damage was entirely on the supply side.

Problem 13 (independent: the four-question test on a live Indian transition). Ten-minute-delivery quick-commerce platforms are routinely called a "disruptor" of India's organized supermarket chains. Run the four-question test and the vulnerability grid, and state what kind of threat this actually is. (Write yours before reading.)

Solution. On Q1, the foothold, quick-commerce mostly entered with a better proposition (speed, convenience) at a higher effective cost per order, aimed at time-poor urban customers. That's a sustaining, head-on attack on convenience, not a worse-but-cheaper foothold. There's a partial new-market angle only for instant top-up occasions the weekly big-box trip never served. On Q2, trajectory, it's improving toward the mainstream basket through assortment and private label, so yes. Q3, economics, is the decisive question. Dark-store unit economics have historically been subsidy-supported, and a structural cost advantage has to be proven per order at maturity before you assume it. Remember the 2015 to 2024 museum of un-closed unit economics. On Q4, asymmetry is weak. An incumbent can list on the same rails and open its own express/dark-store format without cannibalizing a high-margin core, since grocery margins are already thin. So its rational response is co-option, not forced retreat.

Classification: mostly a channel-shift / sustaining battle with a small new-market slice, not textbook low-end disruption. On the vulnerability grid, the incumbent's moat is efficient scale + distribution density, which is architecture-specific, because a warehouse-app world can bypass store-location advantage. So the exposure that matters is the distribution row, not a Christensen dilemma. Direction read: STABLE where the job is "planned weekly stock-up," and a live watch item where the job is "instant top-up." (Check the current dark-store contribution margins before finalizing.)

Problem 14 (independent: the published future, two order books). A vessel segment shows an order book of 72% of the fleet deliverable evenly over three years, scrapping ~1%/year, demand growing ~2%/year (a boom-top reading). Years later a second reading shows an order book of 7% over three years, scrapping ~3%/year, demand ~3%/year (a trough reading). (a) Net annual supply growth in each. (b) Which is the better buy setup, and why. (c) The single input you didn't need.

Solution. (a) At the boom, gross supply growth is (1.72)^(1/3) − 1 = 19.8%/yr, and net of 1% scrapping it's ≈ 18.8%/yr. Against ~2% demand that's a ~17-point-a-year glut already on the water. At the trough, gross is (1.07)^(1/3) − 1 = 2.3%/yr, and net of 3% scrapping it's ≈ −0.7%/yr. So the fleet actually shrinks, and against ~3% demand that's a ~3.7-point-a-year tightening. (b) Buy the trough. Starved supply plus persistent demand is the stage-5-to-6 setup, and you take it through the strongest balance sheet / lowest-cost operator, as Rule 2 says. (c) A precise demand forecast. "About 2 to 3%" was plenty, because the order book did all the work, with the same arithmetic as the shipping example.

Problem 15 (independent: flows vs stocks, in dollars). A US vertical-SaaS category has ARR of $2.0bn growing 15%/year. Its incumbent, Larkspur (synthetic), holds $0.8bn (40% share) but grows only 5%/year, because it's losing new logos to a structurally cheaper product-led rival. (a) Tabulate stock share for Years 0 to 4. (b) Compute the Year-1 flow share (share of the year's growth). (c) When does stock share cross below 30%, and what is the lesson?

Solution. (a) Category: 2.00, 2.30, 2.65, 3.04, 3.50. Larkspur: 0.80, 0.84, 0.88, 0.93, 0.97. So stock share = 40.0%, 36.5%, 33.3%, 30.4%, 27.8%. (b) Year-1 flow = Larkspur's growth ÷ category growth = 0.04 ÷ 0.30 = 13.3%, barely a third of its 40% stock share, so the rival is taking ~87% of new ARR. (c) Stock share falls below 30% only in Year 4, but the flow flagged the loss in Year 1. An installed base votes for the past. Win rates and share-of-new-ARR vote for the future. A moat whose stocks look pristine while its flows crack is eroding regardless.

Problem 16 (timed: 12 minutes, integration). Chatpata Foods (synthetic), a branded-snacks challenger; input costs flat throughout; hurdle 12%.

Score all five signals (0/1/2), compute ROIIC for FY1 to FY4 and FY3 to FY6, name the mechanism, give the direction verdict, and state the reverse-DCF-lite valuation consequence.

Solution. S1 scores 2: gross margin fell 600bps with flat inputs. S2 scores 2, as trade schemes rose from 4% to 9% of sales, with price integrity leaking. S3 scores 2 because share fell 4 points to a structurally different rival (regional factory-direct, ~20% cost edge), verified structural and not a subsidy. S4 scores 2: ROIC fell from 20% to 12%, so the spread has gone to zero at the hurdle. S5 scores 2 too. ROIIC for FY1 to FY4 is (133−120) ÷ (930−600) = 13/330 = 3.9%, and for FY3 to FY6 it's (130−132) ÷ (1,250−800) = −2/450 = −0.4%. Invested capital more than doubled while incremental returns fell to nothing. Your dashboard reads 10/10, with a nameable mechanism: the structural cost bypass of the general-trade channel. Verdict: ERODING, effectively BROKEN, since the spread is at or below the hurdle. Revert to Greenwald's no-barrier rules and value it as operations (CAP ≈ 0 to 2 years).

Now the reverse-DCF-lite. Any price that had embedded a long CAP, say an excess-return strip worth ~40% of value, loses essentially all of it. That's a ~35 to 40% repricing on the duration lever, before headline earnings (still ₹130 cr) visibly break. Any remaining long needs a tripwire: rival share > 15% or ROIC < 10% for a second year means the thesis is dead. Scoring: full verdict + both ROIICs + mechanism inside 12 minutes = fluent. If you scored the signals but missed that S4 crossing the hurdle tips ERODING toward BROKEN, go back over the verdict grammar.

Applied mini-project

The project is the Phase 4 capstone: two moat dossiers, one Indian and one US company from the candidate lists, in the required eight-section structure, each ending in a verdict block with numeric tripwires. Budget ~4 hours per dossier. For data, use the latest annual report or 10-K (skimmed the way you learned early on), one investor presentation, and 10 years of statements from screener.in/Tijori (India) and TIKR/stockanalysis.com + EDGAR (US). Add one share-data source you name, such as an industry body or CRISIL/Nielsen citations in filings and the trade press. Every number in the dossier carries a source, and every estimate is labeled. When you're done, self-grade both against the rubric, compare them with the model dossier (Worked Example 3), and revise once. Then file them in your knowledge system and copy both verdict blocks into the moat section of your Mental-Model & Moat Checklist.

Grading rubric (score each dossier; 2 = fully met, 1 = partial, 0 = absent/wrong):

Pass bar: ≥24/30 per dossier AND no zero on criteria 2, 5, 11, 12 (the critical four). Both dossiers must pass to clear the phase, and this is the master map's M5 milestone. Expect one revision cycle, because it's part of the training. Your gap list from the self-grade is the lesson.

Case Lab: run case XL-4.3 in the app's Cases tab. It's a timed, auto-graded dynamics scenario. You get six years of a synthetic company's numbers plus an industry supply table, and you have to place the cycle stage from the gauges, run the four-question test on the named threat, score all five erosion signals with the computations shown, and commit a direction verdict with numeric tripwires. It rehearses the dossier's dynamics section in miniature and feeds the cycle-stage classifier and erosion-dashboard drills. Clear it before you submit your two capstone dossiers, so the assembly is muscle memory and not a first attempt.

Reading & resources

• Edward Chancellor (ed.), Capital Returns: Investing Through the Capital Cycle (2015): the Introduction (the capital-cycle primer; worth five rereads) plus Part I ("Investment Philosophy") and the commodity/shipping essays. [Paid] [Intermediate] The predecessor volume Capital Account (2004) is optional deepening. [Paid] [Advanced]
• Clayton Christensen, The Innovator's Dilemma (1997): Introduction + chs. 1–2 (disk drives; the failure framework) and chs. 4–6 (upmarket migration, RPV, who wins sustaining vs disruptive fights). [Paid] [Intermediate]
• Christensen & Raynor, The Innovator's Solution (2003): ch. 2 (the fullest minimill/integrated-steel treatment and the "good enough" logic) and ch. 4 (new-market disruption and non-consumption). [Paid] [Intermediate]
• Christensen, Raynor & McDonald, "What Is Disruptive Innovation?", Harvard Business Review, December 2015: the false-positive corrective (Uber, Tesla) from the source. [Free with registration] [Beginner]
• Michael Mauboussin & Dan Callahan, Measuring the Moat (Counterpoint Global, 2024 edition): reread now with dynamic eyes, the industry-analysis, disruption, and threats sections. [Free PDF] [Intermediate]
• Mauboussin & Callahan, The Base Rate Book (2016): ROIC persistence and mean-reversion base rates behind every CAP judgment. [Free PDF] [Advanced]
• Data (India): RBI OBICUS quarterly capacity-utilization survey (rbi.org.in); company investor presentations (capacity/expansion slides); CRISIL/ICRA sector notes via press summaries; Joint Plant Committee (steel), CMA (cement), SIAM (autos); Prime Database issuance summaries in the financial press; screener.in for 10-year capex/D&A. [Free unless noted] [All levels]
• Data (US/global): Baker Hughes weekly rig count; EIA Drilling Productivity Report; Fed G.17 capacity utilization (via FRED); World Steel Association statistics; Clarksons order-book headlines via trade press (TradeWinds, Splash247); the Baltic Dry Index (published daily, free); 10-K Item 2 + investor decks for company capacity. [Free] [All levels]
• Podcasts: Business Breakdowns: search the feed for the episodes on Copart and Old Dominion before drafting a US dossier on either. Acquired: the TSMC and Costco episodes are masterclasses in process power and scale-shared respectively. [Free] [Beginner–Intermediate]
• Live specimens to watch (mid-2026): Indian paints (Birla Opus entry: capital cycle vs a wide moat, in real time); Indian cement's next capacity race; US shale's capital-returns regime; generative AI vs software switching costs (the Adobe dossier question). No book will grade these for you. Your dashboard will. [Free] [Advanced]
• Going deeper: GMO and Marathon quarterly letters (Jeremy Grantham, James Montier; Marathon's Capital Returns essays). Practitioner capital-cycle and bubble-anatomy writing that extends Chancellor with live sector calls. Search the GMO Insights archive for the capacity/returns and "this time is different" pieces, and read them against the capital-cycle section and Mistake 2. [Free] [Advanced]
• Going deeper: Hamilton Helmer & Cid, 7 Powers (2016), the Counter-Positioning chapter. It restates the disruption mechanism from this material as a formal Power: the incumbent that structurally cannot respond without damaging its existing business. It's the cleanest bridge between Helmer's Lens 3 and Christensen's dilemma, and the vocabulary your dossier's disruption audit should borrow. [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": "Building a capital-cycle-plus-erosion-dashboard SCREEN across a coverage list, instead of re-running the six-gauge/five-signal protocol from scratch for one company at a time", "use": "Use an LLM to help assemble the raw inputs (capex, D&A, capacity, demand, gross margin, share, ROIC, invested capital) for a list of companies into one table, then run every company through the same scoring function so the dashboard becomes a repeatable screen rather than a one-off memo.", "tools": ["Claude or ChatGPT to extract capex/D&A/margin/share time series from filings into a structured table", "a notebook with the erosion-dashboard scoring function (the block below)", "RBI OBICUS, CMA/JPC/SIAM sector data (India) or FRED G.17 / Baker Hughes rig count (US) for the capacity/utilization gauges"], "prompt": "Here are six years of capex, D&A, gross margin, market share, ROIC, and invested capital for [company], plus industry capacity and demand growth. Score each of the five erosion signals 0/1/2 per the module's rules, flag the capital-cycle stage from capex/D&A and the capacity-demand gap, and apply the direction-verdict grammar (widening/stable/eroding/broken).", "verify": "Recompute at least the ROIC glide and ROIIC signals by hand from the raw NOPAT and invested-capital figures -- an AI can silently misapply the module's specific false-positive checks (ex-CWIP ROIC, the harvest-trap margin/volume pairing) that turn a raw number into a correctly SCORED signal.", "diy": "The mastery check -- and the capstone dossiers -- require running the full six-gauge cycle read and five-signal dashboard on one company by hand, with no screening tool in the room, and defending the direction verdict and its tripwires unaided. The screen above only extends that same unaided judgment across many companies at once." }

Modern Data Analysis

By hand first. The six cycle gauges and the five erosion signals, each scored 0/1/2 by hand from filed numbers, are unchanged. They're what the mastery check and the capstone dossiers gate.

Today's workflow. A coverage analyst doesn't run the protocol on one company in isolation. Every quarter, the same six gauges and five signals get re-run across a whole watchlist. A professional uses a small, transparent scoring function applied uniformly, so "who deserves a closer look this quarter" becomes a sortable table instead of whichever name is top of mind.

Tools & sources (IN + US). For India, use RBI's OBICUS quarterly capacity-utilization survey (rbi.org.in), capacity data from CMA (cement), the Joint Plant Committee (steel) and SIAM (autos), and screener.in for the ROIC/ROIIC/capex series. For the US, use Fed G.17 industrial capacity utilization via FRED and the Baker Hughes weekly rig count. EDGAR 10-K Item 2 (Properties) and investor decks give you capacity tables, and stockanalysis.com/TIKR covers the financial series.

# recomputed in-session -- R3
def score_erosion(gm_bps, churn_signal, flow_share, roic_glide, roiic_minus_wacc):
    s1 = 2 if gm_bps <= -150 else (1 if gm_bps <= -50 else 0)
    s2 = churn_signal                                   # already scored 0/1/2 upstream
    s3 = 2 if flow_share <= -3 else (1 if flow_share <= -1 else 0)
    s4 = 2 if roic_glide <= -6 else (1 if roic_glide <= -2 else 0)
    s5 = 2 if roiic_minus_wacc <= -4 else (1 if roiic_minus_wacc <= 0 else 0)
    signals = [s1, s2, s3, s4, s5]
    confirmed = sum(1 for s in signals if s == 2)
    total = sum(signals)
    if confirmed >= 2:
        return signals, total, "ERODING" if total < 8 else "ERODING -> BROKEN watch"
    return signals, total, "STABLE" if total <= 1 else "WATCH (1 confirmed signal only)"

def cycle_flood(capex_da, cap_growth, demand_growth):
    return capex_da > 2.0 and (cap_growth - demand_growth) > 2.0

# --- anchor: Worked Example 1's Bharat Ispat capacity/demand CAGR ---
cap_cagr = (146/100)**(1/5) - 1     # -> 7.9%/yr, matches the module's own figure
dem_cagr = (106/88)**(1/5) - 1      # -> 3.8%/yr, matches the module's own figure

# --- extension: run the screen across five synthetic companies ---
companies = {
    "Alpha Cement":   dict(capex_da=2.8, cap_g=11.0, dem_g=5.0, gm_bps=-40,  churn=0, flow=-1.0, glide=-6, roiic_gap=+3),
    "Beta Software":  dict(capex_da=0.9, cap_g=3.0,  dem_g=9.0, gm_bps=-420, churn=2, flow=-9.0, glide=-9, roiic_gap=-6),
    "Gamma FMCG":     dict(capex_da=1.1, cap_g=4.0,  dem_g=4.5, gm_bps=+60,  churn=0, flow=-0.4, glide=+1, roiic_gap=+2),
    "Delta Steel":    dict(capex_da=0.6, cap_g=1.0,  dem_g=4.0, gm_bps=-10,  churn=0, flow=+0.5, glide=+4, roiic_gap=+5),
    "Epsilon Retail": dict(capex_da=1.6, cap_g=9.0,  dem_g=6.0, gm_bps=-150, churn=1, flow=-3.5, glide=-4, roiic_gap=-2),
}
screen = {name: (cycle_flood(c["capex_da"], c["cap_g"], c["dem_g"]),
                  *score_erosion(c["gm_bps"], c["churn"], c["flow"], c["glide"], c["roiic_gap"])[1:])
          for name, c in companies.items()}

Verify. Bharat Ispat's anchor reproduces cap_cagr = 7.9% and dem_cagr = 3.8% to the decimal, matching Worked Example 1. RangSetu Paints' fade table (Worked Example 4) also reproduces when recomputed. PV of the ten visible fading-spread years sums to ₹1,393 cr, and the infinite tail to ₹1,419 cr, against a ₹5,833 cr flat perpetuity. So 75.7% of the naive "moat value" was duration, and the spread crosses the practical CAP line (2 points) at 6.8 years. Both match the figures stated earlier, to the decimal.

Quantitative lens

The screen, run on the scoring rules above. Here are five synthetic companies, with capital-cycle stage and erosion dashboard scored side by side:

Look at two things the table shows that a single company's dashboard can't. Alpha Cement shows barely any erosion in its numbers yet. Its dashboard reads 3/10 with one confirmed signal, one short of the two-signal ERODING threshold. But the capital-cycle gauge flags it. A capex/D&A of 2.8× against a capacity-demand gap of six points a year is a flood already funded and en route. It's largely invisible in this year's margins by construction, because capacity arrives late and together. Epsilon Retail shows the opposite pattern. It has an already-confirmed erosion mechanism (2 signals, right at the threshold) with only a modest cycle reading, so its damage is already in the numbers and not merely inbound. A screen that reported only one axis would miss one of the two companies entirely.

The base-rate caveat, made explicit. Earlier you met the finding that "of firms in the top quintile of ROIC, only about half are still in the top two quintiles a decade later." That's a fact about moat photographs, and it also caps how much confidence any screen output deserves. Even Gamma FMCG and Delta Steel above, flagged "STABLE" with a clean dashboard and no cycle flood, carry the same unconditional base rate as any other currently-strong performer. Roughly half of firms in their position won't hold it for a full decade, for reasons a six-gauge, five-signal screen run on today's filings can't yet see. A screen sharpens where to look harder. It doesn't repeal the base rate.

What the screen can't do. We constructed the five companies and their inputs to illustrate the scoring rules cleanly. None comes from a real filing, so rerun the same functions on real 10-K/annual-report data before you use a verdict in a memo. The screen's thresholds (gross-margin bps cutoffs, the ±3-points-per-year share-flow line, the capex/D&A >2.0× flood trigger) are our own rules of thumb, not statistically fitted cutoffs. They encode judgment calibrated by the worked examples, and a borderline company just under a threshold isn't meaningfully different from one just over it. Most important of all, a screen that scores 0/10 on the erosion dashboard is reporting an absence of the five specific symptoms defined here. That's a long way from a certified absence of risk. Kodak's installed base looked pristine on every stock metric, and the flow metrics that would have caught it weren't the ones anyone was screening for. Screen widely, but verify by hand before betting on any single name's row.

Where this goes next: galaxy cross-links

• G3 · CN4.02 (industry-primers-and-disruption): the consulting-side counterpart to the disruption-classification and industry-dynamics machinery.
• G2 · DA1.02 (pandas-data-wrangling): turns the five-company screen above into a real, sortable multi-company, multi-quarter panel.
• G1 · QM1.01 (probability-distributions-estimation): the base-rate/persistence machinery behind the "half of top-quintile firms fade in a decade" caveat this lens leans on.

Teach it back & journal

Feynman prompt. Explain the capital cycle to a smart 15-year-old using a street of lemonade stands. Start with one hot summer and one stand making great money. Next, describe what every neighbor does, when all the new stands open (and why they all open the same week), what happens to the price of lemonade, who quits, and who is left when the next hot summer comes. Then write the one-page explainer: "Why great profits invite their own destruction, and what a moat actually stops." No jargon is allowed except three terms you must define in passing: supply, barrier to entry, and cost of capital. End with the two questions you'd now always ask before believing any company's profits will last: who is building capacity right now? and is anyone attacking this business with a different architecture?

Journal reflection. Look at your two capstone dossiers and answer in writing. (1) Which erosion signal would you personally be most tempted to rationalize away on a company you loved and owned, and what will you do about that? Your tripwires are the answer, so check that they're numeric and dated and that you've put a reminder in your calendar to check them. (2) Take your model-dossier gap list. Find the rubric criterion you scored lowest on across both dossiers, and decide whether the cause is a knowledge gap (reread that section) or a discipline gap (fix the process). (3) The five weeks of this phase produced a moat checklist you'll carry for decades. Write its single most important line in your own words.

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