Moats II: Greenwald, 7 Powers, Measuring the Moat

Phase 4 · Week 40 · ~20 hours

Where it sits. Fifth module of the strategy phase, and the second of the two moat modules. Your first moat module gave you the first lens, Morningstar/Dorsey's five sources (intangibles, switching costs, network effects, cost advantages, efficient scale), together with the non-negotiable numbers test: no ROIC persistence, no moat, whatever the story says. Three more lenses follow, and the master brief demands all of them on every business. Greenwald (Competition Demystified) collapses Porter's five forces into a single question, are there barriers to entry?, and hands you the cleanest empirical detector in strategy, the market-share-stability test. Helmer (7 Powers) forces every moat claim through a two-part gate: a Benefit to the owner and a Barrier that explains why competitors don't just copy it. He also adds the time dimension of when each power can be born. Mauboussin (Measuring the Moat) converts the whole conversation into the number a valuation actually needs, the Competitive Advantage Period (CAP): how many years the ROIC−WACC spread survives. That last lens closes a loop you have been circling since the returns-on-capital module (which promised "you will estimate CAPs" here) and the terminal-value module, whose reverse DCF found that premium prices are usually bets on duration, not next year's growth. Auditing that claim about moats is the work of the next twenty hours. Your next module adds the forces that erode moats, the capital cycle and disruption, and then the Phase 4 capstone makes you prove two moats end-to-end. By Phase 8, the four lenses compress into teardown question 6, answered in about ten minutes.

Suggested hour budget. Core content: 8h · worked examples: 2h · practice set: 4h · applied mini-project (the Cross-Lens Moat Dossier): 4h · flashcards + mastery check: 2h.

App link. The week feeds the Industry & Business Model branch: the cross-lens exam deck (the gate here, ≥85%: one business described, four verdicts demanded), the share-stability calculator drill (share tables in, barriers verdict out), the power-mapper (company facts in, Benefit + Barrier + which-of-seven out, with "not a power" as the most common right answer), and the CAP sandbox (price in, implied CAP out, wired to the reverse-DCF engine).

Learning objectives

By the end of the week you can:

1. Run Greenwald's collapse: reduce any five-forces map to the single decisive question, do barriers to entry exist here?, then sort every claimed advantage into his three real categories (supply/cost, demand/customer captivity, economies of scale), rejecting everything else as strategy theatre.
2. Explain and detect the strongest combination, economies of scale interlocked with customer captivity, and why local or niche dominance (Walmart's regional density, DMart's clusters, Ghari's Uttar Pradesh) beats thinly spread global scale.
3. Execute the market-share-stability test on any industry from public data: build the share table, compute average absolute share movement per year, apply the ~2-points-per-year threshold, cross-check against incumbent ROIC, and state the verdict with its limits.
4. Prescribe correctly for no-barrier markets: explain why, on a level playing field, operational efficiency is the only strategy and cost of capital the only durable expectation, and what that implies for the multiple you may pay.
5. Apply Helmer's Benefit + Barrier gate to all 7 Powers (scale economies, network economies, counter-positioning, switching costs, branding, cornered resource, process power): for each, name the benefit, name the barrier, and demonstrate why "me too won't do", including the incumbent's-dilemma arithmetic behind counter-positioning.
6. Place powers on the clock: use Helmer's power progression (origination → takeoff → stability) to say which powers a business could possess at its stage, and spot claims that violate the timing (a two-year-old startup claiming brand power; a mature laggard claiming counter-positioning).
7. Run Mauboussin's Measuring-the-Moat workflow (industry map → profitability/profit-pool analysis → five forces → sources of added value) as a repeatable audit that ends in a written verdict.
8. Estimate a CAP and read the market's implied CAP: value a franchise as steady state + excess-return years (the Miller–Modigliani decomposition), solve backwards from price to the CAP the market is granting, judge it against mean-reversion base rates, and connect the answer to M3.05's three-lever diagnosis.

Prerequisites & connections

Builds on. M2.03 gave you ROIC and ROCE built by hand, the ROIC−WACC spread as the economic verdict, and the persistence base rates from Mauboussin's transition matrices. Mean reversion is the default fate there, and roughly half of top-quintile firms are still top-two a decade later. That work ended with a promise: "when you meet a durable spread, your job is to name the barrier." The naming machinery starts here. M3.03 supplied the hurdle rates you'll test spreads against (~12% post-tax INR, ~8 to 9% USD, as of mid-2026, check against your own WACC builds). M3.05 supplied the reverse DCF and its three-lever diagnosis (growth, duration, required return), and lens 4 here is the duration lever made rigorous. M4.03 covered Porter's five forces and profit pools, and Greenwald is a ruthless simplification of it. You need to know the full map before you're allowed the shortcut. M4.04 covered the five sources and the numbers test, which is lens 1. Everything there is assumed. Nothing here re-argues whether ROIC persistence is required. Here the argument is why it happens and how long it lasts.

Feeds forward. In M4.06 the capital cycle and disruption arrive as the moat-erosion forces. M4.06 also teaches you to watch your share-stability tests and CAP estimates decay. The Phase 4 capstone asks for two moat dossiers graded against the rubric, one on an Indian company and one on a US company, with all four lenses run and reconciled. Phase 5's sector playbooks apply these lenses industry by industry. Phase 8's teardown question 6 ("is there a moat?") is this week under a stopwatch. Day 2 of the deep dive (share stability, Mauboussin's map) repeats this week's share-stability test and Mauboussin audit word for word. Competency C4 in the master map, "identify moat existence, source, width, and direction, and prove it in the numbers", is certified on what you learn this week.

In short. One question decides everything. Can a newcomer get in and force returns to the cost of capital? Three lenses answer it from different angles. Greenwald names the barriers, Helmer stress-tests each claim as Benefit-plus-Barrier with a birth certificate, and Mauboussin converts the survivors into a number of years. That number of years, the CAP, is where moat analysis stops being literature and becomes a valuation input.

Core teaching content

4.1 Why you need three more lenses when you already have one

You left M4.04 with a working detector. Name the source (one of five), confirm it in ROIC persistence, and reject the false moats. So why not stop there? Because the five-sources taxonomy, used alone, fails in three ways that cost real money:

1. It classifies but doesn't interrogate. "Brand" is on the list, so analysts write "moat: brand" for every household name, including the ones whose brand produces zero pricing power (you met them in the false-moat drill). You need a gate that demands mechanism. What does the advantage do to cash flow, and what stops imitation? That gate is Helmer's Benefit + Barrier.
2. It's static. The taxonomy says nothing about when advantages form. So analysts pay moat prices for startups that can't have moats yet, and assume mature businesses can conjure new powers at will. Helmer's power progression adds the clock.
3. It isn't a number. A valuation can't discount "wide moat." It can discount fifteen years of ROIC above WACC. Until the moat becomes a duration (Mauboussin's CAP), your Phase 3 machinery and your Phase 4 judgment don't connect. You're left saying "great company, expensive stock" with no way to say how expensive is too expensive. The CAP is the adapter plug.

Underneath all three is Greenwald's method, and it solves a fourth problem. Five-forces analysis sprawls. Run the full M4.03 five-forces exercise properly and you'll produce two pages of forces, each "medium-to-high," and no verdict. Greenwald's collapse turns the sprawl into a decision. Only entry matters, and the rest is commentary.

One habit before we start. These are lenses, and nobody's asking you to join a sect. The four frameworks agree far more than their marketing suggests. Helmer's scale economies ≈ Greenwald's supply-plus-scale ≈ Dorsey's cost advantage, and Helmer's switching costs ≈ Greenwald's captivity ≈ Dorsey's switching costs. You run all four because each one fails differently. Greenwald catches industries pretending to be attractive and Helmer catches advantages pretending to be barriers. Mauboussin catches durations pretending to be forever. When all four agree, you may finally believe the numbers you saw in M2.03.

4.4 Local dominance beats global spread: the geometry of scale

The subtlest and most profitable idea in Competition Demystified is that scale is measured against the relevant market, and the relevant market is usually smaller than the map suggests. Most fixed costs are local or product-niche costs. A distribution network covers a region, and an ad campaign covers a language and a culture. A sales force covers a territory, and an R&D programme covers a product category. So a dominant share of a small market beats a minor share of a huge one. Whenever the cost structure is heavy with fixed costs, 60% of Tamil Nadu beats 6% of India.

Walmart, the canonical case (US). The lazy story says Walmart won by being the biggest retailer on earth. Greenwald's numbers say the opposite. Walmart won when it was small, a regional discounter saturating contiguous small-town markets in Arkansas, Oklahoma, Missouri, because within that region it was overwhelmingly dominant. Here are the mechanics, in illustrative magnitudes from the book's mid-1980s analysis (read the chapter). Stores clustered within a day's truck run of a company distribution centre pushed inbound logistics ~2 to 3% of sales below rivals whose stores were scattered. Advertising bought city by city covered many Walmart stores per rival store, cutting ad cost per sales-rupee by more than half. District managers supervised tightly packed stores, which thinned supervision overhead. Stack the layers and Walmart ran at a ~4 to 5 percentage-point operating-cost advantage against Kmart, even though Kmart was the national giant with more stores in total.

Sam Walton's genius was density. He expanded like spilled paint, one adjacent county at a time, and never leapfrogged into territory where the distribution math didn't work yet. Then comes the proof by contradiction. When Walmart later spread into distant, contested metros and abroad (Germany and Korea, both exited), the density advantage stayed home and returns in those ventures sagged toward ordinary. Same company, same systems, same culture. Different local share, different economics. Walmart's moat was its share of each locality, never "Walmart" as such.

The Indian rhymes. You met one when you dissected business models. DMart expands in deliberate clusters (Maharashtra, Gujarat, Telangana first) so that stores share distribution centres and brand word-of-mouth. That's density before geography, Walton's spilled paint with an Ahmedabad accent. Sharper still is Ghari detergent (RSPL), a Kanpur company that became India's largest-selling detergent brand by share of volume while barely advertising nationally (illustrative: high-teens national volume share, ~as of FY24, check against industry reports). It saturated Uttar Pradesh and the Hindi belt with dense distribution, local-freight economics, and price positioning. Within its territory it out-earned multinationals long before it ventured beyond it. HUL had, and has, more total scale. Ghari had more scale where Ghari fought.

This geometry turns up elsewhere too. Regional cement leaders have it, because freight is ~20 to 25% of cement cost and every plant's moat is a 300-km circle. Regional dairies have it too. Milk sours, so collection density decides cost. City-gas distributors run pipe networks that are local natural monopolies. And South-based banks with dense branches in two states beat a national also-ran's thin spread.

Three instructions fall out of this for you as an analyst. First, define the market before you measure share. A share-stability test run on "global retail" is meaningless when the fixed costs are regional, so run it on the region or niche where the fixed costs actually are. Second, respect niche dominance in valuations. A boring company with 65% of a ₹2,000-crore niche (bearings for railway wagons, dental chairs, exam-prep for one state's civil-service test) may own a sturdier moat than a celebrated firm with 8% of a giant market. Third, treat growth out of the niche as moat-dilutive until proven otherwise. When the regional champion announces national ambitions, it's leaving the territory where its costs are lowest and its customers are captive. Greenwald's framework predicts margin compression. Applause from the market for the "bigger opportunity" is often the sound of a moat being traded for a story, Your returns work made this point with Asian Paints and Birla Opus, and the capital-cycle lens will develop it.

4.5 The market-share-stability test: strategy's cleanest empirical instrument

Everything so far is theory you could argue either way over dinner. Greenwald's escape from dinner-table strategy is something you can observe. If barriers exist, entrants fail and incumbents hold their customers, so market shares barely move. If barriers are absent, share is bought and lost continuously. You don't have to deduce the moat. Industries have been running this experiment for you for decades, and publishing the results.

The procedure (this is Day 2, step 1 of the Phase 8 deep dive):

1. Define the relevant market, meaning the arena where the fixed costs and the customers actually are.
2. Build the share table: the top 4 to 6 players, with share at 3 to 5 dates spanning 8 to 15 years. Company investor presentations and DRHPs are a good start, since they quote CRISIL/ICRA/Frost & Sullivan studies. Industry bodies help too (SIAM/FADA for autos, TRAI for telecom, IBEF, AMFI for mutual funds, IRDAI for insurance), and so does the trade press. For the US, use 10-K competition sections, trade associations, and the market-share exhibits in initiating-coverage notes. Use revenue share where you can get it. Where you can't, use volume or subscriber share, and say which you used, because they diverge when price points differ.
3. Compute the instability metric: for each firm, take the absolute change in share in percentage points, divide by the years elapsed, then average across firms. Also record the largest cumulative gain or loss and any entrant that reached ≥5% share during the window.
4. Apply the threshold: average absolute movement above ~2 percentage points per firm per year signals weak-to-no barriers, and movement under ~0.5 points/year signals strong barriers. Anything between is a judgment zone that needs the second check. That ~2%/yr line is Greenwald's rule of thumb and nothing more, so calibrate it down for slow-moving industries and up for young ones.
5. Cross-check against incumbent profitability: barriers must show up as returns. Stable shares + sector ROIC persistently above the hurdle = barriers confirmed. Stable shares + cost-of-capital returns = a stalemate, which is no moat. Think of oligopolies that compete away the pool. Airlines before consolidation did it, and so did Indian telecom's three-player "stability" at low ROCE. Moving shares + high returns for one firm = a superior competitor conquering a weak-barrier market, as early Walmart did. In that case you reward the firm and give the industry no credit.

Run 1: Indian organised decorative paints (the stable specimen). These shares of the organised decorative market are illustrative, reconstructed from industry reports and company presentations at ~FY15 to FY24 scale. Build the real one from the CRISIL citations in the Asian Paints and Berger annual reports:

Average absolute movement ≈ 0.25 points per firm per year, an order of magnitude inside the threshold. Over nine years there's one meaningful riser, Indigo, at +2 points in a decade, and it had to build a beachhead in small-town putty to get that far. Meanwhile the leader gained share while earning ~25 to 30%+ ROCE throughout. That makes the verdict powerful barriers, and you can name them in Greenwald's vocabulary. Economies of scale in distribution (Asian Paints' ~1.6 lakh dealer touchpoints and four-times-a-day supply logistics; illustrative) interlock with dealer and painter captivity. That captivity comes from habit and credit relationships, plus the tinting machine on the counter, a switching cost bolted to every shop.

Then there's the live experiment. Birla Opus entered in 2024 with a committed ~₹10,000 crore, a deliberate attempt on the scale of the capital cycle to buy through these barriers. Margins across the industry compressed within a year. Be clear about what this test is. It's a backward-looking instrument that proves barriers existed, and whether they hold against a determined giant is a question for the erosion lens that comes next. Watch the FY25 to FY28 share tables the way a cardiologist watches an ECG.

Run 2: US carbonated soft drinks (the fifty-year specimen). These are corporate shares of US CSD retail volume. They're illustrative, modelled on Beverage Digest-style tallies, so pull the current ones:

Average ≈ 0.3 points per firm per year, and the pattern has held in kind since the 1970s, through diet waves, cola wars, private-label assaults, and health scares. No entrant has reached 5% in half a century. Virgin Cola made the last serious try in the 1990s, and it died at well under 1% despite Branson's billions and fame. You couldn't ask for a cleaner controlled experiment. Money and marketing went up against captivity and bottling density, and captivity won. As for incumbent economics, Coca-Cola's ROIC has run in the high teens to 20s for decades (illustrative). These barriers are as strong as barriers get. Habit-grade demand captivity is interlocked with scale economies in bottling, distribution, and advertising. It's what the king combination looks like when it's fifty years old.

There's also a counter-specimen you'll compute yourself. Worked Example 1 runs the full metric on Indian mobile telecom from FY16 to FY25, where one entrant took ~40 points of share in nine years. It shows you what no barriers looks like in the same arithmetic, and why ₹5+ lakh crore of cumulative industry capital (illustrative) earned sub-hurdle returns while shares were in motion.

Limits of the instrument (memorise these, because the mastery check tests them). (1) It looks backward. It certifies that barriers existed in the past, so pair it with the erosion signals still to come before you extrapolate. (2) Stability without profitability is a stalemate and no moat. Always run step 5. (3) Market definition can rig the verdict in either direction. A share table of "premium urban decorative paint" would have hidden Indigo's rise. State your definition and your source. (4) Stability that a regulator grants through licences or quotas is real but leased, so reprice the moat every time the regulator clears its throat. (5) Young industries fail the test by nature, because everyone's share is moving. There its message is "no proven moat yet, price it accordingly," and that's Helmer's origination stage seen from outside.

4.6 No-barrier markets: the level playing field and its one commandment

Most markets have no meaningful barriers. Greenwald guesses the large majority do without them, and the ROIC distributions you built when benchmarking agree. So what then? Competition Demystified is at its most bracing here. Porter's generic-strategies triangle encourages the consulting reflex to go and find a differentiation strategy. Greenwald says that's the wrong answer. On a level playing field, differentiation buys you nothing durable. Whatever you differentiate, an equal-cost competitor can copy, and the returns on your cleverness get competed away at the speed of imitation. Here's what the book leaves a manager of a no-barrier business. It's in our own words, not Greenwald's:

(our compression of Greenwald's argument, not a quotation) There is no strategy here, only operations. Run faster, waste less, execute better, and expect at best the cost of capital plus a temporary bonus for being the best operator in the room.

It sounds like despair. It's actually liberation, in three ways.

For the manager, it ends the strategy-offsite charade. In a barrier-free market, every hour spent on "repositioning" is stolen from procurement, logistics, and yield, the only games that pay. Exemplary no-barrier operators are heroes of efficiency. Think of the best EPC firms, auto-component machinists, or textile exporters you met in Phase 2. Their 2 to 4 point ROCE edge over peers is pure operational virtue. They re-earn it every year, and they know it.

For the analyst, it sets the multiple. A no-barrier business, however well run, deserves a valuation built on returns fading to the cost of capital. In reverse-DCF terms that's a CAP of roughly zero: steady-state value only, EV ≈ NOPAT/WACC (~8 to 10× NOPAT at Indian WACCs, ~11 to 13× at US WACCs, as of mid-2026). Anything above that multiple is a bet that operational excellence is a moat, and you already know it isn't one.

For the investor, it defines a hunting style. No-barrier industries are cyclical-value territory and never compounder territory. You buy the best operator below replacement cost in the trough and sell when the capital cycle turns. Both the capital-cycle and commodities modules formalise this. Avoid the classic error of paying a captivity multiple for an efficiency story. Indian markets love an "execution premium," and they serve this error daily at 40× earnings.

One refinement before we leave Greenwald. Where barriers exist and several incumbents share the wall, the game among them, pricing restraint against price war, is strategic in the full sense. Greenwald spends the back half of his book on it. There are entry and deterrence games and prisoner's dilemmas, the Coke-Pepsi détente that ended the 1980s cola wars and rebuilt both firms' margins, and the Fox-versus-networks entry study. Hold on to his headline. Inside the wall, cooperating on price and competing on cost is the equilibrium that preserves the pool, and price wars among the protected are self-inflicted moat drainage. If you see a protected oligopoly start a price war (Indian telecom 2016 to 2019, US airlines repeatedly before 2010), mark the profit pool down immediately, whatever the share table said last year. The capital-cycle lens picks this thread up next.

4.9 The clock: origination, takeoff, stability: when each power can be born

The second half of 7 Powers ("Dynamics") answers the question the taxonomy alone can't: when does power arise? Helmer divides a business's life into three stages by the growth of its market. Origination comes before explosive growth and is the invention window. Takeoff is explosive growth, when share is cheap because new customers vastly outnumber contested ones. Stability is maturity, when share is expensive because every gain is a rival's loss. Each power has a birth window, and outside it the power essentially can't be established:

• Origination-stage powers: counter-positioning and cornered resource. Both are decisions or endowments made before scale exists. You adopt the heretical model, or secure the patent/licence/team, at the start. A mature laggard "pivoting to counter-position" is violating the definition, because its own legacy P&L is the target. A mature firm can still acquire a cornered resource, but only at auction prices that arbitrage the rent away, so it fails the acquired-below-value question.
• Takeoff-stage powers: scale economies, network economies, switching costs. All three are races run while share is cheap. During takeoff, uncaptured customers flood in and acquiring them costs little. Whoever banks the volume, the network, or the installed base first owns the barrier when the music stops. Rational takeoff-stage losses exist for this reason. In unit-economics terms you spend CAC now and harvest captivity later, and Jio's free year was a takeoff-stage share purchase of historic scale. The same spending in a stable market destroys value, because you're buying share at stability prices from incumbents who fight back. Birla Opus is running that experiment in paints right now, and the capital-cycle lens comes back to it.
• Stability-stage powers: branding and process power. Both are made of time, or hysteresis: years of consistent signalling, decades of process evolution. They cement in maturity, though their first bricks were laid much earlier.

Three reflexes fall out of this, and the mastery check tests all three. First, date every power claim. A two-year-old startup claiming brand power or process power is claiming to have compressed time, so reject it. What it can legitimately claim is counter-positioning, a cornered resource, or a takeoff race it's currently winning. That last one is a claim about the future, so price it as "no proven moat yet."

Second, respect the window. A firm that missed the takeoff land-grab can't cheaply build scale or network power at stability. When a mature also-ran's strategy deck promises just that, the deck is fiction with a budget. Third, demand the origination story. "Power arrives on the wings of invention," so for every claimed power you should be able to say what was invented (Asian Paints' dealer-direct model, Zerodha's flat fee, Netflix's subscription) and roughly when the barrier formed. A power with no birth certificate is usually operational excellence wearing a costume.

4.10 Four power stacks: reading real companies in Helmer's grammar

Real moats are rarely one power. They're stacks, and what's in a stack tells you how durable it is. (The CAP, coming up shortly, turns that judgment into years.) Here are four studies, with all figures illustrative at ~FY24/FY25 scale. Verify them against filings.

Microsoft (US: the fortress stack). Start with network economies. Office file formats and Teams mean every user makes the next one more captive. Then come switching costs (enterprise workflows, data, admin skills), both procedural and financial. Then scale economies, with Windows/Office/Azure development spread over a billion-plus seats and Azure's capex spread over the second-largest cloud base. Its origination dates check out. Microsoft's 1980 to 81 IBM DOS contract was close to a cornered resource moment, and the 1990s PC takeoff converted it into the scale/network/switching triad, a textbook sequence on Helmer's clock. That stack is why ROIC has stayed far above WACC for decades, and why the CAP the market grants Microsoft is among the longest on earth. Keep an eye on platform shifts (cloud, AI). Each one re-opens the takeoff window, so the stack has to be re-earned there. So far it has been.

Netflix (US: the relay-race stack). Counter-positioning against Blockbuster (subscription, no late fees) bought the time. At incumbent capitulation (~2010, Blockbuster's bankruptcy) that power expired on schedule, and the baton passed to scale economies in originals, the per-subscriber arithmetic you saw earlier. That was a deliberate second invention. Weak-to-moderate switching costs (profiles, taste data) help. Powers are mortal and stacks are relay races, so your question is always "what's the next leg, and is it being built while the current one still pays?"

Asian Paints (India: the interlocked stack). It has scale economies in distribution density, and switching costs in the dealer ecosystem (tinting machines, credit, four-times-daily replenishment). Add process power in the fifty-year forecasting-and-logistics engine, plus modest branding, a real but junior premium. Notice that this is Greenwald's king combination restated power by power: scale interlocked with captivity, plus a time-hysteresis process no entrant can buy. Birla's ₹10,000 crore has to defeat the whole stack. For FY25 to 28, the Helmer question is which layer the attack can actually breach. It can buy dealers and machines. It can't buy fifty years of demand data or painter habit. Still, the share table, that ECG from earlier, gets to referee.

IRCTC (India: the single-power stack). There's one power here, a cornered resource: the exclusive right to online rail ticketing, catering, and packaged water for Indian Railways. It passes four of Helmer's five tests brilliantly and fails none. But the fiat that grants it can reprice it. In October 2021 the government proposed taking half of IRCTC's convenience-fee revenue, and the stock fell roughly a quarter in a day before the order was reversed (check the episode; the lesson is exact). A single leased power is a moat with a landlord. Grade such businesses the way Greenwald's warning on regulation said to, as a real advantage repriceable at the owner's pleasure, and cap the CAP accordingly. Microsoft's four-power redundancy is the contrast. Stacks diversify moat risk the way portfolios diversify stock risk.

4.14 Synthesis: the four-lens protocol, end-to-end on one company

Here's the complete protocol the two moat modules have been building. It's the sequence you'll use for the Phase 4 capstone dossiers and for Phase 5 sector work. Compressed to ten minutes, it's also Phase 8's teardown question 6:

Step 0 · Numbers first: 10-year ROIC/ROCE vs the hurdle. No persistent spread, no moat conversation. Value it as a no-barrier business. Step 1 · Lens 1, the five sources: name the candidate source or sources, and reject false moats (hot product, share without returns, execution). Step 2 · Lens 2, Greenwald: define the relevant market and build the share table. Run the stability metric and the asymmetry test to reach a barriers verdict. If there are no barriers, stop there: operations only, CAP ≈ 0 to 3. Step 3 · Lens 3, Helmer: push each surviving claim through Benefit + Barrier and date each power's origination. Assemble the stack and mark the "not a power" rejects. Step 4 · Lens 4, Mauboussin: run the four-pass audit for anything missed (profit pool, government, disruption). Estimate your evidence-based CAP with a fade, invert the market price for the implied CAP, and write the verdict sentence and the falsifiers.

The protocol run: Asian Paints. All figures are illustrative, modelled on ~FY24/FY25 scale, so pull the actual filings, screener.in ten-year table, and the CRISIL citations in the annual report.

Step 0 · numbers. ROCE has run at ~25 to 35% for over two decades and never come near the 12% hurdle, even in years of paint-input inflation. FY22's crude-derivative shock compressed margins and left the franchise alone. Its spread is wide and persistent, so something real needs explaining. There's a wrinkle in FY25. Margins and growth compressed after Birla Opus's entry, the first sustained counter-evidence in the dataset.

Step 1 · five sources. Three candidates come up: cost advantage from distribution scale, switching costs in the dealer and painter ecosystem, and intangibles. Its brand is real but junior, because so much of the paint choice runs through painters and dealers. Two sources get rejected. Network effects go, because a dealer network is scale and dealer N+1 doesn't benefit from dealer N. Efficient scale goes too, because the market is huge and growing.

Step 2 · Greenwald. Define the market as Indian organised decorative paints. Its fixed costs (depots, tinting machines and brand) are national within the organised segment, and the unorganised segment is a separate price tier. From the share table earlier, average movement ≈ 0.25 pts/firm/yr over ~a decade. In that time the leader gained share, and one firm rose 2 points in ten years. On asymmetry, an entrant must replicate ~1.6 lakh dealer touchpoints, four-a-day replenishment logistics, and painter habit while carrying a fraction of the volume over the same fixed costs. For attackers, the flywheel spins backwards. Verdict: barriers of the king-combination type, scale interlocked with captivity, with a live, capitalised challenge running since 2024 that tests this claim directly.

Step 3 · Helmer. Take the Asian Paints stack from the four studies and gate each layer. Scale economies give the lowest distribution and advertising cost per litre, and the barrier is that share is too expensive to buy. Birla is paying that tariff right now to test it. Switching costs give follow-on sales through captive dealers. Here the barrier is that an attacker must buy out tinting machines, credit lines, and painter familiarity shop by shop. Process power gives industry-low inventory days and stockout rates from the fifty-year forecasting engine. Its barrier is hysteresis, since data and routines can't be bought at any price, and it's the layer no entrant's budget line can address. Branding is the junior power, a modest premium that's real mostly as uncertainty reduction at the consumer edge. Origination stories all date cleanly (the dealer-direct model from the 1940s to 60s, computerised forecasting from the 1970s), so there are no timing violations. Stack verdict: three interlocked "cannot" barriers plus one junior power, with no dependence on fiat and no expiring counter-position.

Step 4 · Mauboussin. In Pass 1, the profit pool shows the leader capturing a large majority of industry economic profit, with shares frozen until 2024. In Pass 2, industry ROIC is comfortably above WACC for the top three, so the profit is structural and doesn't belong to a single firm. Pass 3 covers the forces. Entry is the live force, with Birla Opus committing ~₹10,000 crore and JSW and others probing. Government is neutral. A disruption scan finds no counter-positioned model. D2C paint fails on painter mediation, because the painter is the channel. Pass 4 is the sources you gated above.

Now the CAP. Pre-2024 evidence supported 15 to 20 years. A determined, deep-pocketed entrant attacking the scale layer argues for a haircut and a faster fade, so suppose your dossier lands at ~12 to 15 years with linear fade, updated every year against the share ECG. For the implied CAP, take an illustrative ~50 to 55× earnings, where the market has often priced this franchise. At historical spreads the skeleton returns an implied CAP far above 20 years. In plain terms, the price asserts that the Birla attack largely fails and that the spread survives beyond what even the friendly evidence grants.

Here's the protocol output, stated the only defensible way: moat: real, wide, of the most durable known type; direction: contested for the first time in decades; price: granting more duration than the contested evidence supports; falsifiers: FY25 to 28 share table (±1 pt/yr threshold), incremental margins of both combatants, dealer/tinting-machine counts. That sentence isn't investment advice. It's what a finished moat analysis looks like. Every clause can be audited, every disagreement can be located, and the whole thing is wired straight into your reverse-DCF engine. Next comes what erosion looks like while it's happening, and then the capstone makes you produce two of these sentences yourself.

Common mistakes & how experts think differently

1. Treating capital intensity as a barrier. "It costs ₹5,000 crore to build" stops nobody who has ₹5,000 crore, and capital markets fund entrants by the lakh-crore (telecom, airlines, cement). An expert runs the asymmetry test instead. A barrier exists only when the incumbent got something the entrant's money can't buy at the same price: captive customers, resources granted instead of auctioned, fifty years of demand data. Expensive-to-enter and hard-to-enter are different claims, and only the second one pays.

2. Grading benefits and forgetting barriers. Nine of ten "moat" slides list benefits (better product, lower cost, beloved brand) and skip the only question that matters: why won't a competent rival have this in three years? Experts spend 80% of their time in Helmer's second column, and their commonest verdict is "not a power." A benefit without a barrier is a gift to customers and future competitors, and the analyst ends up paying for the wrapping.

3. Calling every lock-in "network effects." A dealer network, a branch network, a delivery fleet: these are all scale economies, which spread cost. Network economies are different, because there the value per user rises with the number of users. So test it. Does customer N+1 get a better product because customer N exists? If not, stop saying network. Getting the label right matters because the two powers cap out differently. Networks tip winner-take-most within a boundary. Scale only tilts costs, and a rival can out-scale it locally.

4. Ignoring the clock. Paying brand-power prices for a three-year-old D2C label is a timing violation. So is crediting process power to a five-year-old startup, or believing a mature laggard's plan to "build the ecosystem." An expert dates every power's birth. Origination powers come early, takeoff powers arrive during the land grab, and hysteresis powers take decades. No birth certificate, no power.

5. Reading stable shares as a moat without the profitability cross-check. Frozen share tables at cost-of-capital returns describe a stalemate. Picture undifferentiated oligopolists in a trench, where nobody owns a franchise behind a wall. Airlines before consolidation looked like this, as did many stretches of two-wheeler and cement history. Greenwald's step 5 isn't optional. Barriers must show up in both the share table and the ROIC series, or they aren't barriers.

6. Extrapolating the share table forward without an erosion scan. A stability test is a rear-view instrument. It proves barriers existed. Paints looked impregnable in FY23, and ₹10,000 crore arrived in FY24. Experts pair every stability verdict with the live entry map, asking who has announced capacity and who is counter-positioned. They also pre-commit to the falsifying evidence: share-movement thresholds and incremental margins. The erosion lens turns this into a system.

7. Treating the CAP as decoration instead of the deliverable. Analysts write ten pages of moat prose, then value the company on next year's EPS and a peer multiple and never connect the moat to duration. An expert's habit runs the other way. Every moat opinion has to compress into N years with a fade, and every price has to be inverted into the market's N, because only then can the two disagree in a falsifiable way. "Wide moat, 40× earnings, seems fine" isn't analysis. "The price implies 25 years, my gated evidence supports 12" is.

8. Letting the multiple smuggle in the duration. Paying 16× NOPAT "because quality" grants a ~20-year CAP (look back at the CAP table), whether you said so or not. Experts read multiples as CAPs in disguise. They drag the implied duration into daylight and then judge it against base rates, under which 20-year survivors are rare and 30-year survivors are a handful of habit-plus-scale franchises per generation.

9. Moat worship: forgetting that a moat is necessary but not sufficient. A certified moat with no reinvestment runway adds little value, because the I in the M&M formula is zero and the spread protects existing profits only. And a certified moat at an implied-CAP price beyond the evidence is a bad investment in a great business. That was the central lesson of your reverse-DCF work, and now it comes with machinery. An expert keeps three dials separate: spread, duration, and price paid for both.

10. Running one lens and stopping. Each lens has a known blind spot. Five-sources classifies without interrogating. Greenwald's industry verdict can miss a single counter-positioned attacker, since the shares were stable until Netflix. Helmer gates claims but doesn't size profit pools, and Mauboussin's checklist can drown a beginner in process. Experts run all four because the failure modes don't overlap. Agreement across lenses is the only moat evidence strong enough to carry a 15-year CAP into a model.

Worked examples

Worked Example 1: The share-stability metric on Indian mobile telecom, FY16 → FY25 (the no-barrier specimen)

This subscriber-share table is illustrative, built to TRAI-report scale. Reconstruct the real one from TRAI's monthly subscription reports, free at trai.gov.in. Vodafone and Idea are shown combined throughout, so the figures compare across their 2018 merger.

Step 1 is the market definition: Indian wireless subscribers, national, because spectrum, networks, and brand are national fixed costs. Step 2 is the table above, with three dates over nine years. Step 3 is the metric. Total absolute movement = 9 + 18 + 41 + 1 + 31 = 100 points across 5 lines = 20 points per line, and over 9 years that's ≈ 2.2 points per firm per year.

Step 4 applies the threshold. You're above the ~2/yr line, and the secondary flags scream louder than the average. An entrant reached 41%, when anything ≥5% is a red flag, so 41% is a massacre. An entire cohort of incumbents holding 31 points of share was destroyed, with two bankruptcies (RCom, Aircel), multiple exits, and one survivor on state-assisted life support. Step 5 is the profitability cross-check. Cumulative industry capex plus spectrum outlays ran on the order of ₹5 to 6+ lakh crore (illustrative), against sector ROCE below the hurdle throughout the window. Then the AGR judgment turned the deficit into a balance-sheet crisis.

Verdict: no barriers to entry during the window, despite the highest entry ticket in Indian corporate history. Do the asymmetry autopsy. Spectrum was auctioned at full price, so the cornered-resource test fails, because the rent went to the seller. Anyone can buy network equipment from vendors. And subscribers proved non-captive the moment MNP plus a free year removed the switching friction. Capital intensity was a cost of playing. Incumbents could do nothing that Jio's ₹1.5+ lakh crore couldn't buy.

There's a postscript, and it keeps the example straight. After consolidation to three private players plus BSNL, tariffs rose in coordinated steps (2019, 2021, 2024) and the leaders' returns began climbing toward respectability. Is a moat forming? Run the framework and ignore the vibes. Shares are now moving slowly, with Airtel and Jio grinding up ~1 pt/yr and Vi bleeding. But this "stability" is two years old, and profitability is one tariff-war decision away from relapse. Any barrier that exists is a truce among the protected, the oligopoly game Greenwald describes, plus rising spectrum-and-capex scale. Customer captivity has nothing to do with it. Greenwald's framework says to watch the FY25 to FY30 table and the ROCE series before you grant a CAP above the single digits. The capital-cycle lens makes the same point from the supply side.

Worked Example 2: Implied CAP on Coca-Cola: the skeleton, its distortion, and the fix

All inputs are illustrative, at ~FY24 scale. Pull the 10-K and rebuild. What you're learning is the method and the sensitivity, so don't fuss over the decimals.

Your inputs are NOPAT ≈ $10bn, WACC ≈ 8%, incremental ROIC ≈ 20% and reinvestment ≈ 30% of NOPAT, which makes I ≈ $3bn/yr. Enterprise value is ≈ $290bn.

1. Steady-state value = 10 / 0.08 = $125bn, the no-improvement Coca-Cola, 43% of the price.
2. Value per CAP year = 3 × (0.20 − 0.08) / (0.08 × 1.08) = 0.36 / 0.0864 ≈ $4.17bn.
3. Implied CAP = (290 − 125) / 4.17 ≈ 40 years.

Forty years. Before you either laugh or genuflect, run the two audits the skeleton demands.

Audit 1 is the buried-investment distortion. KO's true franchise investment includes more than the $3bn of capex and acquisitions the formula sees. It also includes the multi-billion annual brand spend expensed through the P&L, which is the capitalisation lesson from your earnings-quality work. Re-run with an economically capitalised I ≈ $5bn and incremental ROIC ≈ 22%. Value per year = 5 × 0.14 / 0.0864 ≈ $8.1bn, so implied CAP ≈ 165 / 8.1 ≈ 20 years. The truth is a band of roughly 20 to 40 years, and every input choice you just made is visible and arguable. That visibility is why you do it this way.

Audit 2 checks the base rates. A 20 to 40-year CAP asserts membership in the rarest club in capitalism, and almost no company qualifies. A fifty-year share table, habit-grade captivity, and a century of spread persistence make Coca-Cola one of the few arguable members. That's what the market is saying when it never lets the stock get cheap on spreadsheet inputs. Agreeing or refusing on aesthetics isn't your job. State the falsifiers (per-capita CSD volumes, share table, pricing-vs-volume mix) and price the band instead of the myth.

Last, the production fix. Confirm with a reverse DCF that uses an explicit fade, for example a spread eroding linearly to zero over T years. The step-function skeleton and the fade model will bracket the answer, and a quoted bracket is the deliverable you can defend.

Worked Example 3: The ten-minute cross-lens verdict: InterGlobe Aviation (IndiGo)

This is the Phase 8 compression drill, shown once at full length. Figures are illustrative at ~FY24/FY25 scale, so verify them against the annual report and DGCA monthly data.

Step 0 is the numbers. IndiGo has ~60%+ domestic passenger share and is the only Indian airline with a long record of profits, cyclical as they are. ROCE swings from spectacular to negative across fuel and demand cycles, so persistence fails on the spread test, though its performance relative to every peer is unbroken.

Step 1 is the five sources. IndiGo's candidate is cost advantage: fleet commonality, the sale-and-leaseback machine, aircraft utilisation, bulk A320 orders. Network effects are a no, because a flight schedule is scale and passenger N+1 gains nothing from passenger N. There are no switching costs either, since loyalty programmes are rebates and lock nobody in. As for brand, it's fame without a premium. It fails the behaviour test from the first moat module, because nobody pays ₹500 extra to fly the brand.

Step 2 is Greenwald. IndiGo's share table from FY15 to FY25 is violently unstable in its favour. It went from roughly 35% to 60%+ while Jet died, Go First died, SpiceJet shrank to a remnant. Moving shares + one firm's superior returns = superior competitor in a weak-barrier industry. That's Greenwald's early-Walmart pattern, and it isn't a protected industry. Entry remains cheap in the only sense that matters. Anyone can lease aircraft. Slots are a partial constraint, and they're the one real scarcity, so note it. Pilots are hireable. History obliges too, since a new deep-pocketed entrant, Akasa, funded in 2021, reached national scale in two years.

Step 3 is Helmer. IndiGo's claimed scale economies pass the Benefit test, with a unit-cost gap of a few percentage points of CASK against peers (illustrative). They pass only half the Barrier test. A rival can't match costs without matching scale, but customers are completely un-captive, so the king-combination interlock never forms. Process power is arguable. The turnaround-time machine is young-ish and only partly observable. Everything else is not a power.

Step 4 is Mauboussin and the CAP. Here the industry profit pool is ≈ negative-to-thin across cycles. That's the airline industry's structural affliction: commodity product, no pricing power, high operating leverage, suppliers with power. IndiGo's added value is real but supply-side only, which is Greenwald's most perishable family, defended by relentless execution. IndiGo's evidence-based CAP is short, low-to-mid single digits, refreshed continuously by operational excellence and never protected by structure. A valuation should therefore stay near steady-state economics on mid-cycle NOPAT, using the mid-cycle normalisation from the special-situations capstone. Any price granting a 15-year franchise CAP is claiming the industry has changed character, and the next fuel spike or funded entrant gets to referee that claim.

Your verdict fits in a line: dominant operator, thin-moat industry; scale without captivity; grade the company A and the barrier C, and price duration accordingly. At Phase 8 speed this takes ten minutes, because every instrument was pre-built.

Worked Example 4: The scale-economies cost curve, drawn: a regional detergent leader vs a funded attacker (India)

All entities and figures are synthetic, at ₹ scale, and computed for this example. Focus on the mechanism, because the decimals matter less. This is Greenwald's local-scale geometry and Helmer's scale economies turned into arithmetic.

Chamak (synthetic) leads a regional detergent market with 8,00,000 tonnes/yr, 60% of a ~13.3 lakh-tonne arena. Nova (synthetic), a well-funded attacker, sells 1,60,000 tonnes/yr (12%). Both have to carry roughly the same regional fixed-cost bundle of ₹480 crore/yr in advertising, depot network, and salesforce simply to be present across the territory. That lumpiness is what makes the advantage a scale economy. Variable cost is the same for both at ₹40,000/tonne, and the market price is ₹60,000/tonne.

Step 1 · draw the cost curve. Fixed cost per tonne = ₹480 crore ÷ volume:

• Chamak: 480 cr ÷ 8,00,000 = ₹6,000/tonne → unit cost 40,000 + 6,000 = ₹46,000 → +₹14,000/tonne of profit.
• Nova: 480 cr ÷ 1,60,000 = ₹30,000/tonne → unit cost 40,000 + 30,000 = ₹70,000 → −₹10,000/tonne: it loses money selling at the market price.

So the whole advantage comes down to one number, a ₹24,000/tonne unit-cost gap, and it exists for no reason except who spread ₹480 crore over more tonnes. Nobody did anything "better."

Step 2 · the Benefit. Chamak's power flows through Helmer's spend-less-per-unit channel. At the same price it banks ₹14,000/tonne, which is ≈ ₹1,120 crore/yr on 8,00,000 tonnes, while an equally competent rival bleeds. Or it can cut price toward ₹46,000 and still survive, and that standing threat keeps the whole region in line. Benefit: confirmed, quantified.

Step 3 · the Barrier (why me-too won't do). Nova's only cure is volume. Break-even needs fixed cost/tonne ≤ ₹20,000, which means 2,40,000 tonnes. That's an 18% share, +50% on today's volume, just to stop losing money. To reach Chamak's ₹6,000/tonne it has to match Chamak's 8,00,000 tonnes, which means seizing 60% of a market Chamak already holds. Buying that volume means underpricing, so model a five-year land-grab at ₹58,000/tonne, with contribution of ₹18,000/tonne over the ₹480 cr fixed base:

Discounted at the 12% hurdle, that campaign is worth ≈ −₹231 crore, and it only reaches 24% share, still short of Chamak's cost. It gets worse. Chamak, earning ₹1,120 crore/yr, can drop its price to ₹54,000, still above its ₹46,000 cost. That pushes Nova's break-even volume higher and turns the −₹231 crore into a deeper hole at no risk to Chamak. Nova's rational move is not to try. Barrier: confirmed, the prohibitive cost of share gains.

Verdict. Scale economies, gated. Benefit = a ₹24,000/tonne cost edge through the spend-less channel. Barrier = a share-buying NPV of ≈ −₹231 crore that the leader can always deepen. One caveat sets the CAP. This is scale alone. If detergent buyers are non-captive (weak habit, easy shelf-switching), Nova's task is "merely" the cost curve. A patient giant can win that, as Jio proved in telecom (Worked Example 1). Interlock the same cost curve with real captivity (the king combination) and the −₹231 crore never even earns its volume, because the customers won't move at any sane subsidy. With captivity in the mix, the arithmetic that failed to protect telecom does protect paints.

Worked Example 5: Switching-cost math: pricing the lock on an enterprise-software base (US)

A synthetic entity at $ scale, computed for this example. It's Helmer's switching costs reduced to the two numbers that decide it: the customer's cost to leave, and the price umbrella that cost creates.

Meridian Systems (synthetic) sells a vertical ERP to 2,000 enterprise customers at $0.50m per customer per year. A functionally identical rival exists. Does Meridian have switching-cost power, and what's it worth?

Step 1 · price the customer's cost to leave. Ripping Meridian out and installing the rival costs one customer, one time: re-implementation and integration ≈ $1.8m, data migration and parallel-running ≈ $0.6m, retraining and productivity dip ≈ $0.4m, and a risk-adjusted allowance for a botched cutover ≈ $0.7m, for a total S ≈ $3.5m.

Step 2 · amortise it into an annual number (the step most analysts skip). A one-time $3.5m can't be compared with an annual fee until you spread it over the relationship horizon. Over a 7-year remaining horizon at an 8% discount rate, the annuity factor is (1 − 1.08⁻⁷)/0.08 = 5.206, so the switch costs $3.5m ÷ 5.206 = $0.672m/year in present-value terms. The lazy $3.5m ÷ 7 = $0.50m ignores the time value and understates the lock. That's the classic trap.

Step 3 · the Barrier. A me-too rival has to compensate the customer for that $0.672m/yr. Meridian charges only $0.50m/yr, so the rival would have to price at 0.50 − 0.672 = −$0.172m/yr. It would have to give the software away and hand the customer $172k/year on top. That's NPV-suicide, and against a matched product nobody can poach the installed base. Barrier: confirmed, the rival must buy out the switch.

Step 4 · the Benefit, quantified. The lock is a price umbrella. A customer rationally stays until Meridian's fee reaches the switch-indifference ceiling = competitive price + amortised switch cost = 0.40 + 0.672 = $1.072m/yr. Meridian charges $0.50m, well under the ceiling. Against a no-lock world where the competitive price is ≈ $0.40m, the follow-on pricing it actually captures is (0.50 − 0.40) × 2,000 = $200m/year. It also holds back ~$0.57m per customer of unexercised headroom so it doesn't provoke migrations. Benefit: confirmed and quantified, through the charge-more channel on the installed base.

Verdict, with its two limits. Switching costs, gated: Benefit ≈ $200m/yr of captured follow-on pricing; Barrier = a $0.672m/yr wall a me-too rival cannot clear. Now run Helmer's two warnings on it. (1) New customers are level ground. The lock says nothing about the enterprises Meridian hasn't signed yet, so watch share-of-new-wins as the erosion tell. (2) The lock is strong but finite. A rival whose product is really worth ~$0.20m/yr more, given away free, delivers PV of (0.50 + 0.20) × 5.206 = $3.64m > $3.5m and just clears the wall. The barrier repels imitation, then, but a sufficiently superior, subsidised product gets through. That's why switching-cost CAPs belong in the mid band and stop short of the elite one. They protect the base you have, for as long as nobody invents materially past you.

Practice set

Problems 1 to 4 are guided. They apply the framework with scaffolding in place. Problems 5 to 9 are independent, and 10 to 12 are exam-condition (12 minutes each, closed book). Currency is ₹ crore and the hurdle is 12% unless stated. All companies are synthetic unless named.

P1 (guided: the metric). Synthetic industry: organised Indian QSR pizza. Shares five years apart: Firm A 50→44, Firm B 30→33, Firm C 12→17, Firm D 8→6. Five-year ROCE averages: A 9%, B 14%, C 21%, D 6%. (a) Compute the instability metric. (b) State the Greenwald verdict with the step-5 cross-check.

Solution. (a) Absolute movements: A 6, B 3, C 5, D 2 → total 16 points ÷ 4 firms = 4 points per firm ÷ 5 years = 0.8 points/firm/yr. That's the judgment zone, between ~0.5 and ~2. (b) Here the cross-check decides it. Shares are drifting steadily toward the one firm earning far above the hurdle (C, at 21%) while the ex-leader earns below it. It's the moving-shares-plus-one-winner pattern, the early-Walmart type. A superior competitor is conquering a weak-to-moderate-barrier market, and the industry isn't protected. Reward C in your analysis by finding its advantage and gating it, but don't give the industry credit for a barrier. C's own CAP claim has to rest on whether its advantage builds captivity. Industry structure won't carry it.

P2 (guided: asymmetry sort). Barrier or cost-of-playing? (a) A ₹1,500 crore cement plant. (b) Telecom spectrum won at auction. (c) A USFDA-approved plant after three years of inspections. (d) Prescriber habit for a 20-year-old branded chronic-therapy generic. (e) One of a fixed number of state liquor wholesale licences, granted decades ago. (f) Landing slots at a saturated metro airport.

Solution. (a) Cost of playing. An entrant's ₹1,500 crore buys the same plant. (b) Cost of playing. An auction price transfers the rent to the government, so it fails Helmer's non-arbitrage test. (c) Mostly cost of playing. It's expensive and slow but symmetric, and any competent entrant can eventually clear it. Treat it as a delay, worth a short CAP extension, and don't mistake it for a wall. (d) Barrier. This is demand-side captivity (habit + search costs) that an entrant can't purchase at any price, so the asymmetry is real. (e) Barrier while it lasts. It was granted below value (non-arbitraged fiat), but it's leased, so reprice it at the regulator's pleasure. (f) Barrier. Slot access is scarce and can't be replicated. Note that it protects incumbents collectively, and only at that airport, because the moat is local.

P3 (guided: Greenwald families). Sort each claimed advantage into supply / demand-captivity / economies of scale, or reject it. (a) "Our engineers are IIT toppers." (b) Tinting machines on 60,000 dealer counters. (c) Unit costs 18% below peers from a proprietary continuous-process line. (d) "Largest ad budget in the category," in a fixed-cost-heavy category, with 55% share. (e) "Customers rate us 4.8 stars." (f) 92% of users auto-renew without comparing alternatives.

Solution. (a) Reject. Talent is mobile and purchasable, which makes it operational excellence. (b) Demand/captivity. It's a switching cost bolted to the counter. (c) Supply. It's proprietary process technology, so mark it perishable. (d) Economies of scale. Budget per rupee of sales is what scale buys, and the 55% share on fixed-cost advertising is the mechanism. (e) Reject. Satisfaction without captivity or premium is a compliment, and it never reaches cash flow. (f) Demand/captivity. This is habit in its purest measurable form.

P4 (guided: the king combination). Using only facts covered above, explain in ≤150 words why ₹1.5+ lakh crore broke Indian telecom's incumbents but ₹10,000 crore is not guaranteed to break paints.

Solution. Telecom's incumbents had scale without captivity. Subscribers faced near-zero switching costs (MNP, no habit premium for a commodity SIM), so Jio's capital could buy the customers directly, and once volume moved, the incumbents' scale advantage moved with it. Paints pairs scale with captivity. An entrant's money can build factories and even place tinting machines, but painter habit, dealer credit relationships, and four-a-day replenishment service can't be bought quickly at any price. So volume moves slowly, which keeps the entrant's unit fixed costs high, which keeps its offer inferior, which keeps volume moving slowly. Scale alone is a fort with the gate open. Scale interlocked with captivity makes the attacker pay twice. It pays once to build capacity, and again, year after year, to pry away customers who have no price at which they're quickly for sale. Let the FY25 to 28 share table referee.

P5 (independent: the Helmer gate). Power, or not a power? Name it, and give Benefit + Barrier in one line each. (a) "Our app has 50 million downloads." (b) "Enterprise deployments take 14 months and integrate into the client's ERP." (c) "We are India's largest cement producer." (d) "Our granted process patent runs 12 more years and underpins our 8-point gross-margin edge." (e) "Incumbent retailers can't match our 10-minute delivery without destroying their store economics." (f) "Voted 'most trusted brand' three years running." (g) "Unit costs fall ~15% per doubling of cumulative output; we have 8× the volume of #2 in our niche."

Solution. (a) Not a power. An audience is neither a benefit, since no cash-flow line is named, nor a barrier, since downloads are buyable. (b) Switching costs. Benefit: pricing on renewals and modules to the installed base. Barrier: a rival must fund 14 months of migration risk to poach. (c) Not a power as stated. Cement competes in freight-radius regions, so national size spreads no local fixed cost. Restate it as regional density or reject it. (d) Cornered resource. Benefit: the margin edge. Barrier: fiat (patent law), time-limited, so its CAP contribution is capped at ~12 years with a fade after. (e) Counter-positioning. Benefit: a superior model for the use-case. Barrier: the incumbent's rational refusal to cannibalise itself. It's temporary, so demand the next power being built. (f) Not a power. It's fame without a price-premium or captivity mechanism, so run the behaviour test from the first moat module. (g) Scale economies (experience-curve variant). Benefit: lowest unit cost. Barrier: the prohibitive cost of share gains, since #2 must fund 8× volume growth against a leader who can always price below its costs.

P6 (independent: incumbent's dilemma, compute). A leading insurance distributor earns EBITDA ₹300 crore via commission-heavy agents. A direct-online challenger's model, if matched, gives: Year 1 −₹120 crore, Year 2 ₹0, Year 3 +₹60 crore. Harvesting gives ₹250, ₹220, ₹190 crore. WACC 10%. Compute both three-year NPVs and state the barrier this creates.

Solution. Match: −120/1.10 + 0/1.21 + 60/1.331 = −109.1 + 0 + 45.1 = −₹64 crore. Harvest: 250/1.10 + 220/1.21 + 190/1.331 = 227.3 + 181.8 + 142.8 = +₹551.9 crore. Harvesting wins by ~₹616 crore, so a rational board declines to match this year. Re-run the numbers every year and it declines every year. That refusal is the challenger's barrier, the "will not" flavour of counter-positioning. Agency sharpens the dilemma, because executive horizons are shorter than the decay. It resolves only at capitulation, and by then the challenger must have built its next power, or its own window closes too.

P7 (independent: the clock). Timing-valid or timing-violation? (a) A 2-year-old D2C skincare brand claims branding power. (b) A biotech startup claims a cornered resource via an in-licensed molecule. (c) The #4 player in a tipped two-sided marketplace announces it will "build network effects." (d) A 3-year-old SaaS firm in an exploding category runs deep losses to win share. (e) A 60-year-old confectioner claims process power over its fermentation process.

Solution. (a) Violation. Branding is hysteresis, and two years can't compress into an attribution built from long history. (b) Valid. A cornered resource is an origination-stage power. Check the licence terms against the five tests, especially non-arbitrage, by asking what it paid. (c) Violation. Takeoff is over. Past tipping, share gains cost more than they return, and the barrier now works against the latecomer. (d) Valid as a race, unproven as a power. This is rational takeoff-stage spending. Grade it as "no proven moat yet" and watch whether captivity or network value actually accrues. (e) Valid if evidenced. Sixty years satisfies hysteresis, so demand the NUMMI-style proof: rivals with full visibility who still can't match cost/quality.

P8 (independent: CAP arithmetic). NOPAT ₹600 crore, WACC 12%, incremental ROIC 24%, reinvestment 50% of NOPAT, market EV ₹9,000 crore. Compute (a) steady-state value, (b) value per CAP year, (c) the market-implied CAP, (d) the EV a 6-year CAP would justify.

Solution. (a) 600/0.12 = ₹5,000 crore. (b) I = 300, so 300 × (0.24 − 0.12)/(0.12 × 1.12) = 36/0.1344 ≈ ₹268 crore/yr. (c) (9,000 − 5,000)/268 ≈ 14.9 ≈ 15 years. (d) 5,000 + 6 × 268 = ≈ ₹6,608 crore (≈ 11.0× NOPAT). Between (c) and (d) there's a gap of ₹2,400 crore, the price of nine disputed years of moat. Here the whole disagreement between a bull and a bear is a duration argument, and that's where the four lenses aim.

P9 (independent: what must be true). Your gated evidence supports a 10-year CAP, while the market's implied CAP is 25 years. List (i) what an owner of the stock is implicitly asserting, and (ii) three pieces of evidence that would legitimately move your 10 toward 25.

Solution. (i) An owner asserts that the spread survives 2.5× longer than the gated evidence shows. Either the moat belongs to the elite CAP band (interlocked scale-plus-captivity with decades of demonstrated stability), or reinvestment and ROIC will beat the model's inputs. Put another way, the owner is betting that mean reversion, the base-rate fate, will make an exception for identifiable structural reasons. Admiration doesn't count as one. (ii) First, a share table showing stability through a funded attack, which proves the barriers under fire and not just in peacetime. Second, evidence that a second power is interlocking with the first, for example switching costs forming on top of scale as the king combination assembles. Third, a reinvestment runway expanding at constant incremental ROIC. New geographies or adjacencies would be earning the same spread. Check that incremental ROIC the way the first moat module taught you, and ignore management's TAM slides.

P10 (exam: 12 min). An industry's four firms have held shares within ±1 point for 15 years. Sector ROIC has averaged 11.5% against an 11–12% WACC. A colleague calls it "a proven moat, look at that stability." Write the correct verdict and name which lens-steps catch the error.

Solution. Stalemate, not moat. Stability is necessary but not sufficient, because barriers must show up in returns (Greenwald's step 5, and Mauboussin's Pass 2, "is there anything to defend?"). Fifteen frozen years at cost-of-capital returns describe undifferentiated oligopolists whose customers keep them symmetric. Nobody can gain share profitably, and nobody earns economic profit. Think of airlines before consolidation, or much of two-wheeler and cement history. Treat it as a no-barrier business. Value it at steady state (CAP ≈ 0), reward only best-operator economics, and note the one upgrade path. If the incumbents ever achieve pricing restraint and returns rise above the hurdle and entry stays absent, re-run the verdict. The telecom postscript in Worked Example 1 is this same watch.

P11 (exam: 12 min, full protocol). MediTrack (synthetic) sells hospital-information systems to 220 of India's ~600 large private hospitals. Gross retention is 95%, implementations take 18 months, and two global rivals compete, with shares stable for a decade. Inputs: 8-year average ROCE 28%, WACC 12%, incremental ROIC 25%, reinvestment 40%, EV = 18× NOPAT. Run the protocol and deliver the five-sentence verdict, including the implied CAP.

Solution. Step 0: 28% vs 12% for 8 years is a persistent spread, so proceed. Steps 1 to 3: the source is switching costs, procedural (18-month implementations) and relational (clinical-workflow lock). It passes the gate. Benefit: renewal and module pricing on 220 captive sites. Barrier: rivals must fund 18 months of migration risk to poach. Shares have been stable for a decade, with the metric well under 0.5/yr on these facts, and the timing is valid because the installed base was built during the category's takeoff. Step 4 arithmetic: steady state = 8.33× NOPAT; value/CAP-yr per unit of NOPAT = 0.40 × 0.13/0.1344 ≈ 0.387×; implied CAP = (18 − 8.33)/0.387 ≈ 25 years. Five-sentence verdict: "MediTrack owns a gated switching-cost power confirmed by a frozen share table and eight years of 16-point spreads. The moat protects the installed base, not new wins; the 380 unconverted hospitals are contested ground where all vendors fight level. The price implies ~25 years of constant-spread reinvestment, above the 10-to-15-year band this evidence class earns. Owning it here asserts either elite-band durability or under-modelled reinvestment; the falsifiers are share-of-new-wins, net revenue retention, and any rival shipping a credible migration tool. Great business; the argument is entirely about duration, and the market has taken the optimistic side."

P12 (exam: 12 min, market definition). A listed innerwear leader shows a stable ~55% share of the "organised premium innerwear" market for a decade, earning 45% ROCE. Your junior notes that including unorganised players, its all-market share rose from 8% to 14% over the same decade: "shares are moving >0.5/yr, so barriers are weak." Adjudicate: which market definition is right, and what is the verdict?

Solution. Define the market where the fixed costs and customers actually compete. Its brand spend, exclusive-outlet network, and price points compete within organised premium innerwear, while the unorganised segment competes on price for a different customer with different economics. Within the true arena, stable 55% + 45% ROCE = barriers, made of brand-habit captivity + distribution scale (gate them the Helmer way). What the organised segment gains collectively from the unorganised one is category formalisation, a demand tailwind from GST and urbanisation, and it isn't competitive churn. It lengthens the reinvestment runway, with more I at high ROIC, and leaves the barriers intact. Your junior's error is that the instability he measured is migration between arenas, while the test asks about movement within one. Verdict: the moat is confirmed within the relevant market, and formalisation extends the CAP's reinvestment leg. Watch for the day the definition itself dissolves. D2C entrants selling premium product at unorganised prices would be counter-positioning, and the share table to watch would change.

P13 (guided: the scale-cost curve). Synthetic regional biscuit maker. A competitive regional presence requires a fixed-cost bundle (advertising + depots) of ₹300 crore/yr. Variable cost is ₹80/kg and the price is ₹100/kg. The leader sells 5,00,000 tonnes/yr and a challenger sells 1,00,000 tonnes/yr. (a) Fixed cost per kg and unit cost for each. (b) Per-kg profit for each at the ₹100 price. (c) The challenger's break-even volume. (d) Name the power and give the Benefit + Barrier verdict.

Solution. (a) 5,00,000 t = 50 crore kg, 1,00,000 t = 10 crore kg → leader fixed/kg = 300 cr ÷ 50 cr = ₹6; challenger = 300 cr ÷ 10 cr = ₹30 → unit costs ₹86 and ₹110. (b) Leader +₹14/kg; challenger −₹10/kg, so it loses money at the market price. (c) Break-even needs fixed/kg ≤ ₹20 (price − variable), so volume ≥ 300 cr ÷ ₹20 = 1,50,000 tonnes. At 1,00,000 t, the challenger has to grow +50% just to stop bleeding, and 5× to match the leader's cost. (d) Scale economies. Benefit: a ₹24/kg unit-cost edge via the spend-less channel. Barrier: the prohibitive cost of share gains. A challenger can't reach competitive cost without buying volume it can only buy at a loss, while the leader reprices to keep that campaign under water. If biscuit buyers are non-captive, flag that this is scale alone, which a patient giant can beat unless captivity interlocks.

P14 (guided: switching-cost math). Synthetic US cloud-analytics vendor. One-time switching cost per customer: re-implementation $0.8m, migration $0.3m, retraining $0.2m, risk allowance $0.2m. The remaining horizon is 6 years at a discount rate of 9%. The vendor charges $0.30m/customer/yr, a no-lock competitive price would be $0.22m, and the installed base is 1,200 customers. (a) Amortise the switching cost to an annual figure. (b) What must a me-too rival charge to make switching rational, and what does that imply? (c) Compute the follow-on pricing captured and give the Benefit + Barrier verdict.

Solution. (a) S = 0.8 + 0.3 + 0.2 + 0.2 = $1.5m. For 6 yrs @ 9%, the annuity factor = (1 − 1.09⁻⁶)/0.09 = 4.486, so the amortised switch cost = 1.5 ÷ 4.486 = $0.334m/yr. Don't use the naive 1.5 ÷ 6 = $0.25m, because discounting matters. (b) A rival must price at 0.30 − 0.334 = −$0.034m/yr. That means giving the product away and paying the customer ~$34k/yr, which is NPV-negative, so a me-too can't poach the installed base. (c) Captured follow-on pricing = (0.30 − 0.22) × 1,200 = $96m/yr. Its switch-indifference ceiling is 0.22 + 0.334 = $0.554m, so unexercised headroom remains. Verdict: switching costs. Benefit $96m/yr of installed-base pricing; Barrier a $0.334m/yr wall. One caveat: new customers are contested on level ground, so watch share-of-new-wins.

P15 (independent: counter-positioning, compute). A US legacy cable bundler earns operating cash flow of $2.0bn/yr. A streaming challenger's model, if matched, projects EBIT of −$0.9bn, −$0.3bn, +$0.4bn (years 1–3); harvesting the bundle projects $1.7bn, $1.4bn, $1.1bn. WACC 9%. (a) Compute both three-year NPVs. (b) Name the power and its barrier flavor. (c) State the one thing the analyst must demand of the challenger.

Solution. (a) Match: −0.9/1.09 − 0.3/1.09² + 0.4/1.09³ = −0.826 − 0.252 + 0.309 = −$0.77bn. Harvest: 1.7/1.09 + 1.4/1.09² + 1.1/1.09³ = 1.560 + 1.178 + 0.849 = +$3.59bn. So the board rationally harvests, by ≈ $4.36bn. (b) Counter-positioning, the "will not" barrier. Matching is forbidden by the incumbent's own NPV math, because it vaporises the bundle economics. (c) The next power. Counter-positioning is temporary and expires at capitulation, so the thesis has to name what the challenger is building with the time it bought. That might be scale economies in content, or a subscriber base with switching costs. A challenger with no second act has won a stay of execution, and that's no moat.

P16 (independent: CAP with a fade twist). NOPAT $300m, WACC 9%, incremental ROIC 18%, reinvestment 45% of NOPAT, market EV $4,200m. (a) Steady-state value. (b) Value per full-spread CAP year. (c) The market-implied CAP under the step-function skeleton. (d) A colleague argues the spread will fade linearly instead of stopping dead, delivering roughly half the value per year. What CAP would the same price then imply, and which reading is more defensible?

Solution. (a) SS = 300 ÷ 0.09 = $3,333m. (b) I = 0.45 × 300 = $135m, so value/yr = 135 × (0.18 − 0.09)/(0.09 × 1.09) = $123.9m. (c) Implied CAP = (4,200 − 3,333)/123.9 ≈ **7.0 years** of full spread. (d) If a fade halves the per-year value, matching the same $867m of franchise value takes ≈ 14 years of fading spread. Step function and fade bracket the truth between them. A fade reading is more defensible because real spreads erode gradually, and quoting the bracket ("7 years flat, or ~14 fading") beats pretending the skeleton's single number is a measurement.

P17 (independent: the non-arbitrage test). A firm wins a 20-year toll-road concession whose cash flows are worth ₹4,100 crore (PV at award). (a) It paid ₹4,000 crore at a competitive auction. How much economic rent did it corner, and does the concession pass Helmer's non-arbitrage test? (b) A rival firm receives an administratively granted concession of the same ₹4,100-crore value for an application fee of ₹200 crore. Rent captured, and verdict? (c) Which of the two, if either, is a cornered resource, and what warning attaches?

Solution. (a) Rent = 4,100 − 4,000 = ₹100 crore, a sliver, because the auction transferred essentially all the value to the grantor. Non-arbitrage test fails. A resource bought at or near fair value corners nothing. You paid for the rent, so you don't own it, and Indian telecom spectrum is the national case. (b) Rent = 4,100 − 200 = ₹3,900 crore, captured below value, so it passes non-arbitrage. (c) Only (b) is a cornered resource, and it's a leased one. Fiat granted it and fiat can reprice it. That's the IRCTC warning again. Cap the CAP and treat the moat as owned at the regulator's pleasure. The auctioned road in (a) is a fine asset and a fine business. It just isn't a power.

P18 (exam: 15 min, full protocol + implied CAP). PayGrid (synthetic, India) runs a transaction-switching network with 70% of national volume. It's two-sided (banks on one side, merchants on the other). Inputs: 12-year ROCE 30% against a 12% WACC, incremental ROIC 26%, reinvestment 35%, shares stable for a decade, price 22× NOPAT. A regulator periodically caps the interchange PayGrid may charge. Run the four-lens protocol and deliver the five-sentence verdict, including the market-implied CAP.

Solution. Step 0: 30% vs 12% for 12 years is a persistent spread, so proceed. Step 1: the candidate sources are network economies (two-sided liquidity) + scale economies (fixed switching-infrastructure cost spread over volume). Reject "brand." Step 2 (Greenwald): the market is national payment-switching, and shares frozen for a decade put it in the strong-barrier zone. Its asymmetry is that a challenger must assemble both sides at once, because banks won't join without merchants and vice versa. That's the classic cold-start wall. Step 3 (Helmer): network economies. Benefit: each new bank makes the network more valuable to every merchant, so user N+1 gains because N joined, which makes it a true network and not an audience. Barrier: past tipping, no feasible subsidy repopulates an empty rival network, and scale interlocks. But one layer is leased. A regulator sets the interchange the whole model monetises, and that's a fiat overhang. Step 4 arithmetic: SS multiple = 1/0.12 = 8.33×; value per CAP-year per unit of NOPAT = 0.35 × (0.26 − 0.12)/(0.12 × 1.12) = 0.365×; implied CAP = (22 − 8.33)/0.365 ≈ **37 years**. Five-sentence verdict: "PayGrid owns a genuine two-sided network power interlocked with switching-infrastructure scale, confirmed by a decade of frozen share and 18-point spreads. The barrier is a cold-start wall no me-too can climb, so the network itself is close to unassailable. But the cash flow rides on a regulated interchange the state can cut by circular, a leased layer that caps the defensible CAP well below what the structure alone would earn. The price implies ~37 years of constant-spread reinvestment, a duration granted to almost no company in history and blind to the regulatory reprice risk. Great network, real moat; the argument is entirely about the two things the multiple ignores, the fade and the regulator, and the market has priced neither."

P19 (independent: the three-tier stack, barrier present). Nashik Valves (synthetic, India, ₹ crore; hurdle 12%, tax 25%). Reproduction inputs: plant at current replacement cost 540, receivables 300, inventory 350, operating cash 45, trade payables 240. The entrant would need three years of channel and specification-approval spend at ₹60 crore a year. Earnings inputs: through-cycle EBIT ₹210 crore, with depreciation exceeding maintenance capital expenditure by ₹12 crore. (a) Asset reproduction value. (b) Earnings power value of the operations. (c) Franchise value, computed as a residual and again by the excess-return identity. (d) The barrier verdict, and the one input you would stress first.

Solution. (a) 540 + 300 + 350 + 45 − 240 + 3 × 60 = ₹1,175 crore. (b) Adjusted EBIT = 210 + 12 = 222, and after tax NOPAT = ₹166.5 crore, so EPV = 166.5 ÷ 0.12 = ₹1,387.5 crore. (c) Residual = 1,387.5 − 1,175 = ₹212.5 crore. Identity: excess return = 166.5 − 0.12 × 1,175 = ₹25.5 crore, capitalised at 12% = ₹212.5 crore, which agrees. (d) There are barriers, but modest ones. Its reproducible base earns 14.17% against a 12% hurdle, a 2.17-point spread, and franchise is only 15.3% of earnings power. That puts it in the three-to-seven-year CAP band, well short of the fifteen-to-twenty band. Stress the capitalised channel spend first, because it's both the softest input and the largest single swing factor. Drop it from three years to one: reproduction value rises to ₹1,295 crore, franchise falls to ₹92.5 crore, and the barrier claim thins to something a share table would have to carry on its own.

P20 (independent: the three-tier stack, no barrier, and why the DCF lies). Bhiwadi Castings (synthetic, India, ₹ crore; hurdle 12%, tax 25%), a contract manufacturer of machined castings. Reproduction inputs: plant 900, receivables 260, inventory 340, operating cash 40, trade payables 210. There's no brand or channel spend an entrant would need to repeat, because the customers specify to drawing and dual-source by policy. Earnings inputs: through-cycle EBIT ₹150 crore, with depreciation equal to maintenance capital expenditure. Management guides to 30% reinvestment at the current return and asks to be valued on a ten-year growth DCF. (a) The three tiers. (b) What the sign of franchise value means here. (c) Value management's growth plan. (d) The verdict, and where in the capital cycle this sits.

Solution. (a) Reproduction = 900 + 260 + 340 + 40 − 210 = ₹1,330 crore. NOPAT = 150 × 0.75 = ₹112.5 crore, so EPV = 112.5 ÷ 0.12 = ₹937.5 crore. Franchise = 937.5 − 1,330 = −₹392.5 crore. (b) A negative franchise isn't a small franchise. It says the assets in place can't earn their own cost of capital, since the reproducible base returns 112.5 ÷ 1,330 = 8.46% against a 12% hurdle. No rational entrant builds a plant here. That's a fact about the industry and not about this management. It means the industry is carrying more capacity than demand will pay for. (c) Reinvestment is ₹33.75 crore a year at an incremental return of 8.46%, so each year of the plan is worth 33.75 × (0.0846 − 0.12) ÷ (0.12 × 1.12) = −₹8.89 crore, and ten years subtract ₹88.9 crore, taking enterprise value to about ₹848.6 crore. Growth below the hurdle destroys value at a computable rate. It's the iron rule from your cost-of-capital work with a number attached. (d) No barriers, no CAP, and a growth plan that should be refused. Value the business between its earnings power and what the assets would fetch on exit. Note that this is stage 5 of the capital cycle you're about to study. Capacity has to leave before returns can recover, and the catalyst to watch is industry capital expenditure falling below industry depreciation. Nothing Bhiwadi announces counts as that catalyst.

Applied mini-project: the Cross-Lens Moat Dossier

Task. Produce a moat dossier of four pages at most, running the full four-lens protocol on ONE of these two companies. The choice is yours. Your Phase 4 capstone will make you do both countries, so pick the one you won't use there:

• Pidilite Industries (India). For data, use screener.in ten-year financials (ROCE, margins, incremental returns) and the latest annual report and investor presentation from pidilite.com. Those give you segment mix, distribution reach, market-share claims and their CRISIL/industry citations. Add the HP Adhesives DRHP (2021, on SEBI's website), whose industry chapter is a free, sourced map of the Indian adhesives market with shares. Damodaran's industry ROIC data for specialty chemicals rounds it out. Your dossier must answer three questions. Is "Fevicol" branding power, habit captivity, or both? Run the behaviour test on pricing history to find out. Is the carpenter/contractor ecosystem a switching cost? And what would an attacker with ₹5,000 crore be unable to buy?
• Costco (US). For data, use the latest 10-K from EDGAR (Item 1 Competition, membership counts, renewal rates, which Costco publishes). Use the annual report's arithmetic of membership-fee income against operating income too, along with one industry share source (the NRF Top-100 retailers list, free) and Damodaran's retail ROIC data. Your questions: where does Costco's advantage belong among Greenwald's three families? Careful here, because the membership model makes it a hybrid. Is "scale economies shared", deliberately passing scale savings to members, a deepening of the barrier or a forfeit of the benefit? And what CAP does the market's multiple imply, and what would justify it?

Required contents (the capstone rubric is a superset of this one):

1. Share table for the correctly defined market, 3+ dates, with the instability metric computed and sources cited.
2. Greenwald verdict: the named asymmetry (or its absence), the three-families classification, king-combination check.
3. Helmer power-stack table: every claimed power with Benefit, Barrier, origination date; at least one candidate claim explicitly rejected as "not a power" with the reason.
4. Mauboussin audit notes: profit-pool sketch, industry-vs-firm attribution, government/disruption scan.
5. The numbers: ROIC/ROCE series vs hurdle; your evidence-based CAP (band + fade logic vs base rates); the market-implied CAP via the M&M skeleton and a reverse-DCF cross-check.
6. The verdict sentence from the four-lens protocol, plus three named falsifiers you would monitor.

Scoring rubric (score yourself strictly; pass ≥13/16):

Case Lab: run case CL-4.3 (The Two Detergents of Kanpur) in the app's Cases tab for this module: a scale-plus-captivity regional leader versus a capital-rich attacker, walked through Greenwald's asymmetry test, Helmer's Benefit + Barrier gate, and an implied-CAP inversion, so you build the four-lens verdict end to end before the dossier.

Reading & resources

• Greenwald & Kahn, Competition Demystified: chs. 1–4 (the collapse to barriers, the three advantages, how to assess them, including market-share stability) and ch. 5, "Big Where It Counts" (Walmart, Coors: local scale). The back half (entry and pricing games) is optional now, required before M4.06. [Paid] [Beginner→Intermediate]
• Helmer, 7 Powers: The Foundations of Business Strategy: Part I, one power per chapter with the Benefit/Barrier chart; Part II for the power progression and "me too won't do." Read with a pencil; the diagrams are the book. [Paid] [Intermediate]
• Mauboussin & Callahan, Measuring the Moat (Morgan Stanley Counterpoint Global, 2024 update): the audit protocol and checklist; the single highest-value free PDF in this phase. Download and keep. [Free] [Intermediate]
• Mauboussin & Johnson, "Competitive Advantage Period: The Neglected Value Driver" (1997): the original CAP paper; short, foundational. Free copies circulate online. [Free] [Intermediate]
• Mauboussin & Callahan, The Base Rate Book (Credit Suisse, 2016): the mean-reversion evidence your CAP estimates must answer to. [Free] [Advanced]
• Dorsey, The Little Book That Builds Wealth: M4.04's anchor. Skim it again after this week and notice how much sharper the five sources read through Helmer's gate. [Paid] [Beginner]
• Helmer on podcasts: his long interviews on Invest Like the Best and Acquired walk the seven powers with fresh examples, free, in his own voice. [Free] [Beginner→Intermediate]
• Business Breakdowns (podcast): the Asian Paints and Costco episodes pair directly with this week's specimens, and Acquired's Costco and Walmart episodes are book-length audits of local-scale geometry in action. [Free] [Beginner]
• Data for your share tables: TRAI monthly subscriber reports (trai.gov.in) [Free]; DRHP industry chapters on SEBI's site, which are free, CRISIL/Frost-sourced share data for dozens of Indian industries (the HP Adhesives, Indigo Paints, and Nykaa DRHPs are model examples) [Free]; SIAM/FADA (autos), AMFI (funds), IRDAI (insurance) [Free]; US 10-K Item 1 Competition sections via EDGAR full-text search [Free]; Damodaran's industry ROIC/margin datasets (pages.stern.nyu.edu, updated each January) [Free].
• Marcellus Investment Managers' public letters on Asian Paints/Pidilite-class franchises: useful compilations of moat evidence for Indian compounders. Read them as advocacy, and audit every claim with this week's instruments. [Free] [Intermediate]
• Going deeper · Bruce Greenwald, Columbia Business School lectures on competition & barriers to entry (free on YouTube; search "Greenwald competition demystified lecture") [Free] [Intermediate]: the share-stability test and local-scale geometry in the author's own voice. Watch the Walmart/Coors segments alongside the local-dominance and share-stability material to see the cost-curve arithmetic of Worked Example 4 taught live.
• Going deeper · Mauboussin & Rappaport, Expectations Investing (revised 2021) [Paid] [Intermediate]: turns the implied-CAP inversion into a full price-implied-expectations workflow. It's the natural bridge from this week's CAP math to M3.05's reverse DCF, and the book to read before the capstone if the "read the market's CAP" step still feels mechanical.

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": "Turning a single point-estimate implied CAP into a sensitivity grid across the reinvestment-rate and incremental-ROIC assumptions", "use": "Have an LLM help structure a two-way sensitivity table (reinvestment rate x incremental ROIC) from a company's NOPAT, WACC, and EV, so the debate about a price moves from one contested number to a map of which assumption actually swings the answer.", "tools": ["Claude or ChatGPT to draft and format the sensitivity-grid code from raw inputs", "a notebook (the block below) to run the M&M skeleton across a range of assumptions", "screener.in / Tijori (India) or stockanalysis.com / TIKR / the 10-K (US) for NOPAT, invested capital, and EV"], "prompt": "Here are [company]'s NOPAT, WACC, current EV, and its trailing incremental ROIC and reinvestment rate. Build a grid of implied CAP across reinvestment rates from half to 1.5x the current rate, and incremental ROIC from WACC+2pp to WACC+20pp. Flag which single assumption, moved by a realistic amount, changes the implied CAP the most.", "verify": "Recompute the grid's base-case cell (today's actual reinvestment rate and incremental ROIC) by hand using the M&M skeleton and confirm it matches both the AI's grid and your own §4.11 calculation before trusting any other cell.", "diy": "The mastery check hands you NOPAT, WACC, EV, incremental ROIC, and reinvestment rate and asks for the single-point implied CAP by hand, with no grid and no tool in the room. The sensitivity table above only tells you which of your assumptions deserves the most scrutiny before you defend that single number." }

Modern Data Analysis

By hand first. The M&M skeleton and the implied-CAP inversion haven't changed, and they're what the mastery check gates.

Today's workflow. A single implied-CAP number invites a single-number argument ("40 years is too many"). A sensitivity grid is a more useful thing to bring to the table. Hold NOPAT, WACC, and EV fixed at their filed/quoted values. Then sweep reinvestment rate and incremental ROIC across a plausible range. Now the debate becomes "which assumption is actually doing the work," and nobody has to duel over one output.

Tools & sources (IN + US). India: screener.in / Tijori for NOPAT, invested capital, and 10-year reinvestment-rate history, plus company investor decks for management's own incremental-ROIC commentary. US: 10-K/EDGAR for NOPAT and reinvestment, Damodaran's WACC-by-industry tables (pages.stern.nyu.edu) for a defensible discount rate, stockanalysis.com / TIKR for market EV.

# recomputed in-session -- R3. The M&M skeleton run as a function, then a 2D sensitivity sweep.
def implied_cap(NOPAT, WACC, EV, inc_ROIC, reinvest):
    SS = NOPAT / WACC
    I = reinvest * NOPAT
    value_per_yr = I * (inc_ROIC - WACC) / (WACC * (1 + WACC))
    return (EV - SS) / value_per_yr

# --- anchor: Worked Example 2's own Coca-Cola inputs ---
NOPAT, WACC, EV = 10.0, 0.08, 290.0        # $bn
cap_anchor = implied_cap(NOPAT, WACC, EV, inc_ROIC=0.20, reinvest=0.30)   # -> 39.6 (module: "~40 years")

# --- extension: sweep reinvestment rate (rows) x incremental ROIC (cols) ---
grid = {(reinv, iroic): implied_cap(NOPAT, WACC, EV, iroic, reinv)
        for reinv in (0.20, 0.25, 0.30, 0.35, 0.40)
        for iroic in (0.14, 0.16, 0.18, 0.20, 0.22, 0.25)}

# --- the asymmetry: a SAME-SIZE absolute nudge to each input, not a same-size relative one ---
cap_base   = implied_cap(NOPAT, WACC, EV, 0.20, 0.30)
cap_iroic2 = implied_cap(NOPAT, WACC, EV, 0.22, 0.30)   # +2pp on incremental ROIC
cap_reinv2 = implied_cap(NOPAT, WACC, EV, 0.20, 0.32)   # +2pp on reinvestment rate
# cap_base -> 39.6 ; cap_iroic2 -> 33.9 (-14.3%) ; cap_reinv2 -> 37.1 (-6.2%)

Verify. implied_cap(10, 0.08, 290, 0.20, 0.30) returns 39.6 years, which matches Worked Example 2's own "~40 years" (steady-state $125bn = 43% of the $290bn price, value/CAP-year $4.17bn).

Quantitative lens

The sensitivity grid. Implied CAP (years) across reinvestment rate and incremental ROIC, holding Coca-Cola's own NOPAT/WACC/EV fixed:

The bolded cell (30% reinvestment, 20% incremental ROIC) is the Worked Example 2 base case. Read the whole grid and you see straight away that the top-left corner implies CAPs of 60 to 120 years, durations no base rate supports. Even the bottom-right corner implies 21 to 35 years, which is itself a "wide, grudging" grant. No cell in this table gets close to a "no-barrier" CAP of 0 to 3 years. Coca-Cola's price, in other words, is a long-duration bet across almost the whole plausible range of the two inputs. It doesn't hang on one contestable assumption.

Which assumption actually moves the answer. The M&M value-per-CAP-year term is a simple product of the reinvestment amount and the spread. So a given relative change in either input moves CAP by the same proportion. A 10% relative rise in reinvestment rate and a 10% relative rise in the spread both cut CAP by 9.1%, and there's no asymmetry there. The asymmetry that matters in practice is a different one. Analysts argue about these two inputs in absolute percentage points. "What if ROIC is 2 points better" and "what if they reinvest 2 points more" sound like disagreements of about the same size. They aren't. A +2pp move on incremental ROIC (20% to 22%) takes the spread from 12pp to 14pp, a 16.7% relative jump on the smaller base, and it cuts the implied CAP by 14.3%. That same +2pp move on the reinvestment rate (30% to 32%) is only a 6.7% relative jump on the larger base, and it cuts the implied CAP by just 6.2%. So two analysts who disagree by "a couple of points" on the durability of the spread are having a much bigger argument, in valuation terms, than two who disagree by "a couple of points" on the reinvestment rate. Out loud, the two disagreements sound the same size.

Limits. The grid and the M&M skeleton it runs on share every limitation the CAP discussion already named. It's a step function, with full spread for CAP years and zero after, and P16 showed why a gradual fade is more defensible. Every number in this table therefore over-values the mid-years and under-values the true shape of a decaying franchise. A grid tells you which assumption to argue about. It won't tell you the fair value. Its reinvestment-vs-spread asymmetry is a property of the M&M formula's algebra, and it says nothing about which input is harder to forecast in practice. As it happens, a company's near-term reinvestment rate is often the more observable of the two, because it's a management choice visible in this year's capex and working-capital numbers. Multi-year durability of a competitive spread is closer to unknowable. That's a second, independent reason the spread deserves more scrutiny per point of disagreement, on top of the algebra.

Where this goes next: galaxy cross-links

• G1 · QM1.03 (timeseries-montecarlo-ml): turns the fixed sensitivity grid above into a proper Monte Carlo over the reinvestment-rate/incremental-ROIC joint distribution.
• G1 · FI1.02 (duration-convexity), the fixed-income sibling of this exact idea: a CAP is a duration, and this addendum's sensitivity grid is a convexity map for it.
• G2 · DA1.02 (pandas-data-wrangling): scales the implied_cap function above from one company's grid to a cross-sectional CAP table across an entire coverage universe.

Teach it back & journal

Feynman prompt. Write a one-page explainer for a smart 15-year-old: "Why does the same one sweet shop in our town stay busy and profitable for thirty years while three new cafés open and close every year?" Without jargon, your page has to smuggle in three ideas. First comes the only question that matters: what stops a copycat from opening next door and winning? Second is the two-part test any real answer must pass, which is that the shop gets something out of it AND the copycat can't or won't match it. Third, the answer is ultimately a number of years, how long the shop's edge will last, and that's what you'd be buying if you bought the shop. If your explainer needs the words "moat," "hysteresis," or "NPV," you don't own it yet.

Journal prompt. Three entries. (1) Before this week, name the company whose moat you most believed in, then run its claim through Helmer's gate and the share-stability test on paper: did it survive, and which lens did the damage? (2) Which lens do you instinctively over-trust, story-rich Helmer or number-hard Greenwald, and what error will that bias produce in your capstone dossier if unwatched? (3) Write down one price you recently thought was "obviously expensive" and compute what CAP it actually implies; record whether the number changed your confidence, and why.

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