The Schools I: Graham to Buffett/Munger

Phase 6 · Week 52 · ~20 hours

Where it sits. For forty-seven weeks you have been building machinery: reading statements (Phase 1), trusting or distrusting them (Phase 2), pricing what they say (Phase 3), grading the durability behind the numbers (Phase 4), and learning each sector's dialect (Phase 5). You now possess, in raw form, the answer to all five spine questions about almost any business. What you do not yet have is a coherent philosophy for turning those answers into decisions: a set of first principles for how a rational person should relate to a price, to uncertainty, and to their own mind.

That is what Phase 6 installs. This is its first module, the hinge where the whole program turns from technique to judgment, and it lays the foundation stone: the two schools that between them define value investing, and the evolution that connects them. You will learn Graham (intrinsic value versus price, Mr. Market, the margin of safety, net-nets, and the defensive/enterprising split) and then the Graham→Buffett evolution via Munger: why the best student of deep value spent sixty years migrating from "fair companies at wonderful prices" to "wonderful companies at fair prices," and what the four filters, owner earnings, economic goodwill, and the insurance-float model added to Graham's austere original. The second module covers the rest of the schools (Fisher, Lynch, Greenblatt, Klarman, Marks, Pabrai, and the modern quality-compounders) and forces the concentration-versus-diversification decision; the third builds the decision-science latticework; the fourth the behavioral edge; the fifth assembles your personal checklist. The gate is Schools exam part A.

Suggested hour budget. Core content: 8h · worked examples (See's economics, the net-net screen, the margin-of-safety computation): 3h · practice set: 4h · applied mini-project (four filters + margin of safety on a real company): 3h · flashcards + mastery check: 2h.

App link. This module seeds the Philosophy & Behavior branch's first lesson set and contributes the "value vs price" and "four filters" case cards to the case engine. It begins your personal investment checklist (the artifact completed in the capstone): every principle here that ends in a decision rule (demand a margin of safety, never let Mr. Market set your estimate, price is what you pay and value is what you get) becomes a checklist line you will carry for decades.

Learning objectives

By the end you can:

1. State and defend the price–value distinction from first principles: define intrinsic value as the discounted value of the cash a business will hand its owners over its life, explain why market price diverges from it, and articulate why that divergence is the value investor's entire opportunity.
2. Deploy the Mr. Market parable as an operating discipline: describe Graham's manic-depressive business partner precisely, and convert the allegory into three standing rules (the market serves you, it does not instruct you; his mood is your opportunity, not your valuation; you are never forced to transact).
3. Reason about the margin of safety quantitatively: explain why it is the central concept (estimate fallibility plus adverse futures), and how much to demand by scaling the discount to the width of your Phase-3 valuation range and the durability of the business, applied to a conservative estimate rather than an optimistic one.
4. Compute and screen a Graham net-net: calculate net current asset value (NCAV = current assets − total liabilities), apply the two-thirds rule and the diversification requirement, explain why net-nets worked historically and why they are rare today, and identify the value-trap ("melting ice cube") failure mode.
5. Distinguish Graham's defensive and enterprising investors: recall the defensive stock-selection criteria (size, financial strength, earnings and dividend record, moderate P/E and P/B), state what extra work the enterprising investor takes on, and place yourself.
6. Explain cigar-butt investing and its limits: define the "one last puff" trade, show with the compounding math why it fails at scale and why time is the mediocre business's enemy, and name the transition point that pushed Buffett past it.
7. Narrate the Graham→Buffett evolution and Munger's role: give the three forces (scale, the compounding arithmetic of durable high returns, and Munger's quality turn), and use See's Candies as the hinge case that proved a franchise's economics to Buffett.
8. Apply the Buffett/Munger four filters (understandable business, durable moat, able and honest management, sensible price) to a real company, and reconcile owner earnings to the free-cash-flow work of M2.04, distinguish economic goodwill from accounting goodwill, and explain pricing power, float, and "the first rule is don't lose money" as the pillars of the quality school.

Prerequisites & connections

Builds on. M0.01: the five-questions spine and the ROIC>WACC master idea. The philosophy here tells you what to do once the spine is answered. M2.04, free cash flow, especially its owner-earnings section. Owner earnings isn't re-derived here, but you'll see why Buffett made it the unit of value and how it reconciles to the FCF you already compute. M3.04–M3.05, intrinsic valuation and, above all, the reverse DCF. The margin of safety is a discount applied to the Phase-3 valuation range, and "what is the price implying?" is the question every Mr. Market episode secretly asks. M4.04–M4.05, the moat. Filter two of the four filters is the durable competitive advantage you learned to prove in the numbers, and See's economic goodwill is a moat expressed in balance-sheet language. M5.03, insurance float: the float section below is the philosophy behind that module's mechanics. M3.07, capital allocation and reading a proxy. Filter three (able and honest management) is that same skill, now aimed at judgement instead of computation.

Feeds forward. M6.02 completes the schools (Fisher's scuttlebutt and growth, Lynch's GARP, Greenblatt's magic formula, Klarman's absolute value, Marks's cycles, Pabrai's asymmetry, and the Smith/Akre/Sleep quality-compounders). Every one of them is a variation, extension or rebuttal of the Graham-to-Buffett axis you fix here, so this is the reference frame for that one. M6.03 formalises circle of competence (filter one) and the ruin-avoidance behind "don't lose money." M6.05 turns every decision rule you meet here into a line of your written checklist. Phase 8's teardown question 12 ("cheap or expensive vs quality?") and the deep-dive thesis both end in a margin-of-safety judgement that comes straight from M6.01. Competency C9 (investor's temperament and process) is founded here, and C5 (value three ways) gets its philosophy here.

The short version. Price is what you pay and value is what you get. The gap between them, offered daily by an emotional Mr. Market, is your only opportunity. The margin of safety is your only protection against being wrong about how big that gap is. And the great evolution of value investing was the discovery that the most durable margin of safety comes from a wonderful business whose intrinsic value grows past your purchase price year after year, where Graham had looked for it in a cheap balance sheet.

Core teaching content

1. The one idea underneath everything: price is not value

Every school in Phase 6, every valuation method in Phase 3 and every decision you'll ever make as an investor rests on a single distinction that beginners collapse and professionals hold sacred:

Price is what you pay. Value is what you get. They are set by different processes, they are frequently far apart, and the entire craft of investing is the disciplined exploitation of their divergence.

Price is a number a market produces. It's the last transaction between a buyer and a seller, each with their own motives, horizons, information and emotional state, and it updates every few seconds. It's also a fact. At 2:47 p.m. the stock changed hands at ₹1,842, and nobody can argue with that. But a fact isn't an opinion, and it isn't a truth about the business either. A price tells you what the marginal, most-motivated participant would transact at right now, and nothing more.

Value, or intrinsic value, is something else entirely. Graham, and later Buffett, give it a precise definition borrowed from John Burr Williams in The Theory of Investment Value (1938):

The intrinsic value of any business is the discounted value of the cash that can be taken out of it over its remaining life.

You've already built this machine. It's the DCF from the valuation phase: forecast the owner cash flows, discount them at an appropriate rate, sum them. The philosophy adds just one insistence, and you must never lose it. Intrinsic value is a property of the business. It comes from the business's economics, meaning its returns on capital, how long its moat lasts and how much room it has to reinvest. Price is a property of the market's mood about the business. A long, loose, unreliable tether connects the two. Graham taught it and Buffett repeats it in a famous formulation:

In the short run the market is a voting machine; in the long run it is a weighing machine.

Day to day, price registers popularity. It's a vote, swung by fear, greed, narrative and flows. Year over year and decade over decade, the business's actual cash generation becomes undeniable and drags the price toward value. That's the weighing. Your edge as a value investor is to estimate the weight while everyone else counts votes. Then you transact only when the vote has pushed the price far enough from the weight to pay you for the risk of being wrong.

Why this isn't obvious, and why the market can be beaten at all. The efficient-market view, which you'll meet evenhandedly in the portfolio phase, says price already equals the best estimate of value. On that view the gap is an illusion and chasing it is a waste. The value investor's rebuttal is a modest one. It doesn't claim markets are usually wrong, only that they're sometimes wrong at knowable moments. Prices are set by humans, and human crowds get seized by emotion from time to time. Phase 6's whole second half is about that. You don't need the market to be inefficient every day. You need it to be inefficient occasionally and detectably, and you need the patience to act only then. Everything Graham built is machinery for waiting for those moments without fooling yourself in between.

2. Mr. Market: the most useful allegory in finance

Graham's most important teaching device appears in Chapter 8 of The Intelligent Investor. Buffett has called it one of the two ideas an investor really needs, the other being the margin of safety. Learn to recite it. You'll use it every week for the rest of your investing life.

The parable. Imagine you own a small stake in a private business, and one of your partners is a man named Mr. Market. He's obliging to a fault. Every single day, without being asked, he shows up and names a price at which he'll either buy your share or sell you more of his. Your business plods along steadily. Mr. Market doesn't. He's emotionally unstable, even manic-depressive. On his euphoric days he sees only blue skies and quotes you a wildly high price, terrified you'll snatch his stake before the boom. On his despairing days he sees only ruin and quotes a pathetically low price, desperate to dump his share before the roof caves in. None of these moods has anything to do with the business, which is doing just what it did yesterday.

And you're under no obligation to trade with him. Everything else in the parable follows from that. He'll be back tomorrow with a new quote, so his daily price is a service offered to you. He isn't instructing you to do anything. So:

• When he's euphoric and his price is far above your estimate of value, you may sell to him, or simply ignore him.
• When he's despairing and his price is far below value, you may buy from him, or ignore him.
• On the many days when his price is roughly sensible, you do nothing.

The catastrophic error, the one Graham built the allegory to prevent, is letting Mr. Market's mood become yours. You feel bullish because he's bidding high and frightened because he's bidding low. That turns the logic upside down. His high prices should make you more cautious about buying and his low prices more interested, because price moves inversely to prospective return. You saw the same point in the reverse DCF, since a higher price implies more optimistic assumptions and so a worse deal. If you buy because the price is rising and sell because it's falling, you've hired Mr. Market as your analyst. It's the one job he's constitutionally unfit for.

Three standing rules to memorise as checklist lines:

1. The market is there to serve you, not to instruct you. Its job is to offer prices, some of them foolish. Form your intrinsic-value estimate independently, from the business's economics, before you look at the quote. Otherwise the quote anchors your judgement, an early taste of the anchoring bias you'll study properly in the behavioural-finance module.
2. His mood is your opportunity, not your valuation. A falling price is a cheaper offer for the same thing. It tells you nothing about whether you were wrong on value, unless the facts changed. Telling a mood swing from a thesis break is the hardest and most important distinction in practice.
3. You are never forced to swing. This is Buffett's baseball image, and the decision-science module formalises it. Investing is a game with no called strikes. You can stand at the plate all day letting pitches go by, and you only have to swing at the fat one Mr. Market eventually lobs. Most of the time the correct action is none.

The subtle trap: facts versus mood. Rule 2 hides the skill that separates masters from the merely patient. A price can fall for two utterly different reasons. Maybe Mr. Market is having a bad day about an unchanged business. That's a gift. Or maybe the business's intrinsic value really fell, and that's a warning. Mix the two up and it ruins you in both directions. You hold melting ice cubes, or average down on them, and call it "buying the dip." You panic-sell wonderful businesses and call it "risk management." The only defence is to have done the work of Phases 1 to 5 so thoroughly that when the price drops, you can re-underwrite the business and ask whether a fact changed, instead of reacting to the quote. Mr. Market tells you the price has changed. Only your analysis can tell you whether the value has.

6. Cigar-butt investing and its limits

The bridge from Graham to Buffett runs straight through a vivid metaphor. Buffett coined it for the strategy he practised in his early years and later outgrew.

The cigar butt. Buffett named the strategy in his 1989 letter to Berkshire shareholders, reviewing the mistakes of his first twenty-five years. Buy a stock at a low enough price, he argued there, and some stumble in the business will usually hand you a chance to sell at a decent profit, however poor that business proves to be over a longer horizon. The image he chose is what made the idea permanent. Picture a discarded cigar stub on the pavement with one puff left in it. Nobody would pay for that puff, and it's a wretched smoke. But you paid nothing, so as Buffett put it, "the 'bargain purchase' will make that puff all profit."

The logic is pure Graham. A business needn't be good, only cheap enough. You pick up the stub for nothing, take the single free puff (a bounce, a re-rating, a liquidation, a buyout) and move on to the next stub. Net-nets are cigar butts. So were Buffett's own early triumphs: Sanborn Map, Dempster Mill and, fatefully, the Berkshire Hathaway textile mill itself.

Why it works, at first. On small sums, in an inefficient market, a diversified basket of cigar butts is a legitimately profitable, low-risk strategy. You're buying below liquidation value, so arithmetic protects your downside, and across a basket the "one puff" comes often enough to compound decently. That's how Buffett ran his partnership in the 1950s and 60s, and it produced spectacular percentage returns on the modest capital he managed then.

Why it fails: the four limits. Buffett came to call the cigar-butt approach "a mistake" for his purposes, and his reasons make the whole case for the quality turn:

1. It doesn't scale. A net-net might have a ₹50 crore market cap. You can't deploy ₹50,000 crore into names like that, because there aren't enough of them and your own buying would move the price. As capital grows, the cigar-butt universe stops mattering. Buffett has said his early small-capital returns are the ones he's proudest of and the ones he can never repeat. Size is the strategy's enemy.
2. Time is the mediocre business's enemy. This is the deep one, and it's pure arithmetic from Phases 3 and 4. A cigar butt is by definition a poor business that earns low returns on capital, or none. Buy it cheap and have it re-rate quickly, and you win. But if the "one puff" is delayed and you're stuck holding for years, you now own a business earning below its cost of capital. A bad business destroys value every year you hold it, and as the DCF work showed, growth with ROIC < WACC makes the hole deeper. A cheap price is a one-time gain. A bad business is a recurring loss. Time works against the cigar-butt holder and for the owner of a compounder, and that inversion is what defines the two schools.
3. The friction is brutal. The strategy needs constant turnover: buy the stub, take the puff, sell, find the next stub. That means constant transaction costs, constant short-term taxes (the after-tax arithmetic from the valuation phase) and constant re-underwriting effort. A great business bought once and held for twenty years pays the tax just once at the end and lets the whole pre-tax sum compound in between. The cigar-butt investor gives up that colossal advantage every time they sell.
4. The cheapness often signals a real problem. Many cigar butts are cheap because they're dying (the melting ice cube), or because the people running them are happy to waste the assets. The "one puff" never comes. It's a wet stub.

The transition point. Buffett's own accounting names the hinge. He bought the Berkshire Hathaway textile business in the 1960s as a cigar butt and then, out of stubbornness and pique, poured years into it. By his estimate it was a monumentally costly mistake. He didn't lose money on the purchase. The cost was the opportunity cost of two decades of capital and attention sunk into a structurally doomed business when they could have gone into good ones. "Time is the friend of the wonderful business, the enemy of the mediocre," he concluded. The man who taught him to escape that trap was Charlie Munger, and with Munger the evolution begins in earnest.

8. The four filters: Buffett and Munger's acquisition test

Buffett has described Berkshire's entire investment and acquisition process as passing a candidate through four filters. Munger compressed it like this: "We're trying to buy businesses with sustainable competitive advantages, at a low — or even a fair — price, run by honest and able people." Buffett's fuller version names four conditions, and a business has to clear all four. They're filters in series, and you don't average them like a scorecard. They're also the five-questions spine from the orientation week restated as a philosophy. Here the course's whole machinery finally becomes a single judgement.

Filter 1. A business we can understand. Can you explain, in plain language, how this company makes money, what drives it and what could break it? And do you have a defensible view of where it'll be in ten years? If not, you're outside your circle of competence, which the decision-science module formalises. No other filter can rescue you then, because filters 2, 3, and 4 all need judgements you can't make about a business you don't understand. It comes first because it's a gate, the precondition for having any opinion at all, and no strength elsewhere makes up for missing it. Munger's rule is that the circle needn't be large, but you must know where its edge is. A great deal of investing skill is simply having the humility to put the "too hard" businesses in the "too hard" pile and walk away. This maps to spine question 1 (what does it do and how does it make money?) and to Phase 8's teardown, which begins by forcing you to state the business in one line.

Filter 2. A durable competitive advantage. Favourable long-term economics, meaning a business protected by a moat (the whole of Phase 4) that keeps returns on capital durably above the cost of capital. You know this one already. It's the five sources of a moat, confirmed by the numbers test: persistent ROIC, stable margins, stable share. The philosophy adds the word durable. Buffett wants a moat that will still be there in ten and twenty years, because filter 2 is really a bet on how long the ROIC−WACC spread lasts, the Competitive Advantage Period from the moat module. Duration is what a long-term owner is buying. At heart the four filters are a way of finding businesses you can hold for decades without the value eroding, so it's no surprise that filter 2 is the one Buffett and Munger spend the most words on. It maps to spine question 2 (is it a good business?) and to question 4 (what can kill it?), whose answer here is the moat's erosion risks.

Filter 3. Able and honest management. You want people who are both competent operators and trustworthy stewards of shareholders' capital. Both halves matter, and they're different. A brilliant operator who's a dishonest or self-dealing capital allocator will still transfer your money to themselves, and that's where Phase 8's proxy work, the capital-allocation grading from the valuation phase and the India-specific promoter-governance lens all come in. An honest mediocrity will simply waste it slowly.

Buffett's tests, which you'll apply in Phase 8, start with capital. Does management allocate it rationally across the five uses (reinvest, acquire, pay dividends, buy back stock below value, cut debt), graded over a decade? Do they tell the truth in good times and bad? You'll see it when they own their mistakes in candid letters and keep clear of the accounting games from Phase 2. Then check that their incentives line up with owners' through ownership stakes and sensible pay metrics, which you can read in the DEF 14A or the Board's Report. Buffett's own caution here is really about filter 3 acting as a gate on the other three. Send a manager with a dazzling reputation into a business with terrible economics, he observed in the Berkshire letters, and only one of those two reputations survives the encounter: "it is the reputation of the business that remains intact." Management can operate a moat. It can't substitute for one. What a dishonest management can do is destroy a real moat's value to you, by diverting the cash the moat produces. This maps to spine question 3.

Filter 4. A sensible price (with a margin of safety). Price enters only after a business clears filters 1 to 3, and that ordering is the whole philosophy in a single design choice. Graham let price lead, buying anything cheap enough. Buffett and Munger make price the last filter, applied only to businesses already judged understandable, moaty and well-run. The price has to be "sensible" or "attractive." In the language you've been learning, that means it offers a margin of safety against your conservatively estimated intrinsic value, sized to how predictable the business is.

Notice the softening from Graham that Munger's arithmetic justifies. For a business that really is wonderful and durable, a fair price is enough, because the compounding of intrinsic value supplies the dynamic margin of safety over time. You don't need Graham's deep static discount when the business itself is doing the work. But "fair" doesn't mean "any". Buffett's point is that even a wonderful business can be overpaid for. A fully priced moat earns you the market return at best, which was the final mistake in the moat module, and the reverse DCF is the tool that tells you when a great business's price has run past sanity. This maps to spine question 5 (what is the price implying, and is that beatable?).

The filters in series, worked as a decision. The power of the four filters is that they reject fast. A dazzling growth story you don't understand fails at filter 1. Stop there, because you don't need any further analysis. A business you understand perfectly but which has no moat fails at filter 2. Think of an airline or a commodity producer. You may still trade it on mid-cycle value, as you learned at the end of the valuation phase, but it isn't a Buffett-filter compounder. A wonderful, moaty, understandable business run by a promoter with a history of siphoning cash to related parties fails at filter 3. Pass, whatever the price. And a business that clears 1, 2, and 3 but trades at a price whose reverse DCF implies thirty years of accelerating perfection fails at filter 4. It's a wonderful business at the wrong price, so put it on the watchlist and wait for Mr. Market to have a bad day. The four-filters worked example and the practice set both drill this reject-fast habit, because knowing when to stop is most of what the filters buy you.

9. Owner earnings, economic goodwill, and pricing power: the machinery of quality

The quality school needs a unit of value more truthful than reported earnings, and a way of thinking about the asset that captures what a balance sheet misses. Buffett supplied both.

Owner earnings: what an owner can really take out. You built this in the free-cash-flow module, so here's why the philosophy insists on it. Reported net income is an accountant's opinion, which was Phase 2's whole warning, and even cash flow from operations flatters a business that has to spend heavily just to stand still. Buffett's owner earnings, from the 1986 Berkshire letter, is the number a rational private owner would actually care about:

Owner earnings = reported net income
               + depreciation, amortization, and other non-cash charges
               − maintenance capital expenditure (the capex required to
                 preserve the business's competitive position and unit volume)
               ± required change in working capital

Everything hinges on the maintenance-capex line. As the free-cash-flow module stressed, it's an estimate. Buffett himself calls it "necessarily a guess." Owner earnings deliberately leaves out growth capex. That's optional investment, which staying alive doesn't require. It counts only the spending needed to keep the current earning power intact. That makes it the right unit for the quality school, because a wonderful business is, almost by definition, one whose owner earnings are large relative to the capital it must reinvest to maintain them. See's and Nestlé India need almost nothing back, so nearly all their reported profit is extractable owner earnings. A steel mill or an airline has to plough most of its cash flow back just to keep the plant competitive, so its owner earnings are a fraction of reported profit.

Line it up with the free-cash-flow framework and owner earnings ≈ CFO (with working capital normalised) − maintenance capex. For a grower it comes out above free cash flow, because FCF is depressed by growth capex that owner earnings excludes. The FCFF forecasting you did in the DCF work is, in spirit, owner earnings with the maintenance/growth split made explicit and the growth capex modelled instead of dropped. Buffett's punchline is the bridge from Phase 2 into valuation: "we consider the owner-earnings figure, not the GAAP figure, to be the relevant item for valuation purposes."

Economic goodwill: what the balance sheet can't see. This idea, from the appendix to Buffett's 1983 letter, explains why See's was cheap at three times tangible book. There are two kinds of goodwill, and you need to keep them apart:

• Accounting goodwill is a bookkeeping residue. When one company buys another for more than the fair value of its identifiable net assets, the premium goes on the balance sheet as "goodwill." It looks backward, it only ever declines (through impairment, as the long-term-assets module showed), and it tells you what was paid. It says nothing about what anything is worth.
• Economic goodwill is the real thing. It's the capitalised value of a business's ability to earn returns on tangible capital far above the market rate, the franchise value that lets See's earn tens of millions on $40m of assets. It looks forward, it grows (especially in nominal terms during inflation), and it appears on no balance sheet.

That gave Buffett the insight that reshaped his investing. The businesses worth owning are those with large and growing economic goodwill and small tangible-capital needs. Why growing, and why does inflation matter? A business with pricing power can raise nominal prices as inflation lifts costs, so its nominal earnings rise. Meanwhile its small tangible capital base needs only modest nominal reinvestment to keep pace. Its economic goodwill therefore inflates in nominal terms, handing the owner a real asset that compounds. A capital-heavy business is the opposite case. Inflation forces it to replace plant and inventory at ever-higher nominal costs, devouring cash just to stand still, so inflation is its enemy. That's the deep reason Buffett migrated toward capital-light franchises. They're inflation-resistant compounding machines, and their most valuable asset is the very one no accountant records. The intangible-brand moat from the moat module and this economic-goodwill idea are the same thing seen through the two lenses of strategy and accounting.

10. Intrinsic value as discounted owner earnings, and "the first rule is don't lose money"

Intrinsic value, defined and estimated. Put the opening definition of value together with owner earnings as the unit, and you have the quality school's complete theory of value:

The intrinsic value of a business is the discounted present value of the owner earnings it will generate over its remaining life.

It's the DCF from the valuation phase with owner earnings as the cash-flow line, and Buffett is emphatic on two points the philosophy must hold. First, it's a definition. Don't expect it to work like a formula that yields precision. Intrinsic value is knowable only in principle, because it depends on an unknowable future. Two honest, skilled analysts will compute different numbers. You might think that's a flaw in the method. It's simply the nature of the thing, and Buffett calls intrinsic value "an estimate rather than a precise figure."

The second point is what sets it apart from academic DCF. Because the estimate is imprecise, you confine yourself to businesses whose futures are predictable enough to estimate at all. That's filter 1, and it's the reason Buffett avoids hard-to-forecast businesses however exciting they are. You also demand a margin of safety against the estimate's imprecision, which is filter 4. So intrinsic value and margin of safety work as one system. The first says value is a discounted, uncertain estimate of owner earnings. The second says that, because of this, you pay enough below it to survive being wrong. Everything in Graham and Buffett hangs from that single hook.

"Rule No. 1: Never lose money. Rule No. 2: Never forget Rule No. 1." This is Buffett's most quoted and most misunderstood line. Read literally, it's absurd. Everyone takes losses, and Buffett has taken plenty. Read correctly, it's the deepest risk principle in the whole course, and it means avoiding the permanent loss of capital. As Phase 9 will insist, you need to tell apart two things a beginner runs together:

• A quotational loss is when Mr. Market temporarily marks your holding down. To a value investor that's no real loss. It's often an opportunity, a chance to buy more of an unchanged business cheaper. Volatility isn't risk.
• A permanent loss is one where the business's intrinsic value was really and irreversibly impaired: a fraud, a broken moat, a balance sheet that failed, a price so high the value could never grow into it. That's the thing to avoid at almost any cost.

Why "at almost any cost"? Because of the arithmetic of drawdowns, which every investor should feel in their bones. Losses and gains aren't symmetric. A 50% loss needs a 100% gain just to break even. A 75% loss needs a 300% gain, and a 90% loss needs 900%. Compounding is a chain, and a single zero (or near-zero) link breaks the whole chain however strong the others are. One catastrophic loss can erase decades of good decisions.

That's the mathematical case for survival-first investing, and it's the thread that ties everything here together. The margin of safety is the tool that prevents permanent loss. Filter 1 (understand it) prevents the losses that come from not knowing what you own. Filter 3 (honest management) prevents the losses that come from being robbed. Don't-lose-money is the objective all of them serve. The decision-science module's ergodicity and ruin-avoidance, and the risk module's definition of risk as probability × magnitude of impairment, make this idea rigorous. For now, hold it as the emotional and mathematical core of value investing: you get rich by not going broke, slowly and repeatedly avoiding the permanent loss, and letting compounding do the rest.

12. The synthesis: one philosophy, two eras

Step back and pull all of it into a single frame you can carry into the next module and every teardown after it.

Look at the bottom row. Graham isn't obsolete, and Buffett didn't refute him. In a market panic (2008, March 2020, an Indian small-cap crash) Graham's tools come back into their own. The enterprising investor who can compute an NCAV and demand a static margin of safety will find bargains that the quality investor's patience alone won't surface. In ordinary and expensive markets, over decades, Buffett's synthesis is the more powerful engine, because it lets time and compounding carry the load and it defers the tax. The complete Phase 6 investor holds both. You keep Graham's arithmetic rigour and scepticism of price alongside Buffett's insistence on business quality and durability, and you know which to reach for as the environment and your own capital dictate. The rest of the phase refines that combined stance, and the checklist you write at its end will encode it. The next module widens the aperture to the other great schools (Fisher's growth and scuttlebutt, Lynch's GARP, Greenblatt's special situations, Klarman's absolute value, Marks's cycles, Pabrai's asymmetry, and the modern quality-compounders). You'll understand each of them better for having fixed this Graham-to-Buffett axis first.

Common mistakes & how experts think differently

1. Letting the price set the value. The beginner looks at the quote first and reasons backward: "it's up 40%, it must be a great business; it's down 40%, something must be wrong." That hands Mr. Market the one job he can't do. The expert forms an independent intrinsic-value estimate from the business's economics before looking at the price, then treats the price purely as an offer to accept, reject, or ignore. Here's how to tell you've made this mistake. Your conviction about a company rises and falls with its stock chart.

2. Treating a falling price as a broken thesis (or a rising price as confirmation). This is the hardest skill from the Mr. Market section: telling a mood swing, which is an opportunity, from a fact change, which is a warning. Beginners average down on melting ice cubes ("it's even cheaper now!") and panic-sell compounders ("the market knows something"). The expert re-underwrites the business on every big move, asking whether a fact actually changed, and lets the analysis deliver the verdict, never the quote. Price tells you the market's mood. Only your work tells you whether value moved.

3. Confusing a cheap price with a margin of safety. Take a 60% discount to a number you can't actually estimate, because you don't understand the business or its future really is unknowable. That's a made-up cushion, whatever you call it. The expert knows the margin of safety only protects against uncertainty you can analyse, in a business you understand. So filter 1 comes before filter 4, and the deepest-looking discounts (distressed, pre-revenue, fraud-suspect) are often the least safe.

4. Buying cigar butts with big money, or holding them too long. Beginners who discover Graham run net-net screens and buy the cheapest statistical junk. Then they hold it for years waiting for a puff that never comes, collecting the recurring losses of a bad business. The expert knows cigar butts are a small-capital, diversified, fast-turnover strategy for names that really trade below liquidation value. On any serious capital or any long hold, the mediocre business's time-cost overwhelms the entry discount. Time is the friend of the wonderful business and the enemy of the mediocre one.

5. Misreading the Graham-to-Buffett evolution as "quality at any price." In its lazy version, the quality turn drops the fourth filter entirely: "it's a wonderful business, so price doesn't matter." It does matter. A wonderful business at thirty times an earnings stream that the reverse DCF shows to be implausible earns you the market return at best and a permanent loss at worst. That's filter 4 failing. The expert holds business quality (filters 1 to 3) and price (filter 4) in separate registers and never lets the first answer the second. Munger says "fair price," not "any price."

6. Using owner earnings (or any cash metric) without the maintenance-capex judgement. Whatever truth owner earnings tells comes from the maintenance-capex estimate, as the free-cash-flow module warned. Beginners use either reported EPS, which ignores the reinvestment a business needs to survive, or headline "cash flow," which ignores it too. Both flatter capital-heavy businesses. The expert triangulates maintenance capex (depreciation proxy, Greenwald's method, unit math), quotes a range, and treats the disagreement between methods as information about accounting quality.

7. Admiring a moat while ignoring economic goodwill and pricing power. Beginners describe moats qualitatively ("great brand") and stop. The expert turns the moat into the two quantities that make it worth money. Does it produce economic goodwill, meaning high returns on small tangible capital, growing in nominal terms? Does it have pricing power, with price increases that stick while volume holds? A "moat" with neither is a story. A moat with both is a Buffett-filter compounder.

8. Thinking "don't lose money" means "avoid volatility." The beginner reads Rule No. 1 as a mandate to avoid drawdowns, and so sells at the bottom, holds cash in panics, and misses the recoveries. The expert reads it as "avoid permanent loss" and welcomes quotational declines in understood businesses as gifts. The fear is saved for the four real destroyers of capital: not understanding what you own, a broken moat, a failed balance sheet, and a price too high to ever grow into. Volatility is the price of admission. Permanent impairment is the thing to avoid.

9. Forgetting that the safest structural edge (float) depends entirely on the people running it. Awed by Berkshire's float engine, beginners conclude "leverage is fine if it's float." The expert knows float is safe only while the underwriting that produces it stays disciplined, and that careless underwriting turns float into expensive leverage in disguise. It's a special case of a general law. Any cheap capital advantage is real only as long as the character producing it refuses to abuse it. The safety was never in the float. It was in the people.

Worked examples

Worked Example 1: See's Candies: why "3× tangible book" was cheap (the economic-goodwill engine)

The figures here are the actual numbers Buffett reported in his 2007 Berkshire letter (and the 1983 appendix). Pull the letter from berkshirehathaway.com. It's free, and it repays a dozen re-reads. The point of the exercise is to feel, in arithmetic, the difference between a cheap asset and a wonderful business.

Step 1. The purchase, judged by Graham's rules. In 1972 Blue Chip Stamps (Buffett/Munger) bought See's for $25 million. At the time See's had:

• Net tangible assets ≈ $8 million
• Sales ≈ $30 million
• Pre-tax earnings ≈ $4 million, so ~$2m after the ~50% corporate tax of the era

By Graham's arithmetic this wasn't a bargain. Price ÷ tangible book = 25 / 8 ≈ 3.1×, against Graham's defensive limit of ~1.5×, and price ÷ after-tax earnings ≈ 25 / 2 ≈ 12 to 13×. A pure Grahamite would pass, and the young Buffett nearly did, over the final $5m. A net-net this was not.

Step 2. What the balance sheet couldn't see: the return on tangible capital. Look instead at what See's earned on the capital it used. $4m pre-tax on $8m of net tangible assets = 50% pre-tax return on tangible capital (~25% after tax). Candy companies don't earn that. Franchises do. Ordinary businesses earn ~12% on tangible capital, and See's earned four times that, the signature of large economic goodwill. The $17m Buffett paid above tangible book ($25m − $8m) looks like a Graham overpayment. It was really the market value of an asset no accountant had recorded, a brand with pricing power.

Step 3. The payoff, 35 years later. By 2007 Buffett reported See's at:

• Sales ≈ $383 million (up ~13×)
• Pre-tax profit ≈ $82 million (up ~20×)
• Capital required to run it ≈ $40 million

Now the two computations that reshaped Buffett's career:

• Return on tangible capital, 2007: $82m ÷ $40m = 205% pre-tax. The franchise didn't just persist. It intensified.
• Incremental return on incremental capital, 1972 to 2007: earnings rose by ($82m − $4m) = $78m while capital rose by only ($40m − $8m) = $32m. Incremental pre-tax return = 78 ÷ 32 ≈ 244%. Every extra dollar See's needed to grow threw off ~$2.44 of annual pre-tax profit.

Step 4. The river of cash, and the lesson that matters most. Over 1972 to 2007, See's earned roughly $1.35 billion of cumulative pre-tax profit and needed only about $32 million of additional capital to do it. The other ~$1.32 billion was shipped to Berkshire headquarters, and Buffett reinvested it in entirely different businesses. That's the difference everything turns on:

A cigar butt gives you one puff and you're done. See's gave a growing river of cash for fifty years while asking for almost no capital back, and that cash bought the next See's, and the next.

Step 5. Economic goodwill vs accounting goodwill, in inflation. This is Buffett's 1983 point, made concrete. Accounting goodwill from the deal was $17m, and it could only decline through amortisation. Economic goodwill, the capitalised franchise value, grew. By 2007 an $82m pre-tax stream on $40m of tangible capital was worth well over a billion dollars. So economic goodwill had compounded from ~$17m to ~$1bn+ (roughly 60×) while the accounting version withered. And it happened through the high-inflation 1970s. See's could raise prices as costs rose while needing almost no extra tangible capital, so inflation lifted its nominal earnings and with them its economic goodwill. That's the inflation-resistant compounding described earlier. Now take Buffett's comparison, a capital-heavy business earning the same $4m in 1972 but needing, say, $18m of assets to do it. It would have had to pour ever more nominal capital into plant and inventory just to keep pace with inflation, and it would have compounded far less. Same earnings in 1972, opposite destinies, decided entirely by capital intensity and pricing power.

Worked Example 2: Nestlé India: the Indian See's, run through the four filters (illustrative)

See's has no Indian twin, but India's great consumer franchises reproduce its economics. Nestlé India is the cleanest parallel. It's a capital-light packaged-foods business whose brands (Maggi, Nescafé, KitKat, Munch, Cerelac) command pricing power. All the figures below are illustrative, modelled on ~FY24/FY25 scale, so pull the actual annual report. Nestlé India reports to March, and screener.in has the ten-year page.

Step 1. The economics (why it is a See's). Start with the shape of the business:

• Revenue ≈ ₹19,000 to 20,000 crore. Gross margin ≈ 55 to 57%, EBITDA margin ≈ 23 to 25%, net profit ≈ ₹3,300 crore.
• Capital intensity: tiny. Working capital is negative. Distributors and the trade pay fast. Suppliers and the parent are financed by payables. Fixed assets are modest relative to profit, so operating invested capital is small, often a low single-digit percentage of sales.
• Return on capital: enormous, as a result. ROCE/ROE has often run from the high double digits to over 100%, depending on the year and how the small equity base is treated. Nestlé India distributes most of its profit because it can't productively reinvest it, which is the See's signature.
• Dividend payout: very high, often ~70 to 100% of earnings, because the business gushes cash it has no capital-hungry use for.

Step 2. Owner earnings ≈ reported earnings. Reconcile it the way the owner-earnings section and the free-cash-flow module do. Maintenance capex runs roughly in line with modest depreciation, and working capital is negative instead of a growing drain. So owner earnings ≈ net income + D&A − maintenance capex ± ΔWC ≈ close to reported net income (~₹3,300 cr). Compare a steel maker, where owner earnings would be a fraction of reported profit because maintenance capex devours cash. Nestlé India's reported profit can actually be taken out, and that's how it can pay most of it out. A high conversion of reported earnings into owner earnings is the quantitative fingerprint of a See's-type franchise.

Step 3. Pricing power and economic goodwill. Through input-cost cycles (milk, coffee, cocoa, wheat, packaging), Nestlé India has raised prices and adjusted pack sizes while holding volume. Maggi's near-monopoly loyalty is the archetype. Even the 2015 recall crisis, when Maggi was pulled from shelves nationwide, ended with the brand recovering most of its share within a year. Few businesses could have survived that, and the recovery showed how strong the brand is. So Buffett's pricing-power test is passed. The economic goodwill, a multi-thousand-crore franchise on a small tangible base, dwarfs anything on the balance sheet and grows in nominal terms with each price increase.

Step 4. Run the four filters. Take them in order:

• Filter 1, understandable? Yes. It sells packaged foods under known brands, and you can forecast it a decade out with more confidence than almost any business. Pass.
• Filter 2, durable moat? Yes, an intangible brand plus distribution reach. Confirm it in the numbers, as the moat module taught: persistent very high ROCE, gross margin defended through input cycles, stable-to-rising share in core categories. Pass.
• Filter 3, able and honest management? Broadly yes: professional MNC management, clean accounting, consistent capital return. In fairness, and through the India governance lens from the ethics thread, there's one watch-item. Royalty and other payments to the Swiss parent are a recurring related-party flow that minority shareholders should size and monitor. Is it rising faster than sales, and is it fair? That doesn't disqualify the company, but it's a filter-3 question you must actually ask of any MNC subsidiary. Pass, with the royalty flagged.
• Filter 4, sensible price? For a value buyer, this is usually where it fails. Nestlé India almost always trades at a premium multiple, often 50 to 75× earnings, because the whole market can see filters 1 to 3. Run the reverse DCF and the price frequently implies many years of high growth and permanence, a wonderful business at a demanding price. On these figures the four filters say pass on quality, wait on price. That's a demonstration of the method. It isn't a view on the stock today. Put it on the watchlist and let Mr. Market's occasional bad day (a 2015 Maggi-crisis-style panic, a broad market crash) hand you the sensible price the business's quality won't.

The teaching point. Nestlé India shows the four filters working as designed: three emphatic passes and a perennial price problem. It also shows the deepest practical lesson of the quality school. Finding the wonderful business is rarely the hard part. The hard part is waiting for filter 4 to clear, and not overpaying because you fell in love with filters 1 to 3. That patience is the behavioural edge the later modules of Phase 6 build.

Worked Example 3: A Graham net-net screen, and the melting-ice-cube trap (synthetic)

Kaveri Components is synthetic, so we can control the numbers. Run Graham's purest method on it from end to end.

Step 1. Compute NCAV. From the latest balance sheet:

NCAV = 125 − 60 = ₹65 crore, or ₹65 cr ÷ 5 cr shares = ₹13.00 per share. Look at what you've thrown away. The entire ₹90 cr of fixed assets counts for nothing. That's Graham's deliberate brutality.

Step 2. Apply the two-thirds rule. Graham buys at ≤ ⅔ of NCAV:

• Threshold market cap = ⅔ × 65 = ₹43.3 crore, i.e. ≤ ₹8.67 per share.
• Suppose Mr. Market quotes ₹8.00/share, so the market cap = 8 × 5 = ₹40 crore.
• 40 ≤ 43.3, so it qualifies. You'd pay ₹40 cr for ₹65 cr of current assets net of all liabilities. You'd get the ₹90 cr of fixed assets and the entire operating business for free, plus a further ₹25 cr cushion below even the current-asset-net-of-debt figure. You don't have to estimate this margin of safety. It's sitting on the balance sheet.

Step 3. The melting-ice-cube check, the step beginners skip. NCAV is a snapshot, so ask whether the business is burning its net current assets. Take two scenarios:

• Scenario A, roughly break-even. The business loses a little or makes a little, and NCAV holds near ₹65 cr while you wait for a catalyst (re-rating, buyout, activist, liquidation). This is a real Graham bargain. Buy it, inside a diversified basket.
• Scenario B, burning ₹15 cr/year. Now the arithmetic turns against you. Next year NCAV ≈ 65 − 15 = ₹50 cr, the year after ≈ ₹35 cr, then ₹20 cr. Your ⅔-of-NCAV threshold falls with it (the ₹8.67 becomes ₹6.67, then ₹4.67), so the "40% discount" you bought evaporates roughly as fast as you earn it. A cheap price on a melting asset gives you a race between the discount and the burn, and the burn usually wins. Don't mistake that for a margin of safety. Your defence is to prefer break-even net-nets, size positions small, and, Graham's non-negotiable, hold a basket of 20 to 30, so the re-raters and buyouts outweigh the melters. A net-net is a statistical bet, never a conviction bet on one name.

Step 4. Why this is rare today. Run this screen on the broad US or large-cap Indian market now and you'll find almost nothing but Scenario-B melters, illiquid micro-caps, and governance disasters. The obvious Scenario-A bargains were arbitraged away decades ago. The method survives as a crash tool (2008, March 2020, Indian small-cap troughs, neglected Japanese/Korean small-caps) and as the pure expression of the margin-of-safety logic. That's why working it teaches the logic even if you rarely trade it.

Worked Example 4: Sizing the margin of safety to Phase-3 uncertainty (the computation)

Here's the margin-of-safety principle, a deeper discount for a wider valuation range, made numeric. Take two businesses, each with a base-case intrinsic value of ₹1,000/share from your DCF work, but with very different ranges.

Business P, a predictable compounder, is Nestlé-like. It has a wide moat and stable owner earnings. Here's your base/bull/bear from the scenario work in the valuation phase:

The range is narrow, with bear only 10% below base, because the cash flows are stable and the moat protects the spread's duration. The required margin of safety is a modest ~25%, since there's little room for value to be far below your estimate. Apply it to a conservative anchor:

• Off the base case: buy ≤ 1,000 × (1 − 0.25) = ₹750.
• Off the bear case, the more conservative choice: buy ≤ 900 × (1 − 0.20) = ₹720.
• Buy target ≈ ₹720 to 750. At that price even your bear case (₹900) leaves you a ~20 to 25% cushion, and the compounding of a growing intrinsic value, the dynamic margin of safety, adds more each year.

Business C, a cyclical, is a commodity or auto business with no durable moat, so its earnings swing with the cycle. Same ₹1,000 base, but the range is wide:

The bear is 45% below base, because a cyclical's value really could be that low and you have no moat protecting you. The required margin of safety is ~45 to 50%, and the anchor itself must be normalised to mid-cycle. Never use trailing-peak earnings, the cyclical trap from the end of the valuation phase. Apply it:

• Off the normalised base: buy ≤ 1,000 × (1 − 0.45) = ₹550. You demand a price down at your bear case before you'll act.
• Buy target ≈ ₹500 to 550.

The decisive comparison: why one number for both would ruin you. Suppose you used a lazy, uniform "20% discount" and bought both at ₹800:

• Business P at ₹800 is comfortably safe. It's below base (₹1,000) and below even bear (₹900), and the compounding does the rest. Fine.
• Business C at ₹800 is dangerous. You're above the safe buy zone for the normalised base value and far above the ₹550 trough. A normal downcycle takes the value to ₹550, and you sit on a 30%+ permanent-looking loss with no moat to grow you out of it. That breaks "don't lose money."

Same ₹800 price, same 20% headline "discount," opposite verdicts. The margin of safety has to be sized to the width of the range, and the width comes from how predictable the business is. This is the quantitative marriage of Graham's margin of safety to Phase 3's valuation ranges. Expect the mini-project and the mastery check to ask you for this computation.

Worked Example 5: The dynamic margin of safety: how a wonderful business grows into its price (₹, synthetic)

Graham's margin of safety is static, a discount to today's assets or earnings that doesn't change until you sell. The sections on the margin of safety and on the evolution claimed that Buffett's evolved margin of safety is dynamic. The discount widens on its own as a wonderful business compounds its intrinsic value past your fixed purchase price. Here's that claim as arithmetic, on a synthetic Indian franchise.

The setup. Vaishali Consumer Ltd (synthetic) is a predictable, wide-moat, capital-light franchise. It's the kind whose owner earnings grow steadily and whose bear case stays close to its base case. Your Phase-3 work puts intrinsic value at a conservative ₹1,000/share today. It reinvests a little at high returns and raises prices most years, so you judge its intrinsic value will compound at ~13% a year. On an ordinary day, Mr. Market quotes ₹920.

Step 1. The static verdict a Graham purist reaches. Margin of safety today = 1 − 920/1,000 = 8%. To a net-net Grahamite that's barely a margin of safety at all, nowhere near a two-thirds-of-NCAV discount, so he passes. Hold on to that verdict. It's about to look incomplete.

Step 2. Let the business compound. Intrinsic value grows at 13%/yr while your purchase price stays frozen at ₹920. Track the discount of that fixed ₹920 to each year's intrinsic value:

By year 5 intrinsic value has compounded to ₹1,842, and your unchanged ₹920 purchase is at a ~50% discount to value. The "thin" 8% cushion became a 50% cushion without your doing anything. It was the business that did the work.

Step 3. The contrast with a static cheap asset. Now buy a Graham net-net instead, at a fat 33% discount to a flat NCAV. If the business merely treads water, the discount stays 33% forever. That's safe, but it never grows. If it's a melting ice cube like Scenario B in the net-net example, the shrinking NCAV erodes the discount. Time is neutral to hostile for the static bargain and a tailwind for the compounder.

The teaching point. Here's the Graham-to-Buffett evolution in one table. The most durable margin of safety comes from a wonderful business whose intrinsic value grows past your price year after year. A cheap balance sheet can't supply it. That's why Munger's arithmetic lets a fair entry price suffice for a durable franchise. The compounding supplies the safety that a static discount would otherwise have to provide up front. One catch, and don't forget it: this works only if the 13% is real and durable, which is what filters 1 and 2 check. Pay a fair price for a business that doesn't compound and you have neither Graham's static margin nor Buffett's dynamic one. You just have a mistake.

Worked Example 6: The whole pipeline on one company: filters → owner earnings → intrinsic value → buy price ($, synthetic)

The mini-project asks you to run a business through all four filters and end at a buy price. This example walks the whole pipeline once, end to end, on a synthetic US franchise, so you have a template. Every number is arithmetic you can reproduce.

Ridgeline Brands Inc. (synthetic) is a US branded-consumables company (razors, filters, refills), a razor-and-blade model you can explain in one sentence.

Filter 1, understandable? In one line, it sells branded consumable refills that customers rebuy habitually, at a predictable cadence, in a category you understand. That's inside the circle of competence. Pass.

Filter 2, durable moat, proven in the numbers? NOPAT of $340m on $1,300m of invested capital = 26.2% ROIC. It has held above the ~9% US hurdle for a decade, with gross margin defended through an input-cost cycle, which is the pricing-power test. That's a real, durable moat. Pass.

Filter 3, able and honest management? Ten years of rational capital allocation (reinvest at high returns, return the rest), candid letters, aligned ownership, no accounting games. You verify all of it from the 10-K and DEF 14A, as the capital-allocation work taught. Pass.

Filter 4, sensible price? Now, and only now, compute value and compare it to price. Start with the unit of value, owner earnings:

Notice that only maintenance capex comes off. Ridgeline's growth capex is deliberately left out. It's optional investment, and the business doesn't need it to stay alive.

Intrinsic value comes next, using the definition from earlier: the DCF with owner earnings as the cash line. Treat owner earnings as growing at a sober 4% forever, discounted at the 9% hurdle (Gordon form):

• IV = 410 × 1.04 ÷ (0.09 − 0.04) = $8,528m. On 200m shares, that's $42.64/share.

Margin of safety and the buy price. Ridgeline is predictable, with a narrow range, so demand ~25% off the conservative estimate: buy ≤ 42.64 × (1 − 0.25) = $31.98 ≈ $32.

The decision. Mr. Market quotes $39. It's a wonderful business, with filters 1 to 3 all passing, but $39 > $32, so filter 4 fails because the price offers no margin of safety. Verdict: pass on price, watchlist. Don't let the quality of filters 1 to 3 answer the price question. That's the "quality at any price" trap. Months later Mr. Market has a bad day and quotes $30. Now the price clears your $32 target and the business is unchanged. That's the fat pitch. Same company, opposite action, decided entirely by filter 4.

The teaching point. The four filters are a pipeline with price last: three quality gates, then a valuation built on owner earnings, then a discount sized to predictability. The pipeline forces you to treat a great business and a great investment as different findings. Only the last filter tells them apart.

Practice set

Problems 1 to 3 are guided (the solution teaches the method), 4 to 9 are independent and 10 to 12 are timed at ≤6 minutes each. Companies are synthetic unless named. Where they're named, the figures are illustrative. It's the reasoning that counts. Write your answer in full before reading the solution.

Problem 1 (guided: Mr. Market). You value Sundar Textiles at a conservative ₹400/share. Over six months Mr. Market quotes, in order: ₹410, ₹520, ₹300, ₹250, ₹600. Your analysis of the business has not changed at any point. State the correct action at each quote and the single principle governing all five.

Solution. The principle: Mr. Market's price is an offer to accept, reject, or ignore, never an input to your ₹400 value estimate. Now take the five quotes in turn. At ₹410 it's roughly fair, so do nothing. At ₹520 it's ~30% above value, so sell (or trim), or ignore it, and certainly don't buy. At ₹300 it's 25% below value, which makes it interesting. If it clears your required margin of safety, buy. At ₹250 it's ~38% below value, so buy more. This is the fat pitch and no reason to doubt yourself, provided you re-check that no fact changed. That's the facts-versus-mood check. At ₹600 it's 50% above value, so sell. The amateur's tell is feeling best about the stock at ₹600 and worst at ₹250. The professional feels the opposite, because price moves inversely to prospective return.

Problem 2 (guided: the four filters, reject-fast). For each business, name the first filter it fails, or say that it passes all four, and stop there. (a) A pre-revenue quantum-computing startup with a brilliant story you can't evaluate. (b) A well-understood domestic airline with the lowest CASK in its market but no pricing power. (c) A wonderful, moaty consumer brand whose promoter routinely routes sales through a private related company at soft prices. (d) A wide-moat rating agency trading at 55× earnings whose reverse DCF implies 25 years of accelerating growth.

Solution. The filters run in series, so reject at the first failure. (a) Filter 1, understandable. You can't evaluate it. Stop and put it in the "too hard" pile. (b) Filter 2, durable moat. A cost edge in a structurally moat-less industry may make it tradeable on mid-cycle value, but it won't make it a four-filters compounder. It fails filter 2. (c) Filter 3, able and honest management. The related-party siphoning transfers your value to the promoter whatever the moat, so it fails filter 3. Pass at any price. (d) It passes 1 to 3 but fails filter 4, sensible price. A wonderful business at a demanding price earns the market return at best. Watchlist it and wait for Mr. Market. The skill the filters buy is knowing when to stop analysing.

Problem 3 (guided: net-net). Deccan Foundry (synthetic): current assets ₹180 cr (cash 30, receivables 60, inventory 90), total liabilities ₹100 cr, fixed assets ₹140 cr, 8 cr shares, market price ₹9.00. (a) Compute NCAV and NCAV/share. (b) Does it pass the two-thirds rule? (c) It's losing ₹20 cr/year. State your verdict.

Solution. (a) NCAV = 180 − 100 = ₹80 cr, and per share = 80/8 = ₹10.00, with the fixed assets of ₹140 cr counted as zero. (b) Threshold = ⅔ × 10 = ₹6.67/share. The market price of ₹9.00 > ₹6.67, so it fails the two-thirds rule. It trades at 90% of NCAV, well above the ≤67% Graham demands. Even before the cash-burn check, this isn't a Graham buy. (c) The ₹20 cr/year loss makes it a melting ice cube. NCAV falls to ~₹60 cr next year (₹7.50/share), so the stock at ₹9 would be trading above a shrinking NCAV, which is the opposite of a bargain. Verdict: pass, because it's too expensive relative to a melting NCAV. Had it been break-even and priced at, say, ₹6/share, it would qualify, inside a diversified basket.

Problem 4 (margin of safety: computation). You value Anand Compounders (predictable, wide-moat) at base ₹2,000, bull ₹2,300, bear ₹1,800. Your rule: for a range this narrow, demand a 25% margin of safety off the base case. (a) Compute the buy price. (b) At that price, what cushion do you have versus the bear case? (c) Mr. Market offers ₹1,700. Act?

Solution. (a) Buy ≤ 2,000 × (1 − 0.25) = ₹1,500. (b) Against the bear value of ₹1,800, at ₹1,500 you're at 1,500/1,800 − 1 = −16.7%. That's a ~17% cushion below even your bear estimate, plus the dynamic margin from a growing intrinsic value. (c) ₹1,700 is above your ₹1,500 target, so don't buy. It doesn't clear your required discount off base, and it leaves little room against the bear case. Sticking to the rule beats "close enough." The whole point of a pre-set required margin is that you don't renegotiate it with yourself when a nearly-good-enough price turns up.

Problem 5 (margin of safety: cyclical contrast). Girija Metals (cyclical, no moat): bull ₹2,600, base (mid-cycle normalised) ₹2,000, bear (trough) ₹1,100. Your rule: a wide range means a 45% margin of safety off the normalised base. (a) Buy price? (b) Compare it with Problem 4's ₹1,500 buy on the same ₹2,000 base and explain the difference in one sentence. (c) Why must the anchor be normalised to mid-cycle instead of trailing earnings?

Solution. (a) Buy ≤ 2,000 × (1 − 0.45) = ₹1,100. You demand the trough price before acting. (b) Same ₹2,000 base value, but the buy price is ₹1,100 vs ₹1,500, because the far wider valuation range of a moat-less cyclical forces a far deeper discount to protect against a value that could really be near the trough. (c) A cyclical's trailing earnings peak just when the stock looks cheapest on trailing P/E. That's the cyclical trap from the valuation phase, and the Tata Steel table in the moat module shows it. Anchor the margin of safety to peak earnings and you'd demand the least safety at the moment of most danger. Normalise to mid-cycle first, then discount.

Problem 6 (owner earnings vs the trap). Two businesses each report ₹500 cr net income. Alpha (software franchise): D&A ₹40 cr, maintenance capex ₹35 cr, ΔWC negligible. Beta (integrated steel): D&A ₹300 cr, but honest maintenance capex ₹450 cr (replacement at today's prices), ΔWC +₹50 cr. (a) Compute owner earnings for each. (b) Which is the See's-type business, and what does the comparison say about reported EPS?

Solution. Owner earnings = NI + D&A − maintenance capex ± ΔWC. (a) Alpha = 500 + 40 − 35 − 0 = ₹505 cr, about the same as reported and a hair above, as you'd expect from a capital-light business. Beta = 500 + 300 − 450 − 50 = ₹300 cr, only 60% of reported. Maintenance capex above depreciation devours cash, which is the inflation problem from the economic-goodwill discussion. (b) Alpha is the See's/Nestlé-type franchise, where reported profit ≈ owner earnings you can actually extract. Identical reported EPS can hide opposite economics. Beta has to reinvest most of its profit just to stand still, so its owner is far poorer than the ₹500 cr headline suggests. Reported earnings flatter capital-heavy businesses. Owner earnings tells the truth.

Problem 7 (economic goodwill / incremental returns). A franchise earns ₹80 cr pre-tax on ₹40 cr of tangible capital. Ten years ago it earned ₹20 cr pre-tax on ₹15 cr of tangible capital. (a) Return on tangible capital, then and now. (b) Incremental return on incremental capital over the decade. (c) What does this say about its economic goodwill, and would a Graham "1.5× book" rule have let you buy it?

Solution. (a) Then: 20/15 = 133%. Now: 80/40 = 200% pre-tax, a franchise that's getting stronger. (b) Incremental = (80 − 20)/(40 − 15) = 60/25 = 240%. Every extra rupee of capital generated ₹2.40 of annual pre-tax profit, the same pattern as See's in the first worked example. (c) It has large and growing economic goodwill, a franchise worth many multiples of its ₹40 cr tangible capital, none of it on the balance sheet. A Graham 1.5×-book rule would cap your purchase at ~₹60 cr and forbid buying it at its real worth of hundreds of crore. That's the "Graham limitation" Munger pushed Buffett past. Here the balance sheet is blind to the only asset that matters.

Problem 8 (the evolution: reasoning). Buffett calls the Berkshire textile business one of his costliest mistakes, yet he didn't lose money on the purchase. Explain the mistake in the language you've learned here, and state which of the three evolution forces it illustrates.

Solution. The mistake was opportunity cost. There was no loss on the purchase itself. Berkshire's textile mill was a cigar butt, bought cheap with one puff available. But Buffett then sank two decades of capital and attention into a structurally doomed business earning below its cost of capital, a commodity with no moat and no pricing power. That's cigar-butt limit 2: time is the enemy of the mediocre business. Every year he held it, the poor economics destroyed value that the same capital could have compounded in a wonderful business. It illustrates evolution force 2, the compounding arithmetic of durable high returns, by showing its inverse. A bad business is a recurring loss that a cheap entry price can't offset over time. Munger and See's were the fix.

Problem 9 (don't lose money: arithmetic). (a) A position falls 60%. What gain is needed to recover? (b) A position falls 90%. What gain to recover? (c) Use the result to explain why "avoid permanent loss" outranks "maximise upside," and connect it to the margin of safety.

Solution. (a) To recover from −60% you need 1/(1 − 0.60) − 1 = 1/0.40 − 1 = +150%. (b) From −90%: 1/0.10 − 1 = +900%. (c) Losses and gains are asymmetric. The deeper the loss, the disproportionately larger the gain you need, and a single near-total loss can erase decades of compounding. So survival, meaning avoiding permanent impairment, beats chasing upside, because one broken link breaks the whole chain. The margin of safety is the tool that prevents the permanent loss in the first place. You pay enough below conservative value that even being wrong, or an adverse future, leaves you whole. "Don't lose money" is the objective, and the margin of safety is the method.

Problem 10 (timed: 6 min, value vs price). Meridian Foods (wonderful, moaty) trades at 60× earnings. Your reverse DCF shows the price implies 18% growth for 20 years plus permanence. Its business quality is unquestionable. What's the correct posture, and which two "registers" must you keep separate?

Solution. Pass on price, and keep it on the watchlist. This is filter 4 at work. Keep two registers separate. One is business quality, filters 1 to 3: understandable, moaty, well-run, all yes. The other is price against value, which is filter 4. The reverse DCF says the price already capitalises 20 years of near-perfection, so the margin of safety is negative. A wonderful business at this price earns the market return at best, and a permanent loss if growth merely disappoints. Quality doesn't answer the price question. Munger said "fair price," not "any price." Wait for Mr. Market's bad day.

Problem 11 (timed: 5 min, defensive vs enterprising, and NCAV). (a) Classify: an investor who buys ten established, dividend-paying, financially strong companies at moderate multiples and rarely trades, defensive or enterprising? (b) An investor who computes NCAVs and buys a basket of sub-liquidation micro-caps? (c) Compute NCAV from current assets ₹220 cr and total liabilities ₹90 cr. State the two-thirds buy threshold as a market cap.

Solution. (a) Defensive. That's Graham's passive archetype: strength, moderation, low effort, diversified, left alone. (b) Enterprising. It's active and bargain-hunting, doing the work the defensive investor won't, and net-nets are enterprising territory. (c) NCAV = 220 − 90 = ₹130 cr, and the two-thirds threshold = ⅔ × 130 = ₹86.7 cr market cap. Buy only at or below that.

Problem 12 (timed: 6 min, synthesis). In one short paragraph each: (a) what did Buffett keep from Graham, and (b) what did he change, and why? Name the mechanism of the change.

Solution. (a) Kept: price ≠ value, the margin of safety as a principle, Mr. Market, and "don't lose money." That's the bottom row of the synthesis table. These are the permanent foundation, and Buffett applies them daily. (b) Changed: the source of the safety. It moved from Graham's static, balance-sheet cheapness (net-nets, cigar butts, low multiples) to the durability of a wonderful franchise whose intrinsic value compounds past the purchase price, the dynamic margin of safety. Mechanism of the change: three forces. Scale made cigar butts immaterial. The compounding arithmetic of durable high returns on capital rewards buying and holding great businesses over turning over cheap ones. And Munger's quality turn, proven by See's, supplied the conviction. Same religion, reformed liturgy.

Problem 13 (guided: owner earnings and the owner-earnings yield). Ostara Foods (synthetic): reported net income ₹450 cr, D&A ₹70 cr, honest maintenance capex ₹50 cr, required working-capital increase ₹10 cr, total capex (maintenance + growth) ₹140 cr, market cap ₹12,000 cr. (a) Compute owner earnings. (b) Compute the owner-earnings yield (owner earnings ÷ market cap). (c) What would using total capex have done, and why is that wrong?

Solution. (a) Owner earnings = NI + D&A − maintenance capex ± ΔWC = 450 + 70 − 50 − 10 = ₹460 cr. (b) Yield = 460 ÷ 12,000 = 3.83%. That's the honest earnings yield, the cash a rational owner could actually pull out per rupee of price. (c) Using total capex gives 450 + 70 − 140 − 10 = ₹370 cr. It wrongly subtracts the ₹90 cr of growth capex, which is optional investment, since standing still doesn't require it. That understates the franchise's true cash generation and penalises it for the very act of reinvesting in growth. Owner earnings charges only the maintenance capex needed to preserve current earning power. The growth capex belongs in the forecast you build in the DCF, and it shouldn't come off this year's extractable cash.

Problem 14 (guided: the dynamic margin of safety). Harita Compounders (synthetic) is a predictable wide-moat franchise. You buy at ₹600. Intrinsic value today is a conservative ₹750, and you judge it will compound at 11%/yr. (a) Find the initial margin of safety. (b) Find intrinsic value after 6 years. (c) Find the discount of your fixed ₹600 cost to that year-6 value. (d) State the principle and its one precondition.

Solution. (a) 1 − 600/750 = 20%. (b) 750 × 1.11⁶ = ₹1,402.8. (c) 1 − 600/1,402.8 = 57.2%. The discount nearly tripled from 20% without your lifting a finger. (d) This is the dynamic margin of safety. For a wonderful business the cushion widens over time as intrinsic value compounds past a fixed purchase price. It's the source of safety Buffett migrated to, and the reason a fair price can suffice for a durable compounder. The precondition is that the 11% must be real and durable. Filters 1 and 2 test that. A static cheap asset's discount doesn't grow, and a melting one's shrinks.

Problem 15 (net-net and the melting-ice-cube race). Bellandur Castings (synthetic): current assets ₹300 cr (cash 60, receivables 100, inventory 140), total liabilities ₹150 cr, fixed assets ₹200 cr, 10 cr shares, market price ₹8.50. The business is burning ₹18 cr of net current assets per year. (a) NCAV and NCAV/share. (b) Does it pass the two-thirds rule today? (c) If the burn continues, in roughly how many years does NCAV fall below the market cap you paid, and what is the verdict?

Solution. (a) NCAV = 300 − 150 = ₹150 cr, and per share = 150/10 = ₹15.00, with the ₹200 cr of fixed assets counted at zero. (b) Two-thirds threshold = ⅔ × 15 = ₹10.00/share. The market price of ₹8.50 ≤ ₹10.00, so it passes (market cap ₹85 cr ≤ ₹100 cr threshold). (c) You paid an ₹85 cr market cap. NCAV starts at ₹150 cr and burns ₹18 cr/yr, so it crosses below ₹85 cr when 150 − 18t < 85, i.e. t ≈ 3.6 years (year-3 NCAV ₹96 cr, year-4 ₹78 cr). Verdict: the arithmetic bargain is real today, but it's a race between your catalyst and the burn. It's acceptable only inside a diversified basket with a near-term catalyst (re-rating, buyout, liquidation), and never as a lone concentrated bet. A melting ice cube can pass the two-thirds rule and still ruin you if the puff is late.

Problem 16 (intrinsic value from owner earnings, and the buy price; $). Fairmont Filtration Inc. (synthetic, US): owner earnings $180m, discounted at a 9% hurdle with 3% perpetual growth, and 120m shares. It's a predictable business, so you demand a 30% margin of safety. (a) Intrinsic value per share (capitalise owner earnings, Gordon form). (b) Your buy price. (c) Mr. Market offers $24. Act?

Solution. (a) IV = 180 × 1.03 ÷ (0.09 − 0.03) = 185.4 ÷ 0.06 = $3,090m, and per share = 3,090/120 = $25.75. Intrinsic value is the discounted owner-earnings stream. (b) Buy ≤ 25.75 × (1 − 0.30) = $18.03. (c) $24 > $18.03, so don't buy. A wonderful, correctly valued business is still a pass on price when the quote is above your required discount. The margin of safety is a pre-set rule, and you don't renegotiate it because a price is "close enough." Watchlist it and wait for Mr. Market's bad day. That's filter 4 doing its job.

Problem 17 (float and the cost of float). Konkan General Insurance (synthetic): premiums written ₹4,000 cr, claims + expenses ₹3,880 cr, average float ₹9,000 cr invested at 8%. (a) Combined ratio. (b) Underwriting profit or loss. (c) Cost of the float as a % of float, where a negative figure means the insurer is paid to hold it. (d) The total the float engine contributes, and why this is "leverage that can be safe."

Solution. (a) Combined ratio = 3,880/4,000 = 97.0%. (b) Underwriting profit = 4,000 − 3,880 = +₹120 cr. A CR below 100% means an underwriting profit. (c) Cost of float = −120/9,000 = −1.33%. The insurer is paid 1.33% a year to hold ₹9,000 cr of other people's money. (d) Investment income = 8% × 9,000 = ₹720 cr. Add the ₹120 cr underwriting profit and you get ₹840 cr, an effective 9.33% on capital that cost less than nothing and that no counterparty can call in a panic. That non-recallable, negative-cost quality is the "leverage that can be safe." It's safe only because underwriting discipline keeps the combined ratio at or below 100%. Let it drift above 100% and the same float becomes expensive leverage in disguise.

Problem 18 (timed: 8 minutes, all five parts, synthesis). (i) Current assets ₹500 cr, total liabilities ₹200 cr, 20 cr shares. Find NCAV/share and the two-thirds buy price. (ii) Net income ₹900 cr, D&A ₹100 cr, maintenance capex ₹160 cr, WC increase ₹40 cr. Find owner earnings. (iii) A holding falls 84%: gain needed to break even? (iv) A compounder is worth ₹500 today and growing 14%/yr. Find intrinsic value in 5 years, and the margin of safety on a ₹400 purchase measured against it. (v) An insurer's combined ratio is 102%: is its float costing it money or paying it, and which discipline failed?

Solution. (i) NCAV = 500 − 200 = ₹300 cr, and /20 = ₹15.00/share. Two-thirds = ₹10.00/share. (ii) OE = 900 + 100 − 160 − 40 = ₹800 cr. (iii) 1/(1 − 0.84) − 1 = 1/0.16 − 1 = +525%. That's the drawdown asymmetry: an 84% loss needs a 525% gain, so avoiding permanent loss outranks chasing upside. (iv) IV = 500 × 1.14⁵ = ₹962.7, and the margin of safety on ₹400 = 1 − 400/962.7 = 58.5%, the dynamic margin of safety again. (v) CR 102% > 100%, so it's an underwriting loss and the float costs ~2% of premiums. Underwriting discipline failed, through writing underpriced policies to grow float, and safe leverage turned into expensive leverage. Scoring: 5/5 in time = fluent. ≤3 = re-work Worked Examples 5 and 6 and Problems 13 to 17 before the mastery check.

Problem 19 (guided: the seven defensive criteria, run on a listed company). Screen ITC Ltd against all seven of Graham's defensive-investor criteria. The figures are illustrative, rounded to teaching scale from the published annual report and exchange data. Pull the real ones from the annual report and the exchange before you rely on any of them, and re-check them, because every line here is perishable. Revenue about ₹70,000 crore. Current ratio 2.4. Long-term borrowings effectively nil, against working capital of about ₹15,000 crore. Positive net profit in each of the last ten years, and dividends paid without interruption for far longer than twenty. Earnings per share, taken as a three-year average at each end of the decade, went from ₹8.20 to ₹14.00. Trailing earnings per share ₹16.40, book value per share ₹60.00, market price ₹420. (a) Mark each of the seven criteria pass or fail, showing the arithmetic where there is any. (b) State the verdict a literal Graham screen returns. (c) Say what the verdict does and does not tell you about the business, and which criterion is doing all the work.

Solution. (a) Criterion by criterion, in Graham's order.

• Adequate size: revenue of about ₹70,000 crore is among the largest listed non-financials in India. Pass.
• Strong financial condition: current ratio 2.4 against the required 2.0, and long-term debt of effectively nil against ₹15,000 crore of working capital. Pass on both halves.
• Earnings stability: positive earnings in each of the last ten years. Pass.
• Dividend record: uninterrupted for well beyond twenty years. Pass.
• Earnings growth: 14.00 ÷ 8.20 − 1 = +70.73% on three-year averages, against the required one-third. Pass, comfortably.
• Moderate price-to-earnings: 420 ÷ 14.00 = 30.0× three-year average earnings, against Graham's 15× ceiling. Fail, by a factor of two.
• Moderate price-to-book: 420 ÷ 60.00 = 7.00× against the 1.5× ceiling, and the combined test gives 30.0 × 7.00 = 210 against a cap of 22.5. Fail on both readings, the second by more than nine times.

(b) Five of seven pass, and a literal defensive screen rejects the company, because criteria 6 and 7 are absolute gates. You can't trade them off against the other five. (c) The verdict tells you about the price. It says nothing against the business. On every test of size, safety, consistency and growth, this is just the kind of company Graham was pointing his defensive investor at, and the whole rejection comes from the two valuation criteria. That's the lesson of the defensive-investor section made concrete. Graham's numeric thresholds were calibrated to 1970s US markets and, applied literally, exclude nearly every quality franchise the rest of the course teaches you to admire. Keep the architecture: a defensive buyer wants established, financially strong, consistently profitable, dividend-paying companies at undemanding multiples. Don't import the 1.5× and 15× as constants.

Problem 20 (timed: 6 min, the combined test and the Graham number; US, $). Kestrel Fasteners Inc. (synthetic) trades at $48.00. Three-year average earnings per share $3.60, trailing earnings per share $4.00, book value per share $22.00. (a) Compute the price-to-earnings ratio on three-year average earnings, the price-to-book ratio, and their product; state whether the combined test passes. (b) Compute the Graham number, the square root of 22.5 times trailing earnings per share times book value per share, and compare it with the price. (c) Explain why the two tests disagree about how far the price is from the ceiling, and say which input choice drives the gap.

Solution. (a) P/E = 48.00 ÷ 3.60 = 13.3333×, which passes the 15× criterion on its own. P/B = 48.00 ÷ 22.00 = 2.1818×, which fails the 1.5× criterion on its own. The product is 13.3333 × 2.1818 = 29.0909, against the 22.5 cap, so the combined test fails by 6.5909 points. Notice what the combined test bought you. The standalone price-to-earnings pass was misleading, because the earnings come bundled with a book value that doesn't support them. (b) Graham number = sqrt(22.5 × 4.00 × 22.00) = sqrt(1,980) = $44.4972. The price of $48.00 is 48.00 ÷ 44.4972 − 1 = 7.87% above it, so this test also fails, though only just. (c) The Graham number is the same 22.5 rule solved for price, so the two tests are one test with different inputs. The combined test above used three-year average earnings of $3.60, and the Graham number used trailing earnings of $4.00. Solve the 22.5 rule for price on the three-year average and the ceiling is sqrt(22.5 × 3.60 × 22.00) = $42.2137, which is $2.2834 below the Graham number computed on trailing earnings. The gap is entirely the earnings input, and the smoothed figure is the more conservative one. That's why criterion 6 specifies a three-year average instead of the latest year. A Graham number quoted off a peak trailing year is the cyclical trap of Problem 5 wearing a formula.

Applied mini-project: The four filters and a margin of safety, on a real company

The task (≈3 hours). Choose one company you really understand. Filter 1 is a gate, so don't pick a business you can't explain. Each of these is a plausible quality franchise. In India: Pidilite (Fevicol), Asian Paints, Nestlé India, Titan, HDFC Bank. In the US and globally: the disclosed operating businesses of See's parent Berkshire, Coca-Cola, Moody's, Costco, Visa. Then run the complete Graham-to-Buffett judgement:

1. Filter 1, understand it. Write the business in one line (what it does, how it makes money) and say plainly whether it's inside your circle of competence. If it isn't, pick another company. This is the gate.
2. Filter 2, prove the moat in the numbers. Bring in the method from the moat module. Name the source or sources, then show ≥5 years of persistent ROIC/ROCE above the hurdle (~12% INR / ~8 to 9% USD, mid-2026), gross-margin stability through an input-cost cycle (the pricing-power test), and stable or rising share. Compute at least three years yourself from the filings.
3. Filter 3, judge management. Work from the annual report and proxy (DEF 14A) or the Board's Report, plus India's related-party disclosures. Grade the last five years of capital allocation across the five uses. Check candour: do they own their mistakes? For any promoter-controlled or MNC-subsidiary company, size the related-party and royalty flows and say whether minority holders are being treated fairly.
4. Compute owner earnings: net income + D&A − maintenance capex ± ΔWC, with your maintenance-capex estimate stated and defended (depreciation proxy vs unit math). Note where it falls relative to reported profit and to FCF, and what that says about capital intensity.
5. Intrinsic value range and margin of safety. Build a base/bull/bear intrinsic-value range and judge its width. Size your required margin of safety to that width, using the method from the fourth worked example, then compute your buy price and compare it to today's market price. End with a pre-registered falsifier: "I would abandon this thesis if ___ (a measurable number) crossed ___."

Deliverable: a one-page memo with the business in one line, the four-filters verdict (pass or fail for each, with the decisive evidence), owner earnings, the intrinsic-value range, the required margin of safety and buy price against market price, and the falsifier. File it in your knowledge system. It's a rehearsal for the Phase 8 thesis.

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

Case Lab: run case XL-6.1 (The Four Filters and the Fair Price, a synthetic franchise) in the app's Cases tab for this module. You pass a company through the four filters in series, rejecting fast at the first failure. Then you build its owner earnings, size a margin of safety to its base/bull/bear range, and deliver a buy-or-watchlist verdict against the key. It's the mini-project in miniature, and a warm-up for the Phase 8 teardown.

Reading & resources

• Benjamin Graham, The Intelligent Investor (Jason Zweig revised edition, 2003). The core text here. Chapter 8 (Mr. Market, read twice), Chapter 20 (Margin of Safety, the most important chapter in investing), Chapters 14–15 (defensive and enterprising stock selection). Zweig's modern commentary after each chapter is worth as much as the original. [Paid] [Beginner–Intermediate]
• Benjamin Graham & David Dodd, Security Analysis (1934/2008 6th ed.). The rigorous parent of The Intelligent Investor; for the work here, the net-current-asset-value method and the margin-of-safety foundations. Read selectively; it is dense. [Paid] [Advanced]
• Warren Buffett, Berkshire Hathaway shareholder letters, free at berkshirehathaway.com/letters, and the best free investing education in existence. For the ideas here, pull: 1983 (the appendix on economic vs accounting goodwill), 1986 (the appendix defining owner earnings), 1987–1989 (moats, mistakes, "wonderful business at a fair price"), 1992 (intrinsic value defined), and 2007 ("Businesses – The Great, the Good and the Gross", the See's numbers used in Worked Example 1). [Free] [Beginner–Intermediate]
• Lawrence Cunningham, The Essays of Warren Buffett (arranged by theme): the letters reorganised into topics (value, margin of safety, owner earnings, governance). It's the most efficient way to absorb Buffett's thinking as a system instead of in date order. [Paid] [Intermediate]
• Charlie Munger, Poor Charlie's Almanack: the source of the quality turn; for the quality turn itself, the talks where Munger explains the shift from cheap to wonderful and "the big money is in the waiting." (M6.03 mines it further for the latticework.) [Paid] [Intermediate]
• Roger Lowenstein, Buffett: The Making of an American Capitalist and Alice Schroeder, The Snowball: biographies that narrate the Graham→Buffett evolution and the See's hinge from the inside; excellent for making the abstractions stick. [Paid] [Beginner]
• Robert Hagstrom, The Warren Buffett Way: a clean secondary synthesis of the four filters, owner earnings, and intrinsic value if you want the framework before the primary sources. [Paid] [Beginner]
• Aswath Damodaran, "Value Investing" material and the intrinsic-value lectures (his NYU site / YouTube): the bridge from this philosophy to Phase 3's mechanics; his evenhanded critique of value investing is valuable ballast. [Free] [Intermediate]
• Sanjay Bakshi, "Behavioural Finance & Business Valuation" (BFBV) lecture notes and blog (Fundoo Professor): the finest India-contextualized teaching of Graham/Buffett value investing, with Indian franchise and net-net examples; a Phase 6 staple. [Free] [Intermediate]
• Filings & data to pull (all free): Berkshire 2007, 1986, and 1983 letters (the See's economics, owner earnings, and goodwill appendices). The Nestlé India annual report and the screener.in ten-year page, for Worked Example 2's franchise economics and the royalty note. One Indian post-crash micro-cap screen on screener.in filtered for price < NCAV, to see how few real net-nets survive. [Free] [Intermediate]
• Going deeper: Michael Mauboussin & Dan Callahan, Measuring the Moat (Morgan Stanley / Counterpoint Global, free PDF), a rigorous framework for judging filter 2's durability in the numbers (industry structure, value creation, the sustainability of the ROIC−WACC spread), extending M4.04 into the language demanded here. Optional now; the natural companion when you run the mini-project's moat test. [Free] [Intermediate]
• Going deeper: Joel Greenblatt, The Little Book That Still Beats the Market. The "magic formula" (rank on cheapness × return on capital) reads as a systematic, screenable fusion of the Graham arithmetic and the Buffett quality instinct you fix here, and it is the mechanical bridge into M6.02's Greenblatt section. Short, plain, and worth an afternoon. [Paid] [Beginner]

The Modern Analyst's Addendum

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

AI-Augment this skill

{
  "skill": "The price-value distinction, the margin of safety sized to a valuation range's width, NCAV/net-net screening with the melting-ice-cube check, the four filters run in series, owner earnings, economic goodwill, and static vs dynamic margin of safety",
  "use": "Two narrow, genuinely useful jobs, and one absolute ban. (1) EXTRACTION: turn a 10-K's or annual report's raw tables into the exact line items NCAV needs (cash, receivables, inventory, other current assets, total current liabilities, total long-term liabilities) and owner earnings needs (net income, D&A, capex, working-capital change), as a table carrying the page or note reference for each figure — the same task a research associate does, and just as checkable. (2) THE ADVERSARY: draft the strongest available case that a name your screen returned is a melting ice cube (WE3), or that a four-filters 'pass' is really a filter-3 failure hiding in the related-party note (§8) — a list of things to go verify, never a verdict. THE BAN: never ask it to judge whether a business is wonderful, has a durable moat, or is run by honest management — filters 1 through 3 are exactly the judgment this module exists to teach you to make from primary sources, and a fluent AI-written moat narrative reads identically whether the moat is real or not, which is exactly why it cannot stand in for the numbers test.",
  "tools": ["Chat assistants — Claude, ChatGPT for structured extraction from a filing and for drafting the adversarial bear case", "A spreadsheet or notebook — where NCAV, owner earnings, and the margin-of-safety sizing function actually get computed and re-keyed against source", "screener.in's custom-screen builder and the BSE/NSE filing archives, and SEC EDGAR's companyfacts XBRL API and full-text search — the primary documents every extraction must trace back to"],
  "prompt": "Here is the balance sheet extract from <company>'s FY<year> annual report / 10-K: <paste>. Return a table with the exact line-item label, the figure, and the page or note it comes from, for: cash and equivalents, receivables, inventory, other current assets, total current liabilities, total long-term liabilities, net income, depreciation and amortisation, capital expenditure, and the year-on-year change in working capital. Do not compute NCAV, owner earnings, or any ratio, and do not offer a view on whether the company is cheap — extraction and citation only.",
  "verify": "Re-key every extracted figure against the page or note it cites before it enters your NCAV or owner-earnings formula — an extraction slip is invisible downstream, and Graham's whole method is arithmetic on inputs, so a wrong input produces a wrong verdict with no warning attached. Treat the adversarial bear case as a checklist of claims to go verify in the filing yourself (the related-party note, three years of operating cash flow, the customer-concentration disclosure) — never as a finding, because an AI-drafted bear case is exactly as fluent when it is wrong as when it is right.",
  "diy": "The gate is unaided: you compute NCAV from a balance sheet and apply the two-thirds rule by hand, you compute owner earnings and can defend your maintenance-capex estimate, you run the four filters in series on a real business and state which one it fails and why, and you size a margin of safety to a valuation range's width — all with a calculator and no tool in the room."
}

Modern Data Analysis

By hand first. You computed NCAV to the crore from a balance sheet and applied the two-thirds rule. You told a real bargain from a melting ice cube by its cash-burn trend, sized a margin of safety to the width of a base/bull/bear range, and walked the full filters-to-owner-earnings-to-buy-price pipeline on Ridgeline Brands. None of that changes. A screen that returns "cheap" is worthless if you can't re-derive, by hand, why.

Today's workflow. Graham's method is a screen before it's a decision, and these days you run the screen instead of eyeballing it. Write NCAV as a function over a table of balance-sheet lines, and steer clear of a black-box "value score." Join it against a trailing operating-cash-flow series, so the melting-ice-cube check from the net-net example and Problem 15 becomes a computed flag and stops being a habit you might forget. Then write the margin-of-safety sizing rule from the fourth worked example as a function of the one number that actually drives it, the width of your base/bull/bear range.

Tools & sources (IN + US). Use pandas for the screen and the cash-flow join, and numpy for the sizing function. In India, screener.in's custom-screen query builder runs a saved formula such as Market Capitalization < 0.67 * (Current Assets - Total Liabilities) directly against their database. The BSE Corporate Filings archive holds the underlying annual reports, and NSE's filing portal has the shareholding and related-party disclosures a filter-3 check needs. In the US, SEC EDGAR's companyfacts XBRL API carries the CurrentAssets/Liabilities tags that feed NCAV, and EDGAR full-text search (efts.sec.gov) finds the maintenance-capex and related-party language a filter-3 check requires.

# NCAV screen + melting-ice-cube flag + margin-of-safety sizing, as functions -- R3, recomputed in-session
import pandas as pd
universe = pd.DataFrame({
    "name":  ["Kaveri Components", "Bellandur Castings", "Deccan Foundry"],
    "cash":  [40, 60, 30], "receivables": [35,100,60], "inventory": [50,140,90],
    "cur_liab": [45,150,45], "lt_liab": [15,0,55],
    "shares": [5,10,8], "price": [8.00,8.50,9.00],
    "ocf_trend_3yr": [0.5, -18.0, -20.0],   # cr/yr, 3-yr avg change in operating cash flow
})
universe["ncav"] = (universe.cash+universe.receivables+universe.inventory
                     - universe.cur_liab - universe.lt_liab)
universe["ncav_ps"] = universe.ncav/universe.shares
universe["two_thirds_ps"] = (2/3)*universe.ncav_ps
universe["qualifies"] = universe.price <= universe.two_thirds_ps
universe["melting"] = universe.ocf_trend_3yr < 0
print(universe[["name","ncav_ps","two_thirds_ps","qualifies","melting"]].to_string(index=False))
# Kaveri: 13.00 / 8.67 / True / False   Bellandur: 15.00 / 10.00 / True / True   Deccan: 10.00 / 6.67 / False / True

def margin_of_safety_buy(base, bear, mos_floor=0.20, mos_ceiling=0.50):
    """SS3's rule as code: size the discount to the range's width, applied to the conservative anchor."""
    width = (base-bear)/base
    mos = min(mos_ceiling, max(mos_floor, width))
    return base*(1-mos), mos
print(margin_of_safety_buy(1000,900))   # (800.0, 0.20)
print(margin_of_safety_buy(1000,550))   # (550.0, 0.45)

Verify. Run the screen on the three names used here and it reproduces every verdict worked by hand. Kaveri Components, from the net-net example, qualifies (₹13.00 NCAV/share against an ₹8.67 threshold, priced at ₹8.00) and isn't melting, with a +₹0.5 cr/yr cash trend. Bellandur Castings (Problem 15) also qualifies on price (₹8.50 ≤ ₹10.00), but now the melting flag fires (−₹18 cr/yr), at the very point where the hand-worked answer calls it "a race between your catalyst and the burn." Deccan Foundry (Problem 3) fails to qualify at all (₹9.00 against a ₹6.67 threshold) before the melt check even matters. The sizing function reproduces Business C's ₹550 buy price to the rupee, since its 45%-wide range falls inside the ceiling band. For Business P, though, it returns ₹800, where the fourth worked example hand-picked ₹750. A 10%-wide range falls at the floor of Graham's stated 20 to 30% band, and the 25% chosen earlier was a judgement made within that band, one the qualitative rule alone doesn't determine. That gap is worth keeping, so don't patch it away. A screener can enforce a band, but the specific point inside it is still yours to defend.

Quantitative lens

The net-net example and Problem 15 tell you to judge a net-net basket "as a group... the statistics work, not the single name." That's Graham's own justification for diversifying across 20 to 30+ names instead of concentrating. The claim rests on historical net-net track records, and a historical track record is just the kind of number a database can get wrong without anyone noticing. That's because the names that failed hardest are the ones most likely to have vanished from the record you're reading. So quantify the mechanism on synthetic data built to isolate it, the same way the net-net example builds Kaveri Components to isolate the melting-ice-cube logic.

How survivorship bias inflates a net-net basket, in numbers. Build a 30-name basket, Graham's own "20–30+". Give each name an independent 20% chance of melting to a total loss over the holding period (delisting, bankruptcy, a wipeout the two-thirds rule didn't protect you from). If it survives, it earns a return drawn from a distribution centred on +25%, the re-rating or buyout outcome that the net-net example calls a real Graham bargain. Run it 200,000 times.

• The true basket return, with melters counted at −100% just as a real investor experienced them, averages +0.02% (200,000-basket simulation). The closed form gives 0.20×(−100%) + 0.80×(+25%) = 0.0% on the nose.
• The survivorship-biased "apparent" return is +25.0%. That's what an incomplete database computes when it drops the melted names and averages what's left, and dropping the melters is mathematically identical to pretending they never existed.
• The gap is 25.0 percentage points, and it's no fluke of one draw. Across 20,000 independently drawn baskets the mean gap is 25.0pp (std 9.3pp), and the survivorship-biased figure overstates the true one in 99.9% of them.

The size of the bias scales with how often names melt. At a 5% melt rate it's 6.2pp, and at 40% it's 50pp. So the worse the melting-ice-cube problem is in a given era or market, the larger the invisible inflation in any backtest that doesn't explicitly track delistings. And this isn't hypothetical. Tyler Shumway's "The Delisting Bias in CRSP Data" (Journal of Finance, 1997) found that the standard US academic return database of the era was missing the return for most performance-related delistings. He proposed a −30% proxy because omitting them, the "apparent" case above, systematically overstated historical returns. It's the same mechanism this simulation isolates, found in the dataset that much of the US factor-investing literature was originally built on.

Now the limits. The 20% melt rate and the +25% survivor mean are constructed to make the mechanism vivid. Nobody measured them from a real net-net universe. India's own record would need its own study before this became a claim about actual returns. For now it demonstrates the shape of a bias. On screener.in, the live screens show only currently listed companies, and the BSE delisted-companies archive is the only place the melters remain visible. The simulation also assumes melting is independent across names, which the net-net example's basket logic already leans on. In the next module, the hidden-correlation warning gives that assumption its own scrutiny. Here's what the simulation does establish cleanly. "A diversified basket's statistics work" is a claim about a basket that includes its failures. A historical net-net track record that doesn't name its data source's delisting treatment hasn't yet earned the right to be believed.

Do it in code: never trust a backtest, yours or a published one, until you've asked what happened to the names no longer in the database. Price that gap the way the simulation above does before you extrapolate a historical net-net record into a return you expect to earn. See G1/QM1.01, G2/QD1.03 and G2/DA1.02.

Where this goes next: galaxy cross-links

• G1 · QM1.01 Probability, Distributions, Sampling & Estimation: the law-of-large-numbers logic behind "a diversified basket's statistics work even when one name doesn't," and why a 30-name sample is not yet large enough to make survivorship bias average out on its own.
• G2 · QD1.03 Backtesting Engine: where survivorship-bias-aware, point-in-time data (a database that keeps delisted names on the books) turns the survivorship warning here into standing infrastructure instead of a one-off worry.
• G2 · DA1.02 pandas: Data Wrangling & Joins: the join between a balance-sheet table and a cash-flow trend that makes the melting-ice-cube check a computed column instead of a habit you might skip.

Teach it back & journal

Feynman prompt. Explain all four of these ideas to a smart 15-year-old with no jargon, using a single running example: a lemonade stand you might buy a share of from a moody friend. (1) Why the price your friend names each day, sometimes cheap and sometimes crazy-high depending on his mood, isn't the same as what a share of the stand is really worth. (2) Why you'd only buy when his price is well below your careful estimate of worth, and how much below depends on how sure you are (a boring, steady stand vs. a fad). (3) Why a wonderful stand you hold for years beats a cheap-but-bad one you have to keep flipping. (4) Why never taking a loss so big you can't recover matters more than hitting a home run. Keep it to one page. If you catch yourself writing "intrinsic value" or "margin of safety," stop and say it in kid words instead ("what it's really worth" and "buy it cheap enough that even if you're a bit wrong, you're okay").

Journal reflection. The hardest habit in value investing is separating a mood swing from a fact change when a price drops. Think back to a time, in markets or in any decision, when a falling "price" (a stock, a plan, a relationship, a project) made you doubt your own judgement. Did the underlying facts actually change, or only the sentiment around them? Write down how you'd tell the two apart next time, before the emotion hits. Then turn it into a line for the checklist you'll finish at the end of the phase: "When a price I own drops sharply, before I act I will re-check ___ (the specific facts of my thesis) and only sell if ___ changed, not because the quote did." Finally, note which of the two masters here, austere Graham or patient Buffett and Munger, matches your temperament better, and why that self-knowledge matters for the strategy you'll actually be able to stick to.

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