Special Situations + Capstone

Phase 3 · Week 35 · ~20 hours

Where it sits. The last module of the valuation phase, and the M4 milestone of the whole program. You first built the theory: discount rates, cash flows, terminal values, multiples, value drivers. Then you built the machinery, meaning the linked 3-statement model, the DCF, the comps table and the LBO. Every one of those tools quietly assumed the same patient: a going concern with definable free cash flow, roughly stable economics, and a single business you can see whole. Now come the patients who violate those assumptions. Banks, where FCFF is not even definable. Cyclicals, where the latest year is a lie in both directions. Money-losing young companies, where there is nothing to extrapolate and some probability of death. Option-like assets, where flexibility itself is the value. Conglomerates and holding companies, where the sum has parts and India charges a famous discount. Distressed firms, where the waterfall, not the DCF, decides who gets paid. It closes with Damodaran's discipline for keeping stories and spreadsheets honest with each other, and then the Phase 3 capstone: a full three-way valuation (DCF + comps + reverse DCF) of one Indian and one US company, each delivered as a two-page memo and graded against the rubric in Section 8. On the five-questions spine (from the master map), the question here is number 5, what is the price implying, and is that beatable?, asked of the companies where the standard machine answers wrongly, and then certified on companies of your own choosing.

Suggested hour budget. Core content: 7h · worked examples: 1.5h · practice set: 3h · capstone project: 7h · flashcards + mastery check: 1.5h.

App link. The special-situations wing of the valuation sandbox runs on this material: the bank excess-return builder (book-value path, spread, fade, each graded per component), the mid-cycle normalizer (10-year cyclical histories where you must locate mid-cycle earnings before the P/E is revealed), the survival-haircut slider (watch value respond to failure probability and distress proceeds), the SOTP assembler (segment multiples, central costs, minority interest; forget a line and the check fails), the IBC waterfall puzzle (distribute an estate under Section 53), and the capstone itself: the case engine's memo grader scores your two capstone memos against the rubric below.

Learning objectives

By the end you can:

1. Triage any company in two minutes into "standard DCF/comps patient" versus one of six special situations, then name which assumption of the standard machine it breaks and which replacement architecture to use.
2. Value a lender: explain from the balance sheet up why FCFF and WACC fail for banks; build a dividend-discount valuation with the regulatory-capital payout constraint (payout = 1 − g/ROE); build a multi-stage excess-return model (value = book + PV of (ROE − r) × book); and read any bank's P/B against its ROE without a model.
3. Normalize a cyclical to mid-cycle three ways (through-cycle average ROE on current book, mid-cycle unit margin × current capacity, EV per unit of capacity vs replacement cost), and demonstrate numerically why cyclicals look cheapest on P/E at exactly the wrong moment.
4. Value a high-growth, unprofitable company with the Dark Side architecture (top-down TAM → share → take-rate/margin path → sales-to-capital reinvestment → declining discount rate → survival-probability haircut), and run the young-company checklist before trusting any of it.
5. Recognize when real options genuinely add value (exclusivity, resolvable uncertainty, capacity to act), price a simple one with a one-period binomial, and refuse the "strategic value" abuse that rescues bad DCFs.
6. Build a sum-of-the-parts valuation that remembers the lines amateurs forget (central costs, minority interest, taxes on stakes), and apply India's holding-company-discount evidence (roughly 30–70%) with its causes and the catalysts that occasionally close it.
7. Outline India's IBC/NCLT process from default to resolution or liquidation, run a Section 53 recovery waterfall to the rupee, and explain why the equity of a distressed firm is a call option, which is what makes "it fell 95%, how much lower can it go?" the most expensive sentence in investing.
8. Translate a narrative into numbers and back (story → 3P test → drivers → valuation → falsifiers), and pass the Phase 3 capstone: three-way valuations and two-page memos on one Indian and one US company, with every assumption inside a defensible band, graded ≥80% on the Section 8 rubric.

Prerequisites & connections

Builds on. Everything in Phase 3, deliberately. M3.03's cost of capital is the r in the excess-return model and the WACC-glide in young-company valuation. M3.04's g = reinvestment rate × ROIC and sales-to-capital method drive the Dark Side forecast. M3.05's terminal-value constraints (g ≤ economy growth; steady-state ROIC discipline) and reverse DCF are one-third of the capstone. M3.06's justified multiples reappear here as the bank-valuation shortcut, P/B = (ROE − g)/(r − g) above all, and its comps discipline is the second third. M3.07's ROIC-vs-WACC master test explains why each special situation needs special handling (banks lever a small spread; cyclicals' ROIC oscillates through WACC; young companies burn capital ahead of returns). M3.08–M3.09's model stack is the chassis the capstone runs on. From earlier phases: M2.03 (building NOPAT and invested capital), M2.04 (cash conversion), M2.06–M2.08 (quality of earnings, since a bank valuation is only as good as your GNPA scepticism), M1.08 (leases, which carry the retail SOTP segments), and M1.09–M1.10 (consolidation, minority interest, the SOTP bridge).

Feeds into. Phase 4's moat work (a durable spread is what the excess-return model capitalizes). Phase 5's sector playbooks: the banking/NBFC and insurance modules deepen every financials tool sketched here, and the commodities and infra playbooks industrialize mid-cycle normalization. Phase 8's rapid-analysis system uses the two-minute triage below as step zero of every teardown. Competency C5 ("value a business three ways") is certified by this module's capstone; competency C11 (communicate like a top consultant) is graded on the capstone memos.

Core teaching content

4.1 The triage: what the standard machine assumes, and who breaks it

The Phase 3 machine you now own makes four quiet assumptions:

1. Free cash flow is definable. You can separate operating from financing activity, so FCFF means something.
2. The recent past is a usable base. Last year's margin and this year's revenue are sane starting points.
3. The firm will survive. Discounting expected cash flows assumes there are cash flows in every scenario.
4. The firm is one thing: one growth rate, one risk profile, one multiple.

Each special situation breaks exactly one of these, and the fix follows from the break:

Memorize the table. In Phase 8's one-hour teardown, this triage gets 120 seconds and one question: which row am I in? Pick the wrong row and you've made a big error. It's the difference between valuing HDFC Bank on EV/EBITDA, which means nothing for a bank, and on P/B against ROE, the industry standard. Or between buying a steel stock at 5× peak earnings and at 0.5× replacement cost. The first price is a trap and the second is an opportunity.

One more connection before the rows. The M&A and LBO thinking you've already done frames several of these situations. A SOTP is implicitly a break-up analysis. It asks what an acquirer could sell the pieces for, and precedent transactions from relative valuation price that scenario. An LBO value is a floor under a stable-cash-flow business, because it's the price a financial buyer could pay and still earn their hurdle. And India's IBC is, functionally, a court-run distressed-M&A auction. Keep those lenses in your pocket as we go.

4.2 Financials: valuing banks when FCFF refuses to exist

Why the standard machine fails

Try to compute FCFF for a bank and watch each ingredient dissolve:

• For a bank, interest is cost of goods sold. A bank buys money (deposits, borrowings) at one rate and sells money (loans, investments) at a higher one, so net interest income is the gross profit line. FCFF requires you to strip interest out to reach "operating" cash flow. Do that to a bank and you've deleted the business.
• Debt is raw material. For Tata Motors, borrowing is a choice about how to finance the machine. For HDFC Bank, deposits are the machine. FCFF is defined by separating operating from financing, and WACC discounts it, but neither means anything when the liability side is the inventory.
• Capex and working capital can't be defined. A bank's "reinvestment" is regulatory capital, retained to support loan growth. Book equity is the true invested capital, and it's the regulator who sets how much of it each rupee of lending consumes. The CFO doesn't get a say.

So you value the equity directly, discounting equity cash flows at the cost of equity r. Never EV, never WACC, never EV/EBITDA, because a bank has no meaningful EBITDA. You get three tools, in rising order of power: the dividend-discount variant, the excess-return model, and the P/B-vs-ROE shortcut that compresses both.

The bank P&L in six lines (the vocabulary)

Interest income − interest expense = net interest income (NII), summarized by NIM (net interest margin = NII / average interest-earning assets). Add fee and treasury income and you have total income. Subtract operating costs, tracked as the cost-to-income ratio, and you reach pre-provision operating profit (PPOP). Take off provisions for bad loans (the credit cost, as % of average loans) and you're at PBT, then tax, then PAT. You read asset quality in GNPA/NNPA (gross/net non-performing assets, % of loans) and PCR, the provision coverage ratio. Profitability compresses to ROA (PAT / average assets, ~1.0 to 2.0% for good Indian banks), which the equity multiplier (assets/equity, roughly 8 to 12× for banks) levers into ROE: ROE = ROA × leverage. Phase 5's banking module drills these. Here you need them to feed the valuation.

The regulatory-capital constraint (why banks can't pay out what they like)

Basel III, in the RBI's version for Indian banks, requires common equity tier 1 (CET1) of at least 5.5% of risk-weighted assets plus a 2.5% capital conservation buffer. That comes to 8% CET1 and ~11.5% total capital. Systemically important banks carry add-ons on top, and SBI, HDFC Bank and ICICI each pay a D-SIB surcharge. Globally the CET1 floor is 4.5% + 2.5% buffer. These are the mid-2026 requirements, so verify them against the current RBI Master Circular before you use them. The valuation consequence fits on one line:

To grow risk-weighted assets at rate g while holding capital ratios steady, a bank must retain g/ROE of its earnings. So the sustainable payout is:

payout = 1 − g/ROE

Take a bank growing loans 15% at 15% ROE. It can pay out nothing, because every rupee of profit is conscripted into capital. Slow the same bank to 5% growth and it can pay out two-thirds. Growth isn't free for a bank. It eats the dividend that a DDM discounts. This one constraint is what stops a bank DDM from discounting dividends the bank could never pay.

Tool 1: the DDM variant

Value per share = PV of expected dividends, with dividends forced through the constraint above. For a stable bank: V₀ = E₁ × (1 − g/ROE) / (r − g), where E₁ is next year's earnings, g the sustainable growth, r the cost of equity. Multi-stage versions let a fast-growing bank keep its payout low now and open the taps at maturity. Worked Example P3 in the practice set builds one in full.

Tool 2: the excess-return model (the professional's frame)

Rearrange the same algebra around book value and it becomes wiser:

Value of equity = Book value today + Σ PV of excess returns, where excess return in year t = (ROEₜ − r) × Bookₜ₋₁

Read it aloud: a bank is worth its book, plus the present value of everything it earns above its cost of equity on that book, minus everything it earns below. The model forces the three questions that actually decide bank value:

1. Is the book real? This is asset quality. A rupee of overstated loans is a rupee of fictitious book, so your forensic training on earnings games becomes a valuation input here: GNPA recognition, provision coverage, restructured-book games, evergreening.
2. What spread (ROE − r) can the franchise earn, and for how long? Think deposit franchise, underwriting culture, cost control. It's the moat question in banking clothes.
3. How fast can the book compound at that spread? Growth is g = retention × ROE, bounded by the capital constraint and by the economy's credit growth.

Here's a coherence check. Take a synthetic stable bank with book value per share B₀ = ₹100, ROE 16%, r = 13%, g = 5% forever. Retention is g/ROE = 31.25%, so the payout is 68.75%.

• DDM route: E₁ = 16% × 100 = ₹16. D₁ = 16 × 0.6875 = ₹11. V₀ = 11/(0.13 − 0.05) = ₹137.5.
• Excess-return route: excess return in year 1 = (16% − 13%) × 100 = ₹3, growing 5% forever. PV = 3/0.08 = 37.5, so V₀ = 100 + 37.5 = ₹137.5.
• P/B shortcut: P/B = (ROE − g)/(r − g) = (16 − 5)/(13 − 5) = 1.375×, and 1.375 × 100 = ₹137.5.

Same number three ways. It's one model wearing three outfits, and the equivalence holds whenever accounting is "clean surplus", meaning book grows by earnings minus dividends. Nothing in that algebra is specific to a bank. Under the name residual income, the same model values any company off its own book equity. The residual-income section below builds it for a non-financial company. It adds the multistage version and the accounting adjustments the model needs there. The multi-stage bank version, a high spread for some years and then a fade, is Worked Example 1, computed line by line.

Tool 3: reading P/B against ROE (the two-minute version)

You met the justified multiple in relative valuation: P/B = (ROE − g)/(r − g). Anchors to burn in:

• ROE = r ⇒ P/B = 1. A bank that earns its cost of equity to the decimal is worth book to the rupee, and growth changes nothing.
• P/B < 1 is the market saying ROE will run below r, or that stated book is fiction because the assets are overstated. Treat it as a verdict to investigate, never as a discount coupon. Many Indian PSU banks spent 2016 to 2020 at 0.3 to 0.6× book while GNPAs were being recognized. The market was marking both the spread and the book, and at the time it was right to.
• High P/B is a duration claim. Take an HDFC Bank-shaped franchise (illustrative, ~FY25 scale, so pull the actual filing): book ≈ ₹5 lakh crore, ROE ≈ 15 to 17%, priced around 2.3 to 2.5× book (mid-2026). One-stage math at r = 12.5% and g = 5% justifies only (16 − 5)/(12.5 − 5) ≈ 1.5×. So what justifies ~2.4×? Run the multi-stage arithmetic. At r = 12%, a bank compounding book at 12% for 20 years at ~18% ROE, then fading to a 1.3× terminal P/B, is worth ≈ 1 + 20 × (0.06/1.12) + 0.29 ≈ 2.4× book. The price is a claim that the next two decades will look like the best of the last two. That may well be right, since it's what the franchise has done. The excess-return frame turns a mystique ("great bank, deserves premium") into a sentence you can test: spread × growth × duration.
• Now the US contrast, which comes down to rates and has nothing to do with patriotism. A JPMorgan-shaped bank (illustrative, mid-2026) earns ~17% on tangible equity and is priced at ~2.3 to 2.6× tangible book. At the US cost of equity, r ≈ 9.5% with g = 4%, the one-stage justified P/TB is (17 − 4)/(9.5 − 4) ≈ 2.4×. That's near fair without any heroic duration assumption. The multiple that demands twenty exceptional years in INR is routine in USD. Discount rates drive the gap, and quality doesn't, which is the deepest lesson in comparing financials across borders.

The sector regression from relative valuation now gets teeth. Plot P/B (y) against ROE (x) for the 10 to 15 listed Indian banks and fit the line. A bank far below the line is cheap relative to how the market prices ROE today. That makes it a research question about the one thing the regression can't see, which is whether the book and the ROE are real. You answer that question about asset quality in the filings. The scatter plot can't help you.

Where does all this stop? Insurers get their own machinery (embedded value, VNB margins) in Phase 5's financials playbook. NBFCs use these same tools with a harder funding-fragility overlay, because they lack deposit franchises. IL&FS and DHFL, from the forensic case files, are the cautionary tales. The frame carries across all of them. Value the equity directly, as book plus spread × duration, and ask whether the book is real.

4.10 The declining firm: valuing a business that is shrinking on purpose

Every model so far has assumed growth, or at worst no growth. Some businesses shrink, and they shrink for structural reasons that won't reverse: print directories, wireline telephony, physical media, cheque processing, tobacco volumes in most developed markets, coal-fired generation in some. The standard machine handles them badly, in a specific and predictable direction, and the errors run both ways.

Start with reinvestment, which goes negative and becomes a cash inflow. In a growing firm, revenue growth consumes capital: reinvestment = ΔRevenue ÷ sales-to-capital. Run the same identity with a falling revenue line and the sign flips. Receivables shrink, inventory shrinks, plant is not replaced as it wears out, and each of those releases cash. A declining firm's free cash flow is therefore higher than its earnings, often much higher, and for a while it can be a formidable cash machine. Model the release explicitly. Setting capex equal to depreciation out of habit understates value on a shrinking business.

Divestiture belongs inside that same cash flow. Managements of declining businesses sell things: a regional operation, a brand, a building. Treat announced and probable divestitures as what they are, an accelerated version of the same capital release. Value them at what a buyer would pay, and leave book value out of it. Then net out the earnings that leave with the asset. That's the step where most sum-of-the-parts arithmetic on declining firms goes wrong.

Then comes the central move. Make the terminal value a wind-down, and drop the perpetuity. A Gordon terminal value assumes the business exists for ever, which is the very assumption the whole analysis has just rejected. A negative growth rate doesn't repair it: FCFF/(WACC + 2%) still runs to infinity, it just gets there more slowly. For a firm in structural decline, the terminal value should be one of two things. One is a trade sale at a defensible multiple of the terminal year's earnings, to a consolidator who can strip cost out. The other is a wind-down: collect the receivables, sell the inventory and the equipment, pay the costs of closing, and distribute what's left.

Underneath both of those is the liquidation floor: what the assets fetch in a forced sale, less what's owed. Compute it, always, and compute it separately from the going-concern value, because it answers a different question. The going-concern value tells you what the run-off is worth. The floor tells you how much you lose if the run-off is worse than you think. For most declining firms the floor is far below the going-concern value. Read that as a warning, because the case rests entirely on the cash coming out. The whole answer depends on the discount rate and the decline rate.

So there are two ways to be wrong here, and they're opposites. A declining business is a perfectly good investment at a price that lets the run-off cash return your money with a margin. It's a value trap at any price that requires the decline to stop. The trap is seductive because the multiple is always low. A low multiple on a shrinking earnings base is just arithmetic, though. Arithmetic isn't a bargain. Forget the multiple. Ask whether the present value of the cash the business will actually return, on your decline rate and not management's, exceeds the price. Then ask whether you'd still be solvent on the outcome one notch worse than that.

Worked Example 7 below runs a full declining-firm valuation, including the contrast with the perpetuity that a careless model would have used, and Practice P22 asks you to do it solo on an Indian case.

4.11 The special-situations shelf, assembled

Here's the triage table from 4.1, now with the tools filled in. It's the shelf you carry into the capstone and into Phase 8:

Common mistakes & how experts think differently

1. Valuing a bank with the tools that happen to be open in Excel. The FCFF template exists, so it gets used. Out come EV/EBITDA comps for lenders (no meaningful EBITDA), WACC for a business whose debt is inventory, and "capex-light!" conclusions about balance sheets that reinvest through regulatory capital. Experts choose the model before opening the file: equity-direct, book-anchored, spread-driven.

2. Trusting a bank's stated book because the model needs it. The excess-return model is only as good as B₀. A bank hiding 3% of loans as "standard restructured" has overstated the very anchor your valuation stands on. Experts spend more time on GNPA recognition, provision coverage, and related-party lending than on the discount rate. For a lender, forensics is valuation.

3. The peak-P/E purchase (and its mirror, the trough-P/E sneer). Screens surface cyclicals at 5× earnings right at the top and hide them at 60× right at the bottom. Experts flip the ratio's meaning for cyclicals. They grow suspicious of optical cheapness at record margins and curious about optical expensiveness at losses. Then they re-anchor on P/B, EV/capacity vs replacement cost, and normalized earnings.

4. TAM intoxication. "1% of a $500bn market" assumes share is free and the TAM is the firm's to monetize. Experts build share from capacity and retention, take the lower of top-down and bottom-up TAM, and check the take-rate slice. Then they still haircut for survival. A brilliant destination reached by only 80% of the probability mass is worth 80% plus salvage.

5. Double-counting flexibility. Analysts add "option value" on top of a DCF whose explicit forecast already contains the expansion. Or they apply a survival haircut and a punitive 20% discount rate for the same risk. Experts give each risk one home, and only one: cash flows, discount rate, survival probability, or option.

6. The SOTP of best moments. Every segment goes in at its peak multiple. There are no central costs, no minority interest, no tax on stake exits. Then comes a demand that the market pay 100% of this fantasy NAV immediately. Experts value segments mid-cycle and subtract every forgotten line. Then they make the senior move and analyze the discount instead of denying it: causes, catalyst, carry while waiting.

7. "It fell 95%, how much lower can it go?" The answer is 100%, and via the waterfall it usually gets there. Anchoring on the former price, or on the book equity of a firm in CIRP, ignores that equity is rung 8 of 8. Experts price the queue and ignore the chart. They read a distressed stock's price as an option premium on volatility. It isn't an estimate of asset value.

8. Story-price backfitting. You decide the multiple is justified and reverse-engineer a story to bless it: 4,000 Zudios because the price needs them. Experts run the loop in one direction only, from story to numbers to value, and only then compare with the price. They publish falsifiers before outcomes, and they let the reverse DCF speak for the market and keep their own hopes out of it.

9. Averaging the three-way triangle. DCF says 620, comps say 780, so "call it 700." But the disagreement between methods is information to be explained, and averaging throws it away. Comps capture market context the DCF ignores, and the DCF captures economics the comps assume away. Experts write the sentence that reconciles them ("the sector trades a full rate-regime above intrinsic; my DCF is right on economics and the comps are right that I won't get paid soon"), and only then choose a number. The capstone rubric grades that sentence by name.

Worked examples

Worked Example 1: Excess-return valuation of a bank, line by line (synthetic: Vaigai Bank)

Setup (₹ crore). Book equity B₀ = 20,000, on 100 crore shares, so BVPS is ₹200. Cost of equity r = 13%. In stage 1 (years 1 to 5) ROE is 17% and book grows at g = 12%, so retention = g/ROE = 12/17 = 70.6% and the payout is 29.4%. In stage 2 (year 6 on) the franchise matures. ROE fades to 15% and growth to 5% (retention 33.3%), leaving a spread (ROE − r) of 2% forever.

Step 1, the book value path (grows at 12%):

Sum of PV(excess returns, years 1 to 5) = 3,477.7. See how little the PVs shrink? Book compounds at 12% against a 13% discount rate, so a high-growth bank holds its excess-return value remarkably flat in PV terms.

Step 2, terminal excess returns. B₅ = 31,470.4 × 1.12 = 35,246.8. Year-6 excess return = 2% × 35,246.8 = 704.9, growing 5% forever: TV₅ = 704.9 / (0.13 − 0.05) = 8,811.7, and PV = 8,811.7 / 1.8424 = 4,782.4.

Step 3, assemble. Value = B₀ + PV(stage-1 excess) + PV(terminal excess) = 20,000 + 3,477.7 + 4,782.4 = ₹28,260 crore, which is ₹282.6/share, P/B = 1.41×.

Step 4, the DDM cross-check (prove the equivalence once, believe it forever). Dividends in stage 1 are 29.4% of earnings: 1,000.0, 1,120.0, 1,254.4, 1,404.9, 1,573.5. Their PV = 884.9 + 877.1 + 869.4 + 861.7 + 854.0 = 4,347.2. Now get the equity value at year 5 by Gordon. E₆ = 15% × 35,246.8 = 5,287.0, payout = 1 − 5/15 = 66.7%, so D₆ = 3,524.7 and V₅ = 3,524.7/0.08 = 44,058.5. That's B₅ + TV₅ = 35,246.8 + 8,811.7, the same object. PV = 44,058.5/1.8424 = 23,912.9. Total = 4,347.2 + 23,912.9 = ₹28,260 crore. It's identical, as clean-surplus accounting guarantees.

Step 5, use it. If Vaigai trades at ₹240 (1.2× book), the market prices a shorter or thinner spread than your five 17%-ROE years, so check asset quality before celebrating. At ₹420 (2.1× book), the market extends the 4-point spread well past year 5, and the burden of proof becomes a two-decade franchise claim. The model's gift is that every price becomes a sentence about spread × duration.

Worked Example 2: A cyclical, normalized: the full peak-to-trough autopsy (synthetic: Deccan Ispat)

Setup (₹ crore except per-share). A steelmaker with current capacity of 12 mt, up from 8 mt a decade ago. It has 150 crore shares, current BVPS of ₹260, a 10-year average ROE of 11% and a cost of equity of ~13%. Replacement cost of integrated capacity ≈ ₹5,800 crore per mt. Here are ten years of EPS:

Step 1, mid-cycle EPS, three ways.

• Raw average: 296/10 = ₹29.6, but capacity grew 50% over the window, so this blends today's company with a smaller one.
• Average ROE × current book: 11% × 260 = ₹28.6, size-corrected.
• Per-tonne rebuild: mid-cycle EBITDA/tonne ₹8,500 (from the industry cost curve and this firm's own 10-year record) × 12 mt = 10,200. Less D&A 3,000 and interest 1,200 gives PBT 6,000, then PAT (25% tax) 4,500, so EPS ₹30.0.

The three converge: mid-cycle EPS ≈ ₹29 to 30.

Step 2, the peak re-examined. In year 9 the stock traded at ₹504. Optical P/E = 504/84 = 6.0×, which read as "cheap!" Normalized P/E = 504/29.5 ≈ 17×. That's dear for a business whose mid-cycle ROE (11%) is below its cost of equity (13%). Now the asset check. EV = 504 × 150 + net debt 6,000 (paid down in the boom) = 81,600, which is ₹6,800 crore/mt, or 117% of replacement cost. The market was paying more per tonne than a greenfield, an open invitation to new supply. What happened next? Year-10 EPS printed 30 and the stock fell to ₹280, down 44%, while its trailing P/E rose to 9.3×. The 6× buyer bought the crest.

Step 3, the trough re-examined. In year 5 the EPS was −6, the P/E meaningless and the headlines funereal, and the stock touched ₹95. That's P/B ≈ 0.4×. EV = 95 × 150 + net debt 18,000 = 32,250, which is ₹2,688 crore/mt, or 46% of replacement cost. Normalized P/E = 95/29.5 ≈ 3.2×. Every asset-anchored signal said "cycle bottom pricing." The earnings-anchored signal said nothing at all, and that asymmetry is what you should carry away.

Step 4, mid-cycle fair value. Mid-cycle EBITDA 10,200 × mid-cycle EV/EBITDA 5.0× (a commodity converter, with ROIC ≈ WACC across the cycle) = 51,000. Take off mid-cycle net debt of 12,000 and you have 39,000, or ≈ ₹260/share. That's 1.0× book, and it hangs together, because a business earning roughly its cost of capital is worth roughly its capital. With ROE a shade under r, a defensible range is ₹230 to 260. Against that anchor, ₹504 was ~2× fair value dressed as a 6× P/E. And ₹95 was ~0.4× fair value dressed as a bankruptcy headline.

Worked Example 3: A money-losing platform with a survival haircut (synthetic: Zyppin, quick commerce)

Setup (₹ crore). In year 0, category gross order value (GOV) is ≈ 65,000 and growing ~20%/yr. Zyppin holds ~18% share (GOV 12,000), with revenue of 2,160 at an 18% take-rate and EBITDA of −900. It has net cash of 3,500 after its IPO and no debt. There are 220 crore shares at a price of ₹95, so the market cap is 20,900.

Step 1, the destination (year 10), conservative base. Category GOV: 65,000 × 1.20¹⁰ ≈ 4,00,000. Zyppin slips to a solid #3 with 12% share, so its GOV is 48,000. The take-rate matures to 21%, giving revenue of 10,080. EBIT is 3.5% of GOV = 1,680, anchored on mature marketplace economics, and NOPAT (25%) = 1,260.

Step 2, continuing value at year 10. Mature inputs are g = 5% and ROIC 25%, so the reinvestment rate = 20%. WACC is 11%. CV₁₀ = 1,260 × 0.8 / (0.11 − 0.05) = 16,800.

Step 3, the interim path and the rate glide. FCFF runs −800, −550, −300, −50, +200, +450, +700, +950, +1,150, +1,010. The burn narrows, breakeven comes around year 4 or 5, and then cash builds. Discount at 14% for years 1 to 5, gliding to 11% by year 10 (13.4/12.8/12.2/11.6/11.0). Cumulative discount factors: 1.140, 1.300, 1.482, 1.689, 1.925, 2.183, 2.463, 2.763, 3.084, 3.423. PV of interim FCFF = −701.8 − 423.2 − 202.5 − 29.6 + 103.9 + 206.1 + 284.2 + 343.8 + 372.9 + 295.1 = +248.9. PV of CV = 16,800/3.423 = 4,907.8.

Step 4, going-concern value. Operating value is 248.9 + 4,907.8 = 5,156.7. Add net cash of 3,500 and you get 8,656.7.

Step 5, the survival haircut. Zyppin is listed, funded (~4 years of runway) and contribution-positive in its core city cohort, so P(survival) = 90%. Distress value ≈ 2,000, from a failure scenario, which is a fire-sale of brand/tech plus residual cash after wind-down. E(V) = 0.90 × 8,656.7 + 0.10 × 2,000 = 7,791 + 200 = ₹7,991 crore → ₹36.3/share. How sensitive is that to p, the input people pretend is 1.0? At p = 1.0 you get ₹39.4, at p = 0.75 ₹31.8, and at p = 0.50 ₹24.2.

Step 6, confront the price. The market pays ₹95. Reverse the architecture. To justify ~20,900 of market cap you need roughly 30 to 35% terminal share at ~4.5% EBIT/GOV with near-certain survival. In other words, Zyppin doesn't slip to #3. It becomes co-leader of a category that keeps compounding. That's a possible story. Run the 3P test: possible, yes. Plausible, so check the cohort data. Probable is where it gets hard, because the base rates of #3 players becoming co-leaders are thin. So you don't write "overvalued, avoid." You write the sentence "the price buys the co-leader story," plus the falsifiers you'll watch: monthly share data, take-rate trend, contribution margin per order, and the burn against runway.

Worked Example 4: Sum-of-the-parts with every forgotten line (synthetic: Bharat Diversified Ltd)

Setup (₹ crore). Four parts: cement (EBITDA 2,400, 75% owned, with the subsidiary carrying 4,000 of the group's debt), specialty chemicals (EBITDA 1,500), value retail (sales 6,000, EBITDA-thin growth phase), and a 32% non-control stake in a listed NBFC (market value of stake 8,000). Unallocated corporate costs 330/yr pre-tax (≈250 post-tax). Parent + wholly-owned net debt is 12,000, and the cement sub's 4,000 is included in it. There are 50 crore shares, priced at ₹505.

Step 1, value each part with its own lens (mid-cycle multiples from pure-play comps):

Step 2, the bridge (where SOTPs go to die). Less net debt 12,000. Less minority interest. Outsiders own 25% of the cement business, so 25% × (26,400 − 4,000 allocated debt) = 25% × 22,400 = 5,600. Equity NAV = 53,700 − 12,000 − 5,600 = 36,100 → ₹722/share.

Step 3, the discount measured then interrogated. Price ₹505 against NAV ₹722 is a 30% discount, mid-range for an Indian operating conglomerate. Now the stress test. Re-run at bear multiples (cement 9×, chemicals 7×, retail 1.0× sales). Gross EV = 21,600 + 10,500 + 6,000 + 6,800 − 2,000 = 42,900. Minority = 25% × (21,600 − 4,000) = 4,400. NAV = 42,900 − 12,000 − 4,400 = 26,500, which is ₹530/share ≈ the market price. So the market is marking every segment at bear-case multiples at the same time. That's a coherent pessimism. If there's a thesis, it can't be "the discount must close." It has to be that (a) governance is clean (RPT note, promoter pledge, the forensic checks), (b) a catalyst exists (the announced cement demerger would list the largest part separately), and (c) you're paid to wait (dividend yield ~2%). Without (b), 30% discounts in India are just weather, and calling them mispricing is a mistake.

Worked Example 5: Assembling the football field and writing the range verdict (synthetic: Cascade Freight Corp)

Setup (US $m). A synthetic US freight/logistics operator with 300m diluted shares, net debt of $4,000 and EBITDA of $5,000. The current price is $110, so the market cap is $33,000 and the EV at the tape is $37,000. You've already built the three-statement model and the comps table. This example is the step no one teaches: turning three methods into one picture and one sentence. The bridge convention throughout is equity = EV − net debt and per share = equity ÷ 300.

Step 1, the DCF rail (three driver sets, three EVs, three prices). Your base/bull/bear are toggled driver sets, straight out of the narrative loop. They aren't ±10% nudges on a cell:

The DCF says intrinsic value plausibly spans $90 to $133, centered on $110.

Step 2, the comps rail. Six freight peers trade at 6.5× to 8.0× forward EV/EBITDA, normalized for leases and one-offs the way relative valuation taught you. On this firm's $5,000 EBITDA, the low end gives 6.5 × 5,000 = 32,500 EV, so equity is 28,500, or $95.00 a share. The high end gives 8.0 × 5,000 = 40,000 EV, so equity is 36,000, or $120.00. The comps band is $95 to $120.

Step 3, the transaction/LBO rail (the floor). Recent freight deals and a sponsor LBO test both come out near 7.0× EBITDA = 35,000 EV. That's equity of 31,000, or $103.33, the price a strategic or a financial buyer could pay and still clear their hurdle. Read it as a soft floor. Below it, the whole company is cheaper than a private owner would rationally pay.

Step 4, the reverse-DCF marker (what the tape already assumes). Take a steady-state Gordon that reproduces the $37,000 EV, given normalized FCFF₁ ≈ $2,000 and WACC 8.5%. It needs only g = WACC − FCFF₁/EV = 0.085 − 2,000/37,000 = ~3.1% perpetual FCFF growth. That's undemanding, and well below the ~5% the bull case assumes. The price isn't betting on a heroic future.

Step 5, assemble the field and read it. Put every rail on one per-share axis (the memo's Exhibit 1):

Overlaid, the rails tell one story. Intrinsic value clusters at $95 to $120, the LBO floor is at $103, and the tape ($110) falls inside every band. The reverse DCF confirms that the price banks on only ~3% FCFF growth.

The verdict (the sentence the field exists to produce). "Intrinsic and transaction evidence cluster at $95 to $120. At $110 the tape equals my DCF base, sits mid-comps, and stands only ~$7 above the LBO floor: fairly priced, no margin of safety. The price underwrites just ~3% FCFF growth, so the downside is well supported. That makes this a quality-at-fair-value name, and it isn't a bargain. Action: pass at $110. The field turns clearly favorable below ~$100, the bear-to-base zone and the LBO floor, where I would buy near what a private owner would pay. I don't average the rails into a false-precision '$106'. The DCF and the transaction market agree on value, and the comps merely confirm the sector isn't distorting the read."

Worked Example 6: A pure holding company: NAV, the discount, and what you are paid to wait (synthetic: Sahyadri Holdings)

Setup (₹ crore). A synthetic Indian pure holdco that owns stakes and does little else: 45% of listed Konkan Consumer (market cap ₹80,000), 26% of listed Sahyadri Finance (market cap ₹1,20,000), a wholly-owned unlisted estates arm (EBITDA ₹500, peers at 8× EV/EBITDA, no debt), and net cash of ₹1,800. The holdco's own market cap is ₹29,200. Unlike the operating conglomerate in Worked Example 4, nothing here is run centrally. It's a portfolio wearing a ticker.

Step 1, NAV. Take the stakes at market and add the other assets. There's nothing operating to add.

Step 2, the discount. discount = 1 − market cap ÷ NAV = 1 − 29,200/73,000 = 60%, the deep end of India's persistent 30 to 70% pure-holdco band. At a 60% discount the market pays ₹29,200 for ₹73,000 of marked assets: ₹40 for every ₹100.

Step 3, why the ₹100 is not really yours (don't skip to "buy"). Here are the five causes, made concrete. The operating companies' cash reaches you only as the trickle the promoter declares. The stakes are control instruments that will realistically never be sold, so there's no terminal event. A hypothetical sale is taxed at two gates: first dividend receipt, then capital gains. So the NAV is pre-tax fiction unless distribution is intended. The float is thin. And no one can force the gap shut. The ₹73,000 is an accounting truth, and nobody has promised you its cash.

Step 4, what you are actually paid to hold it. Read it two ways:

• Carry while waiting. The holdco pays ₹700/yr in dividends to its own shareholders, so the yield = 700/29,200 = 2.4%. That's the cash you actually receive. Everything else is hope for a catalyst.
• Look-through earnings. Your economic share of the underlying profits is 45% × ₹3,000 (Konkan PAT) + 26% × ₹8,000 (Finance PAT) + ₹250 (estates) = ₹3,680. That makes the look-through P/E = 29,200/3,680 = 7.9×, against the opcos' own ~27× and ~15×. You buy their earnings at roughly a third of the direct price, and the Step-3 frictions are the reason you can.

Step 5, the conditional verdict. "NAV ₹73,000, price ₹29,200, 60% discount, carry 2.4%. This becomes a thesis only with a named catalyst: a demerger, a holdco buyback, a family settlement, a real change in distribution policy. If such a catalyst narrowed the discount from 60% to 40% (NAV unchanged), the re-rating alone is 73,000 × (1 − 0.40) ÷ 29,200 − 1 = +50%. Without one, I'm a minority holder collecting 2.4% while the discount stays weather. Size the position to the catalyst's probability and ignore the discount's width, because 'it's at 60%!' is the most expensive sentence a holdco investor can say."

Worked Example 7: A business in structural decline, valued as a run-off (synthetic: Fairmont Print Media, US $m)

The specimen: Fairmont Print Media Inc (synthetic), a US printed-directory and local-print business. All figures are synthetic. Revenue is $1,200m and falling 8% a year. The EBIT margin is 14% this year and erodes one point a year as scale leaves. Tax is 25%. Sales-to-capital is 2.0, so invested capital opens at 1,200/2.0 = $600m. WACC is 9.5%, net debt is $120m at a 7% pre-tax coupon, and there are 40m diluted shares. The stock trades at $10.00. Management's deck forecasts "stabilisation by year 3." Your job is to value the run-off without adopting anyone's stabilisation.

Step 1: the engine, with reinvestment running backwards. Each year revenue falls 8%, and NOPAT = revenue × margin × 0.75. The capital released is −ΔRevenue ÷ 2.0, which is a cash inflow. FCFF = NOPAT + released capital.

Sum of the explicit present values: $498.97m. Look at the FCFF column against the NOPAT column before going on. In every year the free cash flow exceeds after-tax operating profit by 41 to 58 percent, purely because the balance sheet is emptying. Set capex equal to depreciation here, as a habit carried over from a growth model, and you'd have thrown that away.

Step 2: the terminal value, two defensible candidates. Year 5 EBIT is $79.09m. A consolidator who can strip cost out pays perhaps 4.0× EBIT for what's left, which comes to $316.36m. The alternative is a wind-down of the remaining $395.45m of invested capital at a 55% recovery, less $40m of closing costs: 395.45 × 0.55 − 40 = **$177.50m**. Take the trade sale, since a buyer exists for this kind of asset, and keep the wind-down as the floor. PV of $316.36m at year 5 = $200.96m.

Step 3: assembly. EV = 498.97 + 200.96 = **$699.93m**. The terminal value is 28.7% of enterprise value, well below the 55 to 75 percent band a growth DCF produces. That gap is the signature of a declining firm. Nearly all the value is cash arriving in the next five years, so the answer is unusually robust to the terminal assumption and unusually sensitive to the decline rate. Equity = 699.93 − 120 = **$579.93m**, or $14.50 a share.

Step 4: the contrast that matters. Suppose you had reached for a perpetuity instead, at a "conservative" −2% forever. FCFF in year 6 would be 59.32 × 0.98 + 7.91 = $66.04m, the shrinking NOPAT plus the capital that keeps being released, giving TV = 66.04/(0.095 + 0.02) = $574.28m, PV $364.80m, and an enterprise value of $863.77m. That's 23.4% above the wind-down build. Every dollar of the difference is an assumption that a business you've just described as structurally obsolete still exists in the twenty-second century. A negative growth rate doesn't fix a perpetuity. It only makes the infinity arrive politely.

Step 5: the floor, and the verdict. The liquidation floor to equity is 177.50 − 120 = $57.5m, or $1.44 a share at year 5, against a $14.50 going-concern value. So the floor protects almost nothing, and the entire case is the run-off. Now price that. The equity cash flows, after $6.30m a year of after-tax interest and repaying the $120m out of the terminal proceeds, are 157.62, 136.89, 118.43, 102.00, 283.76. Bought at $14.50 they return 10.70%, roughly the 10.38% cost of equity this stack implies, which is what you'd expect at fair value. Bought at the market's $10.00 they return 26.0%. Bought at $16.00 they return 7.2%, below the cost of equity, and the multiple is still low.

Read it. Two numbers decide this valuation, and the multiple isn't one of them: the decline rate and the terminal exit. Push decline from 8% to 12% and the run-off shortens. Push the exit from 4.0× to 2.0× and $100m leaves the equity. At $10.00, Fairmont is a buy on a stated three-year horizon, with a written trigger to exit if the decline rate steepens for two consecutive years. At $16.00 it's a value trap, for the same reason it looks cheap there. A low multiple on a shrinking base is arithmetic, and you shouldn't mistake it for a discount.

Practice set

Work each problem fully before opening its solution (working agreement #3). Currency is ₹ crore unless stated, and the tolerance is ±2% on numeric answers unless stated.

P1 (guided, conceptual). A colleague values an NBFC by projecting FCFF = EBIT(1−t) + D&A − capex − ΔWC and discounting at WACC. List three distinct reasons the output is meaningless, and name the two architectures you would use instead.

Solution. (1) Interest is the NBFC's cost of goods sold. It isn't a financing charge, so EBIT and "operating" cash flow are undefined when the margin is the interest spread. (2) Debt is raw material. Borrowings fund the loan book the way flour funds a bakery, so there's no operating/financing separation, and WACC, which prices a capital-structure choice, has nothing to refer to. (3) "Capex" and "working capital" don't capture reinvestment. Growth is consumed as regulatory/prudential capital retained against a growing book, and FCFF never sees it. Use a dividend-discount model with the capital constraint (payout = 1 − g/ROE) instead, or the excess-return model (value = book + PV[(ROE − r) × book]), both discounted at the cost of equity.

P2 (guided). A bank has BVPS ₹450, sustainable ROE 15%, cost of equity 12.5%, long-run growth 5%. (a) Justified P/B and fair value? (b) The stock trades at ₹780. What ROE is the market pricing (same r, g)?

Solution. (a) P/B = (ROE − g)/(r − g) = (15 − 5)/(12.5 − 5) = 10/7.5 = 1.33×, so fair value = 1.33 × 450 = ₹600. (b) Market P/B = 780/450 = 1.733×. Invert it: ROE = g + P/B × (r − g) = 5 + 1.733 × 7.5 = 18.0%. The market prices three points more ROE than you do. Now go and find out who's right. Check the NIM trajectory, credit-cost normalization, the fee engine. The formula has turned a price into a research question.

P3 (guided → faded). Suryodaya Bank will earn E₁ = ₹40/share on B₀ = ₹250 (ROE 16%). To fund 14% growth for years 1–5 it retains 87.5% (payout 12.5%). From year 6: growth 6%, payout = 1 − 6/16 = 62.5%. Cost of equity 13%. Value the share.

Solution. Dividends in years 1 to 5 are 12.5% of earnings growing at 14%: 5.00, 5.70, 6.50, 7.41, 8.44. PVs @13%: 4.42 + 4.46 + 4.50 + 4.54 + 4.58 = 22.52. For the terminal, E₆ = 40 × 1.14⁴ × 1.06 = 67.56 × 1.06 = 71.61, so D₆ = 71.61 × 0.625 = 44.76 and V₅ = 44.76/(0.13 − 0.06) = 639.4. PV = 639.4/1.8424 = 347.0. Value = 22.52 + 347.0 ≈ ₹370 (P/B ≈ 1.48×). Look at where the value comes from: 94% of it arrives after the high-growth phase, when the payout taps open. For banks, growth is a deferral of dividends, and it's worth having only because ROE (16%) exceeds r (13%).

P4 (faded). A mining company's last ten years of EPS: 8, 14, 22, 30, 12, −4, 6, 18, 34, 44 (year 10 = latest). Current BVPS ₹210, 10-year average ROE 10%, cost of equity ~13%. The stock trades at ₹264, and a broker note flags "P/E of only 6×, deep value." Assess, both ways.

Solution. Raw average EPS = 184/10 = ₹18.4, so the normalized P/E = 264/18.4 = 14.3×. That's a long way from 6×. Size-corrected, 10% × 210 = ₹21, and 264/21 = 12.6×. Either way, the "6×" is the peak-P/E trap. You'd pay ~13 to 14× mid-cycle earnings for a business whose through-cycle ROE (10%) is below its cost of equity (13%). Those economics merit a discount to book. They don't merit a growth multiple. The note is extrapolating the crest.

P5 (faded). Same company: market cap 8,400, net debt 9,600, capacity 6 mt, replacement cost ≈ ₹6,500 crore/mt. Compute EV per tonne versus replacement, and state what the reading historically signals, then state what reading would signal the opposite end.

Solution. EV = 8,400 + 9,600 = 18,000, and 18,000/6 = ₹3,000 crore/mt = 46% of replacement cost. Commodity cycles have historically bottomed at deep discounts to replacement (40 to 60%). No rational actor adds supply, marginal producers exit, and the survivor's capacity is being sold below its build cost. The opposite signal is EV/tonne above 100% of replacement. Then the market pays more than a greenfield costs, new supply gets financed, and mean reversion is being invited in writing. Neither signal times anything to the quarter. Both steer the direction of your interest.

P6 (faded). A cash-burning ed-tech has a going-concern DCF value of 5,400, P(survival) 80%, distress value 900, and 45 crore shares. (a) Expected value per share? (b) It then raises 800 of fresh equity at your answer-(a) price, lifting P(survival) to 95%. New value per share (fresh cash is additive, and new shares are issued at ₹100)?

Solution. (a) E(V) = 0.80 × 5,400 + 0.20 × 900 = 4,320 + 180 = 4,500, or ₹100/share. (b) New shares = 800/100 = 8 crore, making 53 crore shares. E(V) = 0.95 × 5,400 + 0.05 × 900 = 5,130 + 45 = 5,175. Add the 800 of cash to get 5,975, and 5,975/53 = ₹112.7/share. An 18% dilution raised per-share value by 13%, because survival is worth more than the shares cost. That's the young-company paradox. For a firm near the edge, a "dilutive" raise is often the most accretive event available. Check the runway before you dismiss any raise.

P7 (independent). An explorer holds the license to a zinc block. Developed today it would be worth V = 900 against development cost K = 1,000 (NPV −100: "worthless," says the intern). In one year, zinc resolves: V_up = 1,400, V_down = 650; K unchanged, decision then; r_f = 7%. Value the license as a one-period option, and state which of the three real-option conditions the license satisfies.

Solution. p = (1.07 × 900 − 650)/(1,400 − 650) = (963 − 650)/750 = 0.4173. Payoffs: up, 1,400 − 1,000 = 400. Down, max(650 − 1,000, 0) = 0. Option = 0.4173 × 400/1.07 = 166.9/1.07 = ₹156 crore. The "worthless" block is worth ~₹156 crore because its owner may wait, learn, and walk away. Now the conditions. Exclusivity: yes, it's a license. Resolvable uncertainty: yes, since the price path and drilling data arrive before expiry. Capacity to exercise has to be checked against the balance sheet. An option you can't fund is someone else's option, and they usually pick it up cheaply.

P8 (independent). Kaveri Invest (a pure holdco) owns 30% of listed OpCo A (market cap 40,000), 55% of listed OpCo B (market cap 18,000), an unlisted plantation business (EBITDA 240, peers at 7× EV/EBITDA, no debt), and net cash of 400. Kaveri's own market cap is 10,800. Compute NAV and the holdco discount, then give your two-sentence reading.

Solution. Stakes: 0.30 × 40,000 = 12,000, 0.55 × 18,000 = 9,900, plantations 240 × 7 = 1,680, cash 400, so NAV = 23,980. Discount = 1 − 10,800/23,980 = 55%. Read it this way. A 55% discount is mid-band for Indian pure holdcos (30 to 70%). It describes structural facts (cash that never reaches minorities, tax at two gates, stakes held for control and never sold, no terminal event), and by itself it isn't a thesis. It becomes one only with a named catalyst (buyback, demerger, family settlement, distribution-policy change) or an acceptable payment for waiting (look up the dividend yield). Otherwise, expect the discount you bought to be the discount you sell.

P9 (independent). A liquidation estate under IBC realizes ₹9,000 crore. Claims: CIRP/liquidation costs 400; workmen's dues (24 months) 350; secured creditors who relinquished security 9,500; other employees (12 months) 150; unsecured financial creditors 3,000; government dues (2 years) 1,200; operational creditors 2,200; preference capital 300; equity book value 2,500. Distribute (a) at ₹9,000 crore; (b) at ₹11,000 crore.

Solution. (a) Costs take 400, leaving 8,600. Tier 2 (workmen 350 + secured 9,500 = 9,850, pari passu) exceeds 8,600, so it's paid pro-rata. Workmen get 8,600 × 350/9,850 = 305.6, and secured creditors get 8,294.4 (87.3% recovery). Everyone below (employees, unsecured FCs, government, operational, preference, equity) receives zero. (b) Costs take 400, leaving 10,600. Tier 2 is paid in full (9,850), leaving 750. Employees take 150, leaving 600. Unsecured FCs claim 3,000 and get 600, which is 20%. Government, operational, preference, equity: zero. Equity's ₹2,500 crore of "book value" produced ₹0 in both scenarios. Book equity is a residual, and it gives you no rank in the queue.

P10 (independent). A firm's assets are worth V = 800 today; its only debt is a zero-coupon obligation of face 1,000 due in one year. In a year, assets will be worth 1,300 or 500; r_f = 7%. (a) Value the equity as an option. (b) What is the debt worth, and its implied one-year yield? (c) Why do managements of such firms suddenly develop appetites for risky bets?

Solution. (a) Equity is a call with strike 1,000, so the payoff is 300 up and 0 down. p = (1.07 × 800 − 500)/(1,300 − 500) = 356/800 = 0.445. Equity = 0.445 × 300/1.07 = ₹124.8 crore. It's positive even though assets (800) < debt face (1,000), because an out-of-the-money option still has time value. (b) Debt = 800 − 124.8 = 675.2, which is 67.5% of face, so the implied yield = 1,000/675.2 − 1 ≈ 48%. That's the market's way of saying "distressed." (c) Equity is a call, so its value rises with asset volatility. A coin-flip acquisition that's value-neutral for the firm transfers value from creditors to shareholders (asset substitution). It's why covenants exist and why the IBC suspends boards on admission. It's also why you should read a distressed firm's sudden "bold strategic pivot" as the option holders maximizing vega.

P11 (independent, short-answer). An IPO note for an electric-scooter maker says: "India will have 25 million two-wheeler sales by 2035, mostly electric; the company's brand and charging network will make it the category's Apple; software subscriptions will transform margins." Translate the story into the five Dark Side inputs (state a defensible base-case number or range for each) and write two falsifiers.

Solution (model answer, and your numbers may differ within reason). (1) TAM/volume: 25m two-wheelers × ~85% electric by 2035 ≈ 21m units, and the serviceable premium segment ≈ 35% ≈ 7.4m units. (2) Share: "category's Apple" implies a premium share of 15 to 25%, or ~1.1 to 1.8m units. Sanity-check that against installed capacity and the dealer/charging footprint. (3) Margins: hardware at mature auto-OEM economics (EBIT 8 to 10%). The software claim earns a separate, smaller line (attach rate × ARPU), and it doesn't earn a blanket 25% blended margin. If the note needs software to make the equity case, that's the fragile clause. (4) Reinvestment: sales-to-capital ~2.0 to 2.5 (plants, network). (5) Risk: cost of capital 13 to 14% gliding to ~11%, and P(survival) 80 to 90%. Is it funded? Is it contribution-positive per scooter after subsidy? Falsifiers: (a) premium-segment share below 10% for two consecutive festival seasons; (b) software attach rate < 20%, or subsidy withdrawal pushing contribution margin per unit negative for more than four quarters.

P12 (timed, 20 minutes, the capstone warm-up). Your three-way work on a consumer company gives a DCF base of ₹620/share (bear 540, bull 720). In the comps, the sector median forward P/E applied to its earnings implies ₹780. The price is ₹850. Your reverse DCF says ₹850 requires 14% revenue growth for 10 years at an 18% EBIT margin, against your researched bands of 9 to 11% growth and 16 to 18% margin. Write the six-line triangulation verdict a portfolio manager would accept.

Solution (model answer). "Intrinsic work says ₹540 to 720 with a ₹620 center, and the market pays ₹850. My DCF isn't being shy here. Reversing the price requires 14% × 10-year growth at peak margins, outside my 9 to 11% band and above the sector's own base rates. Comps 'support' ₹780 only because the whole sector trades one rate-regime above intrinsic: relative comfort, absolute stretch (pricing vs valuing). Verdict: quality confirmed, price unbeatable. Pass at ₹850 and revisit below ₹700. Falsifiers that would flip me: two consecutive quarters above 13% organic growth, or gross-margin expansion evidencing pricing power I haven't modeled. What would worry me even at ₹650: promoter pledge creeping above 10%, or receivable days stretching." Yours should hit the same five beats: range vs price · reverse-DCF burden stated · comps reconciled as pricing-not-valuing · action with trigger level · falsifiers both ways.

P13 (guided). Ratnagiri Holdings (synthetic), a pure holding company, owns 40% of listed OpCo A (market cap ₹50,000 crore), 30% of listed OpCo B (market cap ₹90,000 crore), and net cash of ₹2,000 crore. Its own market cap is ₹24,500 crore. (a) Compute NAV and the holdco discount. (b) A group demerger of the kind that has historically narrowed such discounts toward 35% is announced. With NAV unchanged, what return would the re-rating deliver? (c) Say in a single sentence whether the 50% discount is itself a thesis.

Solution. (a) Stakes: 0.40 × 50,000 = 20,000 and 0.30 × 90,000 = 27,000. Add cash of 2,000 and NAV = ₹49,000 crore. Discount = 1 − 24,500/49,000 = 50%. (b) At a 35% discount on the same NAV, value = 49,000 × 0.65 = 31,850, so the return = 31,850/24,500 − 1 = +30%. (c) No. A 50% discount is mid-band for Indian pure holdcos and merely describes the structural facts (cash that never reaches minorities, tax at two gates, control-stakes never sold, no forced terminal event). It becomes a thesis only once you attach the named catalyst and the payment you collect for waiting, which is the holdco's own ~2%-ish dividend yield.

P14 (guided). Assemble a football field for Meridian Components (synthetic, NYSE): 200m diluted shares, net debt $1,500m, price $85, EBITDA $1,900m. Your rails are DCF EVs of $16,000m / $18,500m / $21,000m (bear/base/bull), comps at 9.0× to 11.0× EV/EBITDA, and a precedent-transaction/LBO floor at 8.5× EBITDA. (a) Convert each rail to a per-share value. (b) Where does $85 sit? (c) Write the two-line range verdict, without averaging.

Solution. Bridge: equity = EV − 1,500, and per share = equity ÷ 200. (a) DCF: bear (16,000 − 1,500)/200 = $72.50, base (18,500 − 1,500)/200 = $85.00, bull (21,000 − 1,500)/200 = $97.50. Comps: 9.0 × 1,900 = 17,100 EV, or $78.00, and 11.0 × 1,900 = 20,900 EV, or $97.00 (band $78 to $97). LBO/precedent floor: 8.5 × 1,900 = 16,150 EV, or $73.25. (b) $85 is right on the DCF base, inside the comps band ($78 to $97), and ~$12 above the LBO floor ($73.25). (c) Verdict: "Intrinsic and market evidence agree near $78 to $97 with an $85 center. The tape is fairly priced and the LBO floor (~$73) limits downside, so there's no margin of safety: act only below ~$78 (bottom of comps / bear-to-base zone). I state the agreement and refuse a spurious blended number."

P15 (independent). An operating conglomerate (synthetic) has: an infrastructure segment (EBITDA ₹3,000 crore, peer EV/EBITDA 8×; the segment is 70%-owned and carries ₹6,000 crore of the group's debt), a wholly-owned consumer segment (EBITDA ₹1,800 crore, 12×), and a 40% non-control stake in a listed logistics company (stake market value ₹9,000 crore). Unallocated corporate cost is ₹400 crore/year pre-tax (tax 25%). Parent + wholly-owned net debt is ₹15,000 crore (this already includes the ₹6,000 crore inside the infra subsidiary). There are 60 crore shares, priced at ₹380. Compute equity NAV per share and the discount, remembering every bridge line.

Solution. Segment EVs: infra 3,000 × 8 = 24,000, consumer 1,800 × 12 = 21,600. Listed stake: 9,000 less a 20% non-control/illiquidity/exit-tax haircut = 7,200. Capitalized central costs: 400 × (1 − 0.25) = 300 post-tax × 8 = 2,400, subtracted. Gross EV = 24,000 + 21,600 + 7,200 − 2,400 = 50,400. Now the bridge. Less net debt 15,000. Less minority interest = 30% of the infra equity value = 30% × (24,000 − 6,000) = 5,400. Subtract the segment's own allocated debt before taking the minority's share, or you'll double-count it. Equity NAV = 50,400 − 15,000 − 5,400 = 30,000, or ₹500/share. Discount = 1 − 380/500 = 24%, ordinary for an Indian operating conglomerate. The analysis is the two forgotten lines (central costs, minority interest at value). Summing segment EVs is the easy part.

P16 (independent). Tarn Retail (synthetic, US $m) can open a flagship in a new country: developed today the project is worth V = 300 against a build cost K = 340, so static NPV = −40 ("kill it"). But the flagship is a pilot: in one year the market reads out (V_up = 520 or V_down = 210) and only then need the network capital K = 340 be committed; r_f = 6%. (a) Value the staged pilot as a one-period option. (b) Compare with committing today. (c) Which real-option condition is the binding question here, and what abuse would double-count it?

Solution. (a) p = ((1.06 × 300) − 210)/(520 − 210) = (318 − 210)/310 = 0.348. Payoffs at the decision: up = max(520 − 340, 0) = 180, and down = max(210 − 340, 0) = 0. Option value = (0.348 × 180)/1.06 = $59.2m. The "worthless" −$40m project is worth ~$59m because the owner may wait, learn, and walk away. (b) Committing today, NPV = 300 − 340/1.06 = −$20.8m, so staging is worth ~$80m more than committing. (c) Exclusivity is the binding question. If any rival can open the same follow-on stores once the market is validated, competition strips the premium toward zero, and a generic "option to expand" is worth ~nothing. The option only counts if a lease lock-up, license, or first-mover right makes the follow-on sites exclusive. The abuse would be adding this ~$59m on top of a DCF whose forecast already contains the follow-on stores. That double-counts the same flexibility, so assign each risk one home.

P17 (independent). Your three-way on a paints company (synthetic) gives a DCF base of ₹2,400 (bear 2,050, bull 2,750). Comps (sector median forward P/E × your EPS) imply ₹3,050, and the stock trades at ₹3,200. Your reverse DCF says ₹3,200 needs 15% revenue growth for a decade at a 19% EBIT margin, against your researched bands of 10 to 12% growth and 17 to 19% margin. (a) State the gap. (b) Write the reconciling sentence, without averaging. (c) Give an action with a trigger price and one falsifier.

Solution (model answer). (a) ₹3,200 is above the entire DCF range (₹2,050 to 2,750) and ~33% over the ₹2,400 base. (b) "I don't average ₹2,400 and ₹3,050 into a ₹2,700 'target'. The comps at ₹3,050 only certify that the sector is priced richly (relative comfort). The reverse DCF names the absolute burden: 15% × 10-year growth at peak margins, above my 10 to 12% band and the category's base rates. The DCF prices economics, the comps price the sector's mood, and the reverse DCF prices expectations. They disagree because they measure different things." (c) Action: pass at ₹3,200 and re-examine below ₹2,750, the top of my DCF range, where the field's weight turns. Falsifier: two consecutive years of organic growth above 13%, or gross-margin expansion showing durable pricing power I haven't modeled, would lift my growth band and reopen the verdict.

P18 (timed, 20 minutes, stretch). Take a diversified Indian holding company (synthetic). Your SOTP puts NAV at ₹6,20,000 crore (listed stakes at market ₹5,40,000 + unlisted arm ₹90,000 + net cash ₹10,000 − capitalized central costs ₹20,000). Its market cap is ₹3,10,000 crore, and long-run dividends to holders run ₹7,440 crore/year. A demerger of the largest listed stake into a directly held structure is under regulatory review. In ≤8 lines, give (a) the discount, (b) the yield you're paid to wait, (c) the re-rating return if the discount narrows to 35% (NAV unchanged), and (d) the conditional verdict a PM would accept.

Solution (model answer). (a) Discount = 1 − 3,10,000/6,20,000 = 50%. (b) Carry = 7,440/3,10,000 = 2.4%/yr. (c) At a 35% discount, value = 6,20,000 × 0.65 = 4,03,000, so the return = 4,03,000/3,10,000 − 1 = +30%. (d) "NAV ₹6.2 lakh crore, price ₹3.1 lakh crore, 50% discount, the deep end of the Indian band but not a thesis by itself. The catalyst is the demerger under review. If it narrows the discount to ~35%, the re-rating is ~+30%, and I collect 2.4%/yr while waiting. Action: a catalyst-contingent position, sized small until the demerger clears its regulatory review. Falsifier: if the demerger is withdrawn or blocked, the discount reverts to structural weather and the thesis is void. I underwrite the catalyst, and I don't count on mean reversion on faith."

P19 (guided). Residual income, single stage, three routes (India). Konkan Instruments (synthetic): book equity ₹150 crore, ROE a stable 15%, cost of equity 11%, sustainable growth 4.5% forever. (a) Compute the implied retention and payout. (b) Value the equity by the residual-income route. (c) Confirm it by the dividend route and by the justified price-to-book ratio, and state in one sentence what the split between book value and the residual-income stream is telling you.

Solution. (a) retention = g/ROE = 4.5/15 = **30%**, so payout is 70%. (b) RI₁ = (15% − 11%) × 150 = ₹6.00 crore, growing at 4.5%. V = 150 + 6.00/(0.11 − 0.045) = 150 + 92.31 = **₹242.31 crore**. (c) Dividend route: E₁ = 0.15 × 150 = ₹22.50 crore, then D₁ = 22.50 × 0.70 = ₹15.75 crore, so V = 15.75/0.065 = **₹242.31 crore**. Justified P/B: (ROE − g)/(r − g) = (15 − 4.5)/(11 − 4.5) = **1.615×**, and 1.615 × 150 = ₹242.31 crore. The split is 150 of book against 92.31 of residual income, so 38% of the value is a claim that a four-point spread survives. The other 62% is already on the balance sheet, and that's why this model is easier to defend than a DCF on the same company.

P20 (independent). Multistage residual income with a persistence factor (US, $m). Ridgeline Controls (synthetic): book equity $80m, ROE 20% held for three years, retention 40%, cost of equity 10%. After year 3, residual income decays with a persistence factor ω = 0.50. (a) Build the three-year residual-income schedule with its book roll-forward. (b) Compute the continuing value and the total equity value. (c) Recompute on ω = 1 and say in a single sentence what economic claim the difference represents.

Solution. (a) Book rolls forward by NI × 0.40. Year 1: opening 80.000, NI 16.000, RI = (20% − 10%) × 80 = 8.000, PV 7.273, closing 86.400. Year 2: opening 86.400, NI 17.280, RI 8.640, PV 7.141, closing 93.312. Year 3: opening 93.312, NI 18.662, RI 9.331, PV 7.011, closing 100.777. Explicit PV = $21.42m. (b) RI₄ = (20% − 10%) × 100.777 = $10.078m, so CV₃ = RI₄/(1 + r − ω) = 10.078/(1.10 − 0.50) = $16.796m and PV = 16.796/1.10³ = $12.62m. V₀ = 80 + 21.42 + 12.62 = **$114.04m**, a price-to-book of 1.43×. (c) At ω = 1 the continuing value is 10.078/0.10 = $100.78m, PV $75.72m, and V₀ = **$177.14m**, a P/B of 2.21×. The $63m difference is one claim and one only: that a ten-point return spread never erodes. Anyone quoting the higher number owes you the name of the moat that holds it.

P21 (independent). The clean-surplus break (India). Nilgiri Chemicals (synthetic) reports opening book equity ₹500 crore, profit for the year ₹80 crore, dividends paid ₹30 crore, and closing book equity ₹538 crore. Cost of equity 12%. (a) What should closing book equity have been, and what does the gap tell you? (b) Compute year-one residual income on reported profit and on the corrected basis. (c) By how much would a valuation built on the reported series overstate value, and where in the annual report do you find the culprit?

Solution. (a) Clean surplus requires 500 + 80 − 30 = ₹550 crore. Reported closing equity is ₹538 crore, a gap of −₹12 crore, so ₹12 crore of value left equity without passing through profit. (b) On reported profit: RI = 80 − 0.12 × 500 = **₹20.00 crore**. On comprehensive income of 80 − 12 = ₹68 crore: RI = 68 − 60 = **₹8.00 crore**. (c) The overstatement is ₹12 crore in year one, worth 12/1.12 = **₹10.71 crore** today, and it compounds if the same item recurs. The culprit is in the statement of changes in equity and the other comprehensive income section. It's typically a foreign-currency translation difference, a defined-benefit remeasurement, or a fair-value movement on investments. One caution before you charge it. Confirm the gap isn't a transaction with owners, such as a buyback or a share-based-payment credit, which belongs outside the earnings series entirely.

P22 (independent). A declining firm, run off (India). Godavari Directories (synthetic): revenue ₹800 crore falling 10% a year for four years; EBIT margin 12% flat; tax 25%; sales-to-capital 1.6, so opening invested capital is ₹500 crore; WACC 12%. At the end of year 4 the remaining invested capital is wound down at a 50% net recovery. (a) Build the four-year free-cash-flow schedule, showing the capital released each year. (b) Compute enterprise value and the terminal share of it. (c) Recompute the terminal value as a −3% perpetuity instead, and say which number you would defend and why.

Solution. (a) Revenue: 720.000, 648.000, 583.200, 524.880. NOPAT = revenue × 0.12 × 0.75: 64.800, 58.320, 52.488, 47.239. Capital released = −ΔRevenue/1.6: 50.000, 45.000, 40.500, 36.450. FCFF: 114.800, 103.320, 92.988, 83.689. PVs at 12%: 102.500, 82.366, 66.187, 53.186, summing to ₹304.24 crore. (b) Invested capital closes at 500 − 171.95 = ₹328.05 crore, and the wind-down is 328.05 × 0.50 = ₹164.03 crore, PV 164.03/1.12⁴ = ₹104.24 crore. EV = 304.24 + 104.24 = **₹408.48 crore**, of which the terminal is 25.5%. (c) A −3% perpetuity gives FCFF₅ = 47.239 × 0.97 + 9.84 = ₹55.66 crore, TV = 55.66/(0.12 + 0.03) = ₹371.09 crore, PV ₹235.84, and an enterprise value of ₹540.07 crore, which is 32% higher. Defend the ₹408 crore. The perpetuity assumes a directories business still printing in the 2200s, and no negative growth rate repairs that. The wind-down assumes only that someone buys or scraps the assets at the end of a horizon you can actually see.

Applied mini-project

THE PHASE 3 CAPSTONE, the M4 milestone. This is the gate for the entire phase and the certification exercise for competency C5 ("value a business three ways"). Everything you've learned in the phase converges here. Budget ~7 hours of the week's 20. The model templates you've already built make that possible. Reuse the chassis instead of rebuilding it.

The assignment

Produce a three-way valuation (DCF + comparables + reverse DCF) of one Indian company AND one US company from the candidate list below. Deliver each as (a) a working model and (b) a two-page memo, then self-grade against the rubric. Every key assumption has to fall inside a defensible band, justified by the company's own 5-to-10-year history, by peer base rates, or by an explicit structural argument written in the model. An assumption outside all three is an automatic rubric deduction.

Candidate list (pick one per market; do not pick a company you fully modeled in M3.04–M3.09)

Stretch tier (optional third company, extra credit, using the tools above): HDFC Bank or JPMorgan (excess-return model), Tata Steel or Nucor (mid-cycle normalization), Grasim (SOTP + holdco), Uber (young-growth graduate, with survival haircut now ~1.0, though the Dark Side architecture still fits). The model-answer memo below is on Infosys. It's deliberately on neither list, so it can guide you without being copyable.

Data pull (be exact): India: the latest annual report (BSE/NSE filings page or the company's investor-relations site), the last two quarterly earnings decks + con-call transcripts, screener.in for the 10-year spread. US: the latest 10-K and last two 10-Qs (SEC EDGAR), the earnings-call transcripts, stockanalysis.com or TIKR for the long history. Record the price and date on the memo's first line. Every number in the model must trace to one of these sources (working agreement #7).

Model requirements (both companies; the audit checklist your file must pass)

1. Chassis: your full three-statement model or the "3-statement-lite" version (revenue → EBIT → NOPAT → reinvestment → FCFF with a working-capital and capex schedule). Inputs/calcs/outputs separated. No hardcodes inside formulas. A checks block that actually checks (balance, non-negative cash, FCFF ties).
2. Forecast horizon: 10 years (5 explicit + 5 fade), driver-based, from volume × price or a segment build. A single cell saying "revenue grows x%" doesn't count.
3. Growth logic: g = reinvestment rate × ROIC respected throughout. Reinvestment via sales-to-capital or explicit capex/WC schedules. Year-10 ROIC must be a number you can defend against the company's own history and peers.
4. WACC, built as in your cost-of-capital work: currency-matched risk-free rate, bottom-up beta from ≥4 peers (unlever/relever), synthetic-rating cost of debt if unrated, market-value weights. INR model for the Indian company, USD for the US one. Never mix.
5. Terminal value both ways: Gordon (g ≤ currency risk-free rate, with steady-state reinvestment consistent with terminal ROIC) and an exit-multiple cross-check. Flag it if TV > 85% of EV, then explain or shorten the fade.
6. Equity bridge: EV − net debt − minority interest − preferred + non-operating assets (cross-holdings at market less tax haircut) ÷ diluted shares (treasury-stock method, RSUs added).
7. Sensitivity: a WACC × terminal-g table (≥5×5) and a one-way table on your single most-loaded driver.
8. Scenarios: base/bull/bear as toggled driver sets (not ±10% on everything), each with a one-line story from the narrative loop. Probability-weight them for your expected value.
9. Comps table: ≥6 true comparables (business-model peers, both markets allowed where relevant), forward multiples on a consistent basis, lease/SBC/one-off normalization as in relative valuation, median + harmonic mean, and your target priced off the regression or median with the residual explained.
10. Reverse DCF: solve for (a) the implied revenue growth at your margin assumptions and (b) the implied margin at your growth assumptions, such that DCF value = market price. Write the market's implied story in ≤2 sentences each.
11. Special-situation overlay if triggered: if your pick has a cyclical segment, normalize it (4.3). Destination-value a loss-making segment (4.4). For material cross-holdings, SOTP the bridge (4.6). The triage table is part of the audit.
12. Version hygiene: put the date, price, source list on the cover tab. Keep an assumptions log with the band and justification for each of the big five (growth, margin, reinvestment, WACC, terminal).

The two-page memo (structure is mandatory; ~850–950 words + 2 exhibits)

Conclusion first (C11, the Minto rule). Any reader must get your answer in the first five lines.

1. Header + verdict box (5 lines): company, ticker, date, price, market cap. Then your value range (bear/base/bull), the one-line verdict ("fairly priced; pay below ₹X"; "priced for a story I can falsify; pass"; "market underwrites less than the evidence supports; interesting below $Y") and the expected return at the current price.
2. The business in five lines: what it does, the profit engine, the moat claim, unit economics, capital-allocation record. No adjectives without numbers.
3. The three drivers that matter (table): your band vs the market-implied number for each, with the evidence for your band (history/peers/structure). Three drivers, and not ten, because choosing them is the analysis.
4. Triangulation (table + 4 lines): DCF range, comps-implied value, reverse-DCF implied expectations, and the sentence that reconciles them (which method you trust here and why, and never an average).
5. What the price implies, and your variant view (5 lines): the market's story in plain words (from requirement 10), where you differ, and why you might be the one who's wrong (the strongest bear/bull counter to your own view).
6. Risks & falsifiers (5 lines): three dated, observable falsifiers, one pre-mortem sentence ("it's 3 years on and this memo was wrong because…"), and your re-examine trigger price.
7. Exhibits: the football field (bear/base/bull DCF, comps band, current price, reverse-DCF marker) and the driver table. Appendix line: model tab map.

Grading rubric (score each company separately; 0 = absent/wrong, 1 = present with gaps, 2 = professional)

Pass = ≥23/28 (≈82%) per company, with no zero on criteria 1 to 4. A mechanical error that flips the verdict caps the attempt at fail until it's fixed. The protocol comes from the master map. Finish completely, then read the model-answer memo below, then score each line 0/1/2 in writing. Log the score and your top miss in your journal. Both companies must pass to clear the module and Phase 3. Passing unlocks the next phase.

Model-answer example memo (the standard to write toward)

Third company: Infosys, not on the candidate list. All figures are illustrative, modeled on ~FY25 scale and mid-2026 prices, so pull the actual filings and re-derive before using any number. The memo shows the deliverable format, and the model behind it follows the requirements above.

INFOSYS LTD (NSE: INFY), valuation memo. Date: [mid-2026]. Price ₹1,520, diluted shares ~415 crore, market cap ≈ ₹6.3 lakh crore, net cash & non-operating investments ≈ ₹47,000 crore.

Verdict. Intrinsic range ₹870 / ₹1,025 / ₹1,180 (bear/base/bull, INR DCF), with a comps band of ₹1,380 to 1,660. At the illustrative ₹1,520, comps put the stock mid-pack against peers, while the DCF range is below the price at INR discount rates. The price embeds either ~13.5% revenue growth for a decade (against the memo's 6 to 9% band) or a ~9.5% cost of equity, which is a global investor's and not the memo's. The check that settles it: which of the two the evidence supports, a decade of 13%+ growth against the company's own base rates or a cost of equity that low for an INR investor. A specimen memo on a listed company stops at that arithmetic. The action line belongs to your own memo on your own dated figures.

The business in five lines. India's second-largest IT-services exporter: ~₹1.63 lakh crore revenue (~$19bn), ~85% from application services, digital transformation, and managed services under multi-year contracts. Top-listed clients are concentrated in BFSI and North America. Profit engine: the pyramid, meaning offshore delivery at ~21% EBIT margins with utilization, pricing, and pyramid mix as the levers. Moat: switching costs on embedded contracts and an offshore cost position, real but shared with four scaled rivals. Pricing power is modest (moat = durability of a middling spread, ROIC ex-cash > 35%). Unit economics: revenue per employee ~$57k, ~85% variable-cost structure, negative-lite working capital, capex ~2% of revenue, a cash machine converting ~85% of NOPAT to FCFF. Capital allocation: ~85% payout via dividends + buybacks for a decade. No empire-building. Acquisitions small and digestible.

The three drivers (band vs implied).

Triangulation. DCF (10y, FCFF, sales-to-capital 3.0, terminal g 5% ≤ r_f, terminal ROIC 25%): PV explicit ₹1.72 lakh crore + PV terminal ₹2.06 lakh crore (TV = 55% of EV, healthy) gives EV ₹3.78 lakh crore. Add net cash of ₹0.47 and you get ₹1,025/share base. Comps (TCS 25×, HCLT 21×, Wipro 17×, LTIM 27×, TechM 20×, Accenture 24× forward P/E, median ~22×): Infosys at 22.0× forward is priced mid-pack for its quality tier, so it's "fair" at ₹1,420 to 1,660 relative to a sector that is itself expensive in INR terms. The stable-growth ceiling from relative valuation, at r 11.5% and g 5%, is ~15×, and the entire sector trades above its INR ceiling on scarcity + flows. Reverse DCF: ₹1,520 requires ~13.5% × 10y revenue growth at 21% margins, above Infosys's best post-2015 decade, or my base case at a 9.3% discount rate. Reconciliation: the methods disagree because they price different things. The DCF prices INR economics. The comps certify only relative position inside a premium-rated sector. The reverse DCF names the premium: a growth-acceleration story and/or a global-capital discount rate. I trust the DCF for value and the reverse DCF for the entry decision, and the comps tell me I won't be paid quickly for being right.

What the price implies, and the variant to test. The market's story: GenAI-led demand recovery lifts industry growth back to low-teens and Infosys holds share and margins. The variant: GenAI cuts both ways, since it accelerates bookings and deflates pricing per unit of work. Net acceleration to 13%+ for a decade has no precedent in the company's own base rates. Where the variant could be wrong: a weaker rupee alone adds an extra 3 to 4% of INR growth, and the 6 to 9% band partly embeds that. If cc growth returns to 8 to 10% and the rupee slides, the bull DCF (₹1,180) is still below the illustrative ₹1,520. The memo's "wrongness budget" is mostly the discount rate, so that's the input to test first.

Risks & falsifiers. (1) If cc revenue growth exceeds 10% for two consecutive fiscal years, my growth band is falsified, so revisit. (2) If EBIT margin breaks and holds above 23% for four quarters, with GenAI productivity accruing to vendors and not clients, my margin band is falsified. (3) If a mega-deal cycle lifts bookings/book-to-bill above 1.2× for a year, base rates shift. Pre-mortem: "three years on, this memo was wrong because I treated a platform shift as a pricing headwind, and IT services re-rated as AI implementers." Re-examine trigger: any of the three falsifiers above.

Exhibits (model tabs): football field [Exhibit 1]; driver bands vs implied [Exhibit 2]. Model: INFY-3way-v1.xlsx, tabs: Cover/Log · Inputs · Drivers · DCF · WACC · Comps · Reverse · Field.

Why does this memo score 26 to 28 points out of 28? The verdict and expected return are in the first five lines (crit. 12). Three drivers each carry an evidence-based band and a market-implied number (5 to 7, 10). The triangulation sentence explains the disagreement causally instead of averaging it (11). The falsifiers are dated and observable, and the pre-mortem argues against its own author (13). Every number traces to the model tabs (14). Notice that the verdict is a pass. Most careful three-way valuations end right here, and writing "no edge" well is the skill the capstone certifies.

Case Lab: run case CL-3.3 (The Discount That Wouldn't Close) in the app's Cases tab for this module. It is a guided special-situations triangulation of a synthetic Indian holdco-plus-cyclical group: build the SOTP, apply and interrogate the holding-company discount, lay the pieces on a football field against a reverse-DCF, and write the reconciling verdict. The case grader scores your NAV bridge (central costs and minority interest present?), your discount reading (description vs catalyst-backed thesis), and your no-averaging sentence against the Section 8 rubric.

Reading & resources

• Aswath Damodaran, Investment Valuation (3e): Ch. 21 (financial-service firms), Ch. 22 (negative earnings), Ch. 23 (young companies), Ch. 28–29 (options to delay/expand/abandon), Ch. 30 (distressed equity as an option). The reference text for this entire module. [Paid] [Intermediate–Advanced]
• Damodaran, The Dark Side of Valuation (2e/3e): the life-cycle chapters, especially the young-company, cyclical/commodity ("Ups and Downs"), and financial-service ("Mark to Market") chapters. [Paid] [Intermediate]
• Damodaran, Narrative and Numbers (2017): whole book; the Uber, Ferrari, and Alibaba cases are the worked templates for the narrative loop. [Paid] [Beginner–Intermediate]
• Damodaran's free NYU resources: the Valuing Financial Service Firms and Valuation of the Young and the Restless lecture videos + the eqexret.xls (excess return, banks) and fcffginzu spreadsheets; his country-risk and ERP pages (update each January). pages.stern.nyu.edu/~adamodar. [Free] [All levels]
• McKinsey, Valuation (8e, 2025): the banks chapter, the cyclical-companies chapter, and the flexibility/real-options chapter (chapter numbers move between editions; use the TOC). [Paid] [Advanced]
• Mauboussin & Callahan (Counterpoint Global): The Base Rate Book (growth base rates for the young-company work and the capstone bands) and Measuring the Moat (spread durability, the bank section's duration question). [Free PDFs] [Intermediate]
• IBBI quarterly newsletters + ibbi.gov.in data: live CIRP outcomes, recovery rates, timelines; the primary source for the distressed base rates. [Free] [Intermediate]
• RBI Financial Stability Report (half-yearly): system GNPA, provisioning, capital ratios; the context for any Indian bank valuation. rbi.org.in. [Free] [Intermediate]
• Stephen Moyer, Distressed Debt Analysis: the professional text if distressed becomes your field; far beyond the defensive mandate here. [Paid] [Advanced]
• Schilit, Financial Shenanigans (4e): re-read Ch. on acquisition and cash-flow tricks before any bank/holdco capstone stretch pick. [Paid] [Intermediate]
• Filings behind the examples here: HDFC Bank & ICICI annual reports + Basel III Pillar 3 disclosures (bank capital in the wild); Tata Steel annual reports FY18 to FY25 (a full cycle on one screen via screener.in); Swiggy/Eternal (Zomato) quarterly shareholder letters (young-platform disclosure done well); Reliance segment note + annual report; Bajaj Holdings AR + its two listed stakes' prices (recompute the discount today); any NCLT resolution-plan order via ibbi.gov.in case records (read one waterfall in the wild). [Free] [Intermediate]
• Zerodha Varsity, Fundamental Analysis module (free refresher on Indian filings and bank metrics vocabulary). [Free] [Beginner]
• Going deeper: Berger & Ofek, "Diversification's Effect on Firm Value" (Journal of Financial Economics, 1995; free via SSRN / university mirrors) [Free] [Advanced]. The classic diversification-discount study cited in the SOTP section. Read it for the measurement method and the (later-contested) causes behind the SOTP and holdco worked examples.
• Going deeper: Aswath Damodaran, "The Promise and Peril of Real Options" (NYU Stern working paper, free PDF at pages.stern.nyu.edu) [Free] [Advanced]. The exclusivity / resolvable-uncertainty / exercisability conditions and the double-counting abuse in full; the deep reference behind the real-options section and Worked Example / Practice P16.

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 triage of when the standard machine breaks, excess-return and DDM valuation of banks, mid-cycle normalization for cyclicals, the Dark Side architecture with a survival haircut, real options and their abuse tests, sum-of-the-parts with the forgotten lines, India's holding-company discount, the IBC waterfall and equity as an option, and the narrative-to-numbers loop with published falsifiers",
  "use": "This module is a set of six different jobs, and a tool's usefulness varies more across them than anywhere else in the phase. Where it earns its place. (1) TRIAGE FIRST: paste a company's segment note, balance-sheet shape and revenue history and ask which row of the section 4.9 shelf it belongs in, and — more usefully — which rows it *partly* belongs in, because the expensive mistakes are hybrids (an NBFC inside a conglomerate, a cyclical with a young platform bolted on). (2) HISTORY ASSEMBLY for normalization: ten years of margins, ROE and capacity, plus the peak and trough dates, is the single most tedious input in section 4.3 and the one a tool can pull with citations. (3) THE SOTP SKELETON: segment notes, listed-stake percentages and their current market values, minority-interest terms, unallocated corporate costs and the location of debt — the lines Worked example 4 calls the place SOTPs go to die, all of them findable and all of them dull. (4) STAKE AND STRUCTURE MAPPING for a holdco: cross-holdings, promoter entities and pledges from shareholding patterns. (5) FILING FORENSICS for a bank: restructured-book language, provision-coverage movements, divergence-report disclosures. What a tool must never do here is set the probability of survival, the mid-cycle margin, the terminal share or the discount's catalyst — those are the four numbers the whole module is about, and each one is a judgement with a distribution behind it.",
  "tools": ["Chat assistants — Claude, ChatGPT for the triage question, for building the segment and stake skeleton of an SOTP, and for stress-testing your narrative against the 3P test before you write it down", "Document/PDF AI — for ten years of segment notes and asset-quality disclosures, IBBI resolution-plan documents and NCLT orders, Section 53 claim schedules, and the shareholding patterns and related-party notes a holdco analysis stands on", "Coding assistants — for the excess-return and DDM engines with their equivalence asserted, the normalization methods run side by side, the survival-haircut distribution, and the discount-catalyst arithmetic below"],
  "prompt": "For <COMPANY>, extract from the last ten annual reports: revenue, EBIT margin, ROE and closing book value per share for each year; the physical capacity or unit base at each year-end; and the dates of the highest and lowest margin years. Cite the statement and page for every figure. Separately, list each reported segment with its revenue, EBITDA and capital employed, every listed shareholding with the percentage held, and every line of unallocated or corporate cost. Do not compute a mid-cycle number, do not choose a multiple, do not tell me whether the company is cheap, and do not estimate any probability.",
  "verify": "Two verifications, and in this module the second one is the module. First, the model must be the right *kind*: an equity-direct, book-anchored model for a lender with no EV or EBITDA anywhere in it; a normalized base for a cyclical with the window checked to contain a real peak and a real trough; a survival probability stated explicitly rather than silently set to one. A tool will produce an EV/EBITDA comp table for a bank without hesitation, because the template exists. Second, every judgement input needs a range, not a point, and the ranges must be stated in the memo: the mid-cycle margin, the terminal share, the probability of survival and the catalyst's timing are all guesses, and the honest output of this module is a conditional sentence — 'the price buys X; here is what would falsify X' — never a target. Never accept a base rate on IBC recoveries, cyclical amplitudes or holdco discounts from a chat window; pull the IBBI quarterly newsletter, the filings and the price history yourself, because a confident unsourced frequency is the failure mode most likely to survive review (Guardrail 1-5).",
  "diy": "The gate is unaided: you value a bank three ways and prove the equivalence, normalize a cyclical by three methods and reconcile them, build a Dark Side model with an explicit survival haircut, price a one-period real option and state its three exclusivity conditions, assemble an SOTP including capitalized central costs and minority interest, rank the Section 53 waterfall from memory, and write the story-to-drivers-to-falsifiers loop. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You valued Vaigai three ways to the same ₹28,260 crore. You normalized Deccan Ispat by three methods that converged on ₹29 to 30 of mid-cycle EPS. You put a survival haircut on Zyppin and watched the price turn out to be buying a different story entirely. You built an SOTP with the lines that kill SOTPs, computed a 60% holdco discount and asked what you're paid to wait, ranked the Section 53 waterfall, and priced a pharma option at ₹620 crore against ₹452 crore of commitment. All of it stays. The modern layer here is unusually narrow, and that's deliberate. The six situations here are defined by their judgement inputs, and no tool improves a judgement. What code adds is the ability to show, in numbers, how much each judgement is worth. That's the only sound way to decide where to spend your research.

Today's workflow. Three habits, none of them glamorous.

• Write the model that matches the row, and make the mismatch fail loudly. An assert "EV" not in bank_model is a joke until you've seen a bank comped on EV/EBITDA in a real note. Put the excess-return and DDM engines in one file with their equivalence asserted, so clean surplus becomes a test and stops being a belief.
• Report every judgement input as a range that propagates. Mid-cycle margin, terminal share, survival probability and discount-catalyst timing are the four numbers all six situations turn on. Each one deserves a low-base-high that flows through to a value band, and the band is the deliverable.
• Make the price the question. Every worked example above ends by inverting: what share, what spread, what duration, what probability does the price require? That inversion is a root-find. Once it's a function, you can run it on every name, and not only on the one you already had a view about.

Tools & sources (IN + US). For code: numpy for the engines, scipy.optimize.brentq for every inversion (the implied share, the implied spread duration, the implied survival probability), pandas for the ten-year normalization history and the SOTP skeleton, and matplotlib for the football field and the discount-through-time chart.

For India, start with the RBI Master Circular on Basel III capital adequacy, which gives the CET1 floor that sets a bank's payout constraint. RBI's Financial Stability Report and bank annual reports carry GNPA, provision coverage and any divergence disclosure. IBBI, at ibbi.gov.in, has the quarterly newsletter with live CIRP recovery rates. It also holds resolution plans and NCLT orders. Quarterly shareholding patterns on NSE and BSE show promoter holding and pledge percentages. The Ind AS 108 segment note in the annual report is the floor of any SOTP. screener.in has the ten-year margin and ROE history a normalization needs. Stock-exchange filings carry demerger and scheme announcements, the catalysts the SOTP section says are the only thing that closes a discount.

For the US, use EDGAR for 10-K segment notes (ASC 280) and the FR Y-9C style capital disclosures in bank filings. FDIC and Federal Reserve data give banking-sector aggregates. FRED covers the cycle itself: DGS10 for the rate regime, USREC for the NBER recession dating that tells you whether your normalization window actually contains a trough, and the producer-price series for your specific commodity. Find that one through FRED's own PPI-by-industry browser instead of guessing at it. Damodaran's current-year datasets give the sector anchors a mid-cycle multiple is checked against.

# What the standard machine does to a bank — R3, recomputed in-session on §4.2's own algebra.
tax, r_debt, r_equity, g = 0.25, 0.055, 0.13, 0.05   # funding cost, cost of equity, mature growth
def wacc(equity_multiplier):
    w_d = 1 - 1 / equity_multiplier      # deposits and borrowings as a share of the balance sheet
    return w_d * r_debt * (1 - tax) + (1 - w_d) * r_equity
for L in (6, 8, 10, 12, 14):
    print(f"{L:2d}x leverage   WACC {wacc(L):7.4%}   WACC - g = {wacc(L)-g:+8.4%}")
#  6x leverage   WACC 5.6042%   WACC - g = +0.6042%
#  8x leverage   WACC 5.2344%   WACC - g = +0.2344%
# 10x leverage   WACC 5.0125%   WACC - g = +0.0125%   <- a Gordon denominator of 1.25 basis points
# 12x leverage   WACC 4.8646%   WACC - g = -0.1354%   <- the denominator is NEGATIVE

Verify. Prove every engine against the worked figures above before it touches a live name. Start with Vaigai. The excess-return path 800.0 · 896.0 · 1,003.5 · 1,123.9 · 1,258.8 discounts to 3,477.7. Terminal book is 35,246.8, the terminal excess return 704.9, and the terminal value 8,811.7, discounting to 4,782.6. The total is ₹28,260 crore = ₹282.60 a share at 1.413× book. The DDM route lands on the same number through dividends of 1,000.0 · 1,120.0 · 1,254.4 · 1,404.9 · 1,573.5 and a year-5 equity value of 44,058.5. The bank mini-example should return 137.50 all three ways. For Deccan Ispat, check mid-cycle EPS of 29.60 · 28.60 · 30.00 by the three methods, and the peak's optical 6.00× against a normalized 17.08×. Enterprise value per tonne should be 117.2% of replacement cost at the peak and 46.3% at the trough. Mid-cycle fair value comes to ₹260.00, 1.0× book, which is the coherence check that a business earning its cost of capital is worth its capital. For Zyppin: interim present value +249, continuing value 16,800 discounting to 4,908, going concern 8,657, and the haircut ladder at ₹36.32 · ₹39.35 · ₹31.79 · ₹24.22 a share for p of 0.90, 1.00, 0.75 and 0.50. For Bharat Diversified: gross enterprise value 53,700, minority interest 5,600, net asset value 36,100 = ₹722.00 a share against a ₹505 price, and the bear-multiple rerun landing at ₹530.00, with the market marking every segment at bear multiples at once. For Sahyadri: net asset value 73,000, discount 60.0%, carry 2.397%, look-through price/earnings 7.93×, and the 60%-to-40% catalyst worth +50.0%. The binomial gives a risk-neutral probability of 0.5525, an option worth ₹619.63 crore, commitment at ₹452.34 crore and flexibility of ₹167.29 crore. The waterfall pays ₹238.4 crore to workmen and ₹5,561.6 crore to secured creditors, a 79.45% recovery, with zero below. Vindhya, on residual income, comes to ₹480.00 crore by all three routes in the stable case. In the fading case, explicit residual income of 14.400 · 12.084 · 9.401 · 6.356 · 2.959 present-values to 34.900, year-6 residual income is 3.087, and equity comes to ₹289.50 · ₹278.27 · ₹276.46 crore at persistence factors of 1.00, 0.60 and 0.00. The clean-surplus break is worth ₹12.81 crore, or 4.6% of the base case. For Fairmont's run-off, free cash flows of 163.92 · 143.19 · 124.73 · 108.30 · 93.70 present-value to 498.97. The trade-sale terminal of 316.36 is worth 200.96 today, against a 177.50 liquidation floor. Enterprise value is 699.93 and equity 579.93, or $14.50 a share, with the terminal only 28.7% of value. The same build on a −2% perpetuity reaches 863.77, 23.4% higher. Equity internal rates of return run 26.0% · 10.70% · 7.2% at prices of $10.00, $14.50 and $16.00.

Quantitative lens

The claim running through all of this is that the standard machine breaks in six places. You can make that sharper than "breaks". In each case there's a number that says how it breaks and how much the judgement input is worth, and those numbers reorder where the research goes. All figures below come from the engines above, run on the synthetic companies in the worked examples. They're properties of those inputs. They don't measure any real bank, cyclical or holdco.

On a bank, the standard machine returns no answer at all. The bank section says FCFF's ingredients dissolve, so price the dissolution. A bank funds itself at, say, 5.5% pre-tax and carries a 13% cost of equity. Its weighted average cost of capital at an equity multiplier of 6× is 5.604%, at 8× 5.234%, at 10× 5.013% and at 12× 4.865%. Set a mature growth rate of 5%, modest for a bank, and the Gordon denominator WACC − g runs +60 bp, +23 bp, +1.25 bp, and then goes negative. The crossing point is an equity multiplier of 10.14×, and Indian banks run 8 to 12×. So at ten times leverage the standard machine's denominator is a hundred and twenty-fifth of the correct one, and the "value" it prints is 640× too large. At twelve times it prints a negative number. That's no modelling inaccuracy you could correct with care. The machine's failure is structural, and it begins right at bank-like leverage, before you've even reached the fact that a bank has no meaningful EBITDA, no separable capex, and no working capital. The instruction to choose the model before opening the file now has a number attached.

And on a bank, forensics really is valuation. Here's the exchange rate. Differentiate the excess-return model with respect to book: ∂V/∂B = (ROE − g)/(r − g), which is P/B itself. On Vaigai that's 1.413, so every rupee of fictional book destroys ₹1.413 of market value. The fictitious rupee was also earning a fictitious two-point spread forever, and both die together. Now bring in leverage, and the number gets uncomfortable. At a 10× equity multiplier, one percentage point of the loan book held as standard when it isn't is 10% of book equity, so it's 10% of value, ₹2,826 crore on Vaigai. Against that, a full 100 basis points on the cost of equity costs 13.04%. At 10× leverage, then, one point of unrecognised bad loans is worth about 77 basis points of cost of equity, and at 8× leverage, 96 basis points. An analyst who spends an afternoon refining a bank's beta and ten minutes on the divergence report has allocated backwards. And 1.413× is a floor. A discovered overstatement usually means past earnings were overstated too, so the forward ROE that multiplies the surviving book is lower as well.

The survival haircut is an expected value nobody ever receives, and it's the smallest lever in the model. Zyppin's going concern is ₹39.35 a share and distress ₹9.09, so at p = 0.90 the expected value is ₹36.32. But the distribution has only two points. ₹36.32 occurs with probability zero. The median outcome is ₹39.35, and the other 10% of the time you hold ₹9.09. A number that can't occur is a poor thing to put on a football field without a note beside it. Then measure what p is worth: ∂V/∂p = (going concern − distress)/shares = ₹30.26 per unit, or ₹0.30 per percentage point. One percentage point of terminal market share, by contrast, is worth ₹1.87 a share, a factor of 6.2. So the input the young-company section rightly calls the step everyone skips is worth, on this model, about a sixth of the input everyone argues about. Don't take that as a reason to skip it. It tells you what the step is for. Its job is to force a number onto a page where it would otherwise stay silently at 1.0, and arithmetic precision is beside the point. The difference between a stated 0.90 and an unstated 1.0 is ₹3.03 a share. That's 8% of the value, and the kind of quiet optimism this material is written against. The decisive check is elsewhere. At the market's ₹95, the survival probability that would justify the price solves to 2.84, and there's no such probability. The disagreement with the market is entirely in the destination (share, take-rate, margin). Knowing which input holds the disagreement is the whole output of a Dark Side model.

A holding-company discount pays you nothing at all unless it narrows, and the arithmetic says how much and how fast. Sahyadri trades at a 60.0% discount, so the market pays ₹40 for every ₹100 of marked assets. Its holders receive 2.397% in dividends on price while owning a look-through earnings yield of 12.60%. So the cash that actually arrives is 19.0% of the earnings you notionally own, and the other 81% stays inside a structure with no terminal event. Generalise it. A holdco's dividend yield equals underlying yield × pass-through ÷ (1 − discount), so the discount is a yield multiplier: 1.43× at 30%, 1.61× at 38%, 2.50× at 60%. The break-even pass-through that leaves you no worse off than owning the operating companies directly is simply 1 − discount: 70%, 62%, 40%. That one line turns "it trades at a 60% discount" into a testable question with a number in it: does this holdco pass through more than 40% of what it receives? Then the catalyst. The 60%-to-40% narrowing above is worth +50.0% in total, but a total isn't a return. Over three years that's 14.5% a year. Over five it's 8.45%, over ten 4.14%, and over twenty 2.05%. The same re-rating that looks like an outstanding thesis at three years is a rounding error at twenty. And the fifth cause, no terminal event, means timing is the one variable you can't estimate. "Size the position to the catalyst's probability and ignore the discount's width" is right, and the table above shows why.

And the real-option double count has a price. The real-options section calls it the single most common sin: adding option value on top of a discounted cash flow that already contains the expansion's cash flows. On the pharma numbers above, the option to wait is worth ₹619.63 crore and committing today is worth ₹452.34 crore. They're alternatives. You either commit now or hold the right and decide later, so the correct answer is the larger, ₹619.63 crore. Adding them gives ₹1,071.96 crore, which overstates the truth by 73.0%. That's the size of the "strategic optionality" line in a valuation that has already booked the strategy.

Now the limits, and they carry more weight here than anywhere else in the phase, because every situation here is one where the machine is already known to be unreliable. The bank arithmetic uses a stylised funding cost and a single leverage number. A real bank's liabilities are a ladder of deposits, borrowings and capital instruments at different rates, and the point survives that detail intact. But the 640× figure is a property of these specific parameters, and it moves violently near the crossing. A near-degenerate denominator does that, so read the number as "the machine has no answer" and never as an estimate. The 1.413× book-quality exchange rate is a derivative of a steady-state model at one point. It's a local sensitivity and not a schedule, and a large enough write-down changes the growth and spread assumptions that produced it. Zyppin's ₹0.30 per point of survival probability depends entirely on the gap between the going-concern and distress values, which is itself two guesses. The two-point distribution also simplifies a continuum of outcomes between "thrives" and "wound up", and the real distribution has a long middle that the binary framing erases. The holdco yield multiplier assumes the discount is stable and the pass-through persists. Both are decisions taken by people whose interests, as the holdco section warned, may not be yours. A promoter who cuts the payout moves your carry without touching the discount at all. The catalyst table prices a given narrowing over a given horizon. It says nothing about the probability of either, and putting a probability on an Indian promoter's decision to simplify a structure isn't a quantitative exercise. Every worked company above is synthetic, with clean books and a known distress value. That's the opposite of the situations this material exists for. In real special situations the book is the thing in doubt, the mid-cycle is the thing being argued about, and the distress value is the thing nobody computes until it matters. Most importantly, none of these numbers is a verdict. The narrative loop's whole purpose is that the output of a special-situations model is a conditional sentence with published falsifiers. Everything above serves the job of writing that sentence more precisely. Knowing that a rupee of hidden book costs 1.41 rupees tells you where to look. It doesn't tell you what you'll find.

Do it in code: choose the model before opening the file, and let a mismatched one fail an assertion. Assert the DDM and excess-return equivalence instead of trusting it. Invert every model for the input the price implies. Report the survival haircut as a distribution with its median beside its mean. And convert any discount into its yield multiplier and its break-even pass-through before you call it an opportunity. See G1/DV1.02, G1/QM1.01 and G3/CN3.03.

Where this goes next: galaxy cross-links

• G1 · DV1.02 Options Pricing & the Greeks: the binomial sketch at full depth, and Merton's insight that levered equity is a call on the firm's assets: what actually drives an out-of-the-money distressed stock's price, and why volatility transfers value from creditors to shareholders.
• G1 · QM1.01 Probability, Distributions & Estimation: expected value over a bimodal outcome, why the mean of a two-point distribution is a number that never occurs, and how to report a haircut model so nobody reads the mean as a forecast.
• G1 · QM1.02 Hypothesis Testing & Regression: the P/B-on-ROE sector fit the bank section asks for, with the standard errors and degrees of freedom that decide whether a bank sitting below the line is a finding or a fitting artefact.
• G3 · CN3.03 When Frameworks Mislead: the consulting-side twin of the opening triage: recognising the situation your standard tool was not built for, before you have spent a week producing a confident wrong answer with it.
• G1 · AL1.01 Hedge Fund Strategies: distressed and event-driven investing as the credit skill the distressed section calls it: buying at the right rung of the queue, and how the professionals size the catalyst risk that a holdco discount and a resolution plan both come down to.

Teach it back & journal

Feynman prompt. Explain three things to a smart 15-year-old, using a lemonade stand, a neighborhood money-lender, and a beach umbrella-rental cart. (1) Why is the money-lender valued on what she has (her lending float and the extra she earns on it), while the lemonade stand is valued on what it makes? (2) Explain why the umbrella cart's amazing monsoon-year profit isn't its real profit, and how you'd find the real one. (3) The lender's cousin owns a locked box holding shares of the lender "worth ₹100." Why will he only ever get ₹50 if he tries to sell the box, and what could unlock it? Then write the one-page explainer: "When the standard tools lie, and what to reach for instead." No formulas are allowed except book + extra earnings. If a sentence needs "WACC" to survive, you haven't understood it yet.

Journal reflection. After both capstone memos are graded, answer in writing: (1) Which single assumption in each model did you most want to nudge upward, and what was the story you were protecting? (2) Your two companies' reverse DCFs each told you what the market believes. Write down, honestly, whether you found yourself adjusting your band toward the market's number after seeing it (anchoring). That's the bias your falsifiers exist to contain. (3) Which special-situation row (bank, cyclical, young, option, SOTP, distressed) do you find most temperamentally tempting, and which most repels you? Your future mistakes are hiding in the first answer, and your future missed opportunities in the second. File both memos, with rubric scores and falsifier watch-lists, in your company files, since Phase 8 will reuse them.

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