DCF II: Terminal Value & Reverse DCF

Phase 3 — Corporate Finance & Valuation · Week 30 · ~20 hours (one immersive week)

Where this sits: the cash-flow work before this taught you to build the numerator: FCFF defined properly, forecasts disciplined by g = reinvestment rate × ROIC and sales-to-capital. The cost-of-capital work gave you the denominator, a WACC you can defend input by input. This module is where the machine finally produces an answer: a terminal value built under constraints you can derive, an enterprise value carried across the bridge to a per-share number, and the same machine run backwards, to read what a market price is silently assuming. That last one changes how you see every stock quote for the rest of your life. This is a crown module: the mastery gate is ≥90%. The reverse DCF you learn here becomes question 5 of the five-question spine ("what is the price implying, and is that beatable?") and Day 2 of the deep-dive playbook in Phase 8. Next: relative valuation, which will lean constantly on the multiple-inversion algebra you build here.

Learning objectives

By the end you can:

1. Compute a Gordon terminal value from a steady-state FCFF, TV = FCFF_T+1 / (WACC − g), and derive, not recite, its three constraints: g capped by the economy (≈ the currency's risk-free rate), reinvestment forced to g / ROIC_terminal, and terminal ROIC faded toward WACC unless a named moat defends it.
2. Prove algebraically that growth adds nothing when ROIC = WACC, and quantify how much terminal value an analyst manufactures by ignoring terminal reinvestment ("free growth").
3. Cross-check a Gordon terminal value with an exit multiple in both directions: compute the implied exit multiple from your TV, and extract the perpetual growth a quoted exit multiple smuggles in. Then explain why the exit multiple can never be the primary method in an intrinsic valuation.
4. Assemble a complete DCF to a per-share value: explicit-period PVs, terminal value, mid-year convention, the enterprise-to-equity bridge (cash, debt, minorities), diluted shares. Then report the output as a defensible range built from a 2-way sensitivity grid and coherent scenarios, not a point.
5. Run a probabilistic read: build and interpret a two-way sensitivity table, weight internally consistent scenarios, and explain what a Monte-Carlo output distribution does and does not tell you.
6. Execute a full reverse DCF: hold the valuation machine fixed, solve by hand-iteration for the growth (or margin, or duration, or required return) that reproduces the market price, translate it into operational language, and judge it against base rates. That is the expectations-investing discipline.
7. Value a dividend payer with the DDM and the H-model (formula, worked example, error bounds), and explain why banks get dividend-based valuation rather than FCFF (previewing M3.10 and M5.01).
8. Audit any DCF, yours or a sell-side analyst's, against the 15 classic mistakes, naming the one-line fix for each.

Prerequisites & connections

Builds directly on:

• M3.01: the growing perpetuity PV = CF₁/(r − g) is the entire mathematical content of terminal value; everything else here is a set of rules about what CF₁, r, and g are allowed to be. Fisher (real vs nominal) returns when we cap g.
• M3.03: the two WACCs you built from scratch (≈11.0% INR for the UltraTech-style cement leader, ≈7.8% USD for the Home Depot-scale retailer) are used as-is here. M3.03 promised that "M3.05 forces ±1-point WACC sensitivity on every DCF"; this week keeps that promise.
• M3.04: FCFF construction, the g = reinvestment rate × ROIC identity, and sales-to-capital forecasting. The explicit-period forecasts here use that same machinery.
• M2.03/M2.04: ROIC built from raw statements, and FCF conversion; the terminal-value constraints are those ideas pushed to infinity.

Feeds into:

• M3.06: relative valuation. The steady-state-multiple algebra you derive here (what P/E or EV/EBIT a perpetuity can justify) is the intrinsic anchor for every comp you'll ever run.
• M3.09/M3.10: the full modeled DCF, and the special cases (banks, cyclicals, money-losers) where today's machinery must be modified.
• Phase 8: Day 2 of the deep dive ends with a reverse DCF; the 12-question teardown's question 11 ("what is the market implying?") is this week's work compressed to five minutes.
• The weekly teardowns from this week onward add a "what's priced in" line. The master schedule flags week 26 as the point where that becomes mandatory.

Core teaching content

4.1 Where a DCF is won or lost

Run the arithmetic on any completed DCF and one fact jumps out. The terminal value, the single number capitalizing everything past your forecast horizon, is typically 60–80% of the enterprise value. In the two builds you'll complete this week, the terminal value is 64% of the Indian company's value and 62% of the American one's; shorten the explicit horizon to five years and it rises toward 80%:

Beginners tend to react to this in one of two wrong ways. Some decide the DCF is a sham: "it's all terminal value, so the whole thing is one made-up number." Others never notice. They spend forty hours polishing quarterly working-capital forecasts, while terminal assumptions they set in ninety seconds decide three-quarters of the answer.

An expert starts somewhere else, with a fact about economics. Businesses are long-duration assets. A going concern earns most of its lifetime cash far beyond any horizon you can forecast. So the terminal value isn't a fudge bolted onto a model. It's a plain admission that most of the value of owning a business lies in the distant future. That's also why equities are volatile. Long-duration assets reprice hard when discount rates or long-run expectations move, and you've already met this duration logic once. When someone tells you "TV dominates", then, don't despair. Spend your care where the value is. In practice that means three things, and they're the three acts of the week:

1. Constrain the terminal value, so it can only say things that are economically possible.
2. Report ranges. A number that is 70% long-run assumption deserves error bars.
3. Invert the machine. The terminal assumptions drive the value, so the market price is mostly a statement about those assumptions, and you can read it.

Here's one habit to start now. Every DCF you build should report "PV of TV as % of EV" as a standard output. A high number isn't wrong in itself, and for a 10-year horizon, 55 to 75% is normal. You're watching for an unusual number, because that one is telling you something. Above about 85%, your explicit forecast is adding almost nothing. Either the horizon is too short for the story, or reinvestment is crushing your near-term FCFF. That can be correct (Amazon looked like this for years), but you should know it's happening. Below about 40%, the horizon is very long, or you're valuing a melting ice cube, or your terminal assumptions are unusually harsh. Think of the ratio as a smoke detector. It tells you where to look, and the verdict is still yours to reach.

4.7 The mid-year convention: a small honest upgrade

The standard DCF discounts year 1's FCFF by (1+WACC)¹, as if the entire year's cash arrives on December 31. Real cash arrives throughout the year, roughly evenly for most businesses. So end-of-year discounting systematically understates value, by about half a year's discounting on every flow. To fix it, discount each flow at t − 0.5. Or, equivalently, compute the whole end-of-year EV and multiply once:

EV_mid-year = EV_end-of-year × (1 + WACC)^0.5

At WACC 7.8% the factor is 1.0383 (+3.8%). At 11% it's 1.0536 (+5.4%). The gain looks small. It's still larger than most inputs people argue about for an hour, and it's systematic. Skip it and every valuation you ever do is biased in the same direction. Three notes. (1) Apply the factor to the enterprise value only, meaning the explicit PVs plus the PV of TV. The bridge items (cash, debt, minorities) are stocks measured today and take no adjustment. So when debt is large, the per-share uplift can be bigger than the factor itself. (2) Skip the convention where cash really does bunch, or when you want deliberate conservatism, but say so when you do. A hyper-seasonal retailer collecting in Q4 barely deserves half of it. (3) State the convention once, at the top of the model, the same way you state your net-debt convention.

From here on, mid-year is the default everywhere. Every build, worked example, practice solution and mastery key applies the factor unless it says otherwise in the same breath, and so does the full DCF workbook later on. Year-end discounting is the stated exception. You'd use it where cash really bunches, where a lender or a mandate requires the conservative read, or where a comparison has to run against someone else's year-end model. Two habits keep the exception safe. Apply the convention to the terminal value as well as the explicit flows, because a mid-year explicit period bolted onto a year-end terminal value is neither convention. And never let the choice pass for insight. A 5.4% uplift is a 5.4% uplift, and that's all it is.

4.8 The last mile: from enterprise value to a per-share number

The DCF's PVs add up to the value of the operations for all claimholders. Shareholders get what's left after everyone senior is paid, plus whatever the operations didn't need. You'll build the bridge by hand once here, and the full DCF workbook automates it later:

Equity value = EV (mid-year adjusted)
             + cash & non-operating assets      (excess cash, listed investments at market,
                                                 idle land — anything your FCFF never used)
             − debt (market value; incl. lease liabilities if — as in M3.03 — your WACC's
                                                 D included them and EBIT was post-rent)
             − minority interest                (consolidated NOPAT you don't own; at market
                                                 or an earnings-multiple estimate, not book)
             − preferred, unfunded pensions, and other debt-like claims

Per share = Equity value ÷ DILUTED shares outstanding (options/RSUs in; M1.09's machinery)

A few notes on the bridge. Count everything once and only once. Cash shouldn't sit inside your FCFF forecasts, because interest income is non-operating, but if it did, don't add it again. If you netted cash against debt in the WACC weights, keep the same convention here. Take non-operating assets at market, with a haircut for tax and access. An Indian promoter-controlled company's cross-holdings may deserve a 20 to 30% holding-company discount, which comes back with the special cases. Minorities come off because consolidation put 100% of the subsidiaries' NOPAT into your FCFF, and you own less than that. And per-share comparisons use the diluted count from the same date as the price quote.

Now we run the whole machine, end to end, twice.

4.9 End-to-end build 1: the UltraTech-style cement leader (INR)

This is the company whose 11.0% INR WACC you assembled from a blank sheet. All figures are illustrative, modeled on ~FY24/FY25 scale. It's a teaching specimen, so don't read it as a recommendation, and pull the actual annual report, screener.in and live rate pages before you believe any number. The setup carries over from your WACC build and your forecasting method:

Run the sanity checks before computing, the same reflexes as in your forecasts. Year-1 reinvestment = ΔRev/1.2 = 7,200/1.2 = ₹6,000 cr, so RR = 6,000/9,900 = 61%. Implied incremental ROIC = g/RR = 10%/61% ≈ 16.5%, near the recently delivered ROIC ≈ 15%. That's consistent, so carry on.

The engine room. In each row revenue grows, EBIT = margin × revenue, NOPAT = EBIT × 0.75, reinvestment = ΔRevenue ÷ 1.2, FCFF = NOPAT − reinvestment, and the PV is taken at 11%:

All figures are ₹ cr. Watch the shape. As growth fades, reinvestment releases and FCFF accelerates even while NOPAT growth slows. That's the signature of a maturing, capital-intensive business.

Terminal value, by the rules. NOPAT₁₁ = 20,035 × 1.045 = ₹20,937 cr. RR = 4.5/12 = 37.5%. FCFF₁₁ = 20,937 × 0.625 = ₹13,086 cr. Now the steady-state audit. Year 10 growth = 4.5% = terminal g. Margins are flat, and tax is at the marginal rate. Reinvestment jumps from 5,752 to 20,937×0.375 = 7,851. That's because the year-10 fade left reinvestment below steady state, and the TV formula correctly puts it back instead of capitalizing the holiday.

TV₁₀   = 13,086 / (0.11 − 0.045) = ₹2,01,317 cr      (implied exit: 7.5× EBIT₁₀ — plausible for a
PV(TV) = 2,01,317 / 1.11¹⁰       = ₹70,901 cr          steady cement asset; cross-check passed)

Assembly.

PV explicit (Σ)            ₹39,793 cr
PV terminal                ₹70,901 cr        → TV share = 70,901/1,10,694 = 64% (normal band)
EV (end-year)              ₹1,10,694 cr
× mid-year 1.11^0.5        ₹1,16,623 cr
+ cash & investments 8,000 − debt 20,000 − minorities 1,000
Equity value               ₹1,03,623 cr
÷ 29 cr diluted shares     ≈ ₹3,570 per share

The sensitivity grid. Here's the promised ±1 point of WACC, crossed with terminal g, in ₹ per share:

*The 5.0% column exceeds the cap. It's shown on purpose so you can see how little it changes, and it stays a red-flagged column, never a base case.

Read the grid the way an analyst would. The defensible range is roughly ₹3,000 to 4,400, with a base of ₹3,570. WACC, in the rows, drives most of the swing. Terminal g, in the columns, barely moves it, because the reinvestment constraint has muzzled g. In a free-growth model the columns would explode. Now look at something the grid can't hide. The market price is ₹11,034 (₹3,20,000 cr ÷ 29 cr), and your most generous corner is ₹4,493. Even a heroic, hand-tuned bull machine (12% growth, margins to 20%, ROIC_T 15%) reaches only ≈ ₹5,466.

You haven't made an arithmetic error. With a machine you can defend input by input, you've found that the price is paying for something your defensible assumptions can't reproduce. It could be a longer super-normal runway. It could be consolidation pricing power, or a structurally lower required return. Or it could be an error, the market's or yours, so check the variant possibilities. Is capacity-led revenue about to step-change? Is your margin normalization stale? The reverse DCF, a few sections on, will measure which of these it is. Until then, hold the discomfort. It's the most instructive feeling in valuation, and in India's premium-multiple regime you'll feel it often.

4.10 End-to-end build 2: the Home Depot-scale retailer (USD)

Same machine, with the American economy's parameters and the 7.8% USD WACC from your build. Illustrative, at ~FY24/FY25 scale, so pull the 10-K from EDGAR. Here are the inputs.

• Revenue $153bn, EBIT margin 14.1% flat, tax 25%.
• Sales-to-capital 3.0, from a mature box network plus high-return pro-supply investments. Implied incremental ROIC = 3.0 × 14.1% × 0.75 ≈ 32%, consistent with the ~30%+ delivered, which points to a real scale and density moat.
• Growth 5% in yrs 1 to 5, from housing-stock aging plus pro-market share, fading to 2.7% by year 10.
• Terminal g 2.7% against a cap of 4.2%. That's roughly inflation plus a whisker, right for a mature format.
• ROIC_T 18%. The moat is named (density economics + brand + pro ecosystem), and it's still well faded from 30%+.
• Bridge: cash $4bn, debt incl. leases $50bn, 0.99bn diluted shares.

All figures in $bn. It's the same maturing shape, with a gentler slope.

Terminal: NOPAT₁₁ = 24.67 × 1.027 = $25.33bn. RR = 2.7/18 = 15%. FCFF₁₁ = $21.53bn. TV = 21.53/0.051 = $422.2bn. The implied exit is 12.8× EBIT₁₀, a rational buyer's multiple for a moaty mature retailer, so the cross-check passes. PV(TV) = 422.2/1.078¹⁰ = $199.2bn, a TV share of 62%.

EV end-year = 120.2 + 199.2 = $319.4bn  → × 1.0383 = $331.6bn
+ 4 − 50 = equity $285.6bn  → ÷ 0.99bn shares ≈ $288 per share      (market price ≈ $354)

Two-way grid #2. This time it's the operating drivers, in $ per share, with WACC 7.8% and g 2.7% held:

The WACC×g grid spans $226 (8.8%, 2.2%) to $400 (6.8%, 3.2%). So the defensible range is ≈ $235 to $350, the base is $288, and the $354 price is at the top edge, slightly beyond it. The diagnosis differs from the Indian build. Here the price is reachable, but only in the corner where nearly everything goes right: 7% growth and margin expansion, or a sub-7% required return. Solve it properly with the reverse-DCF method and the price implies ≈8.6% near-term growth on this machine, against the ~5% you can underwrite.

Two companies, two verdicts, one lesson. The DCF's job was never to tell you the "right price". Its job is to tell you what the current price asks you to believe, and to give you a range within which you'd be a buyer. And no, the machine doesn't always say "expensive". Run it on the same specimens' March-2020 or 2009 prices and the base case clears the market with room to spare. The output tracks the asking price, and mid-2026's asking prices for premium franchises are high. The practice set includes a case priced below its no-growth value, so you'll see the other side too.

Common mistakes & how experts think differently

The 15 classic DCF mistakes. Here are fifteen ways real DCFs go wrong, each with its one-line fix. The first seven are in the terminal value, where the money is. The rest are in assembly and mindset. The mastery check tests all fifteen, the app's error-spotting deck drills them, and your Phase-8 checklist inherits them. An expert can audit a stranger's DCF against this list in four minutes. From this week, so can you.

Beyond the checklist, experts think differently in four ways.

The terminal value gets the most senior hour, not the last five minutes. Beginners burn their care on year-2 gross margin, which is worth rupees, and default the TV, which is worth the company. Experts turn that around. The three terminal dials (g, ROIC_T, and the horizon at which steady state begins) get the evidence review, the debate, and the documentation, because the value is there, as the first section showed.

They read every DCF backwards first. Hand an expert anyone's model and these are the first four moves. PV(TV) as % of EV is the smoke test. The implied exit multiple asks whether a rational buyer would pay it. The implied perpetual g, if the TV were re-expressed as Gordon, catches smuggled growth. And the WACC−g gap shows whether the denominator has degenerated. Four numbers and ninety seconds catch most broken models before anyone looks at a single forecast row.

They treat the output as a distribution and the price as a data point. The deliverable is never "₹3,570". It's "₹3,000 to 4,400 under assumptions I can defend. The price is ₹11,034. The gap is a duration-of-moat bet I can now research." Value is a range, and the price is the precise number. Respect each for what it is.

They let the market speak first. The reverse DCF runs before the forward one hardens into a position. The market isn't necessarily right. But when you know what you're disagreeing with, in units of growth, duration, and required return, you have a variant perception. Without that, all you have is a mood.

Worked examples

Worked example 1: Full reverse DCF, India: a Nestlé-India-style listed FMCG

The specimen: India's archetypal premium-multiple franchise. It has dominant brands (dairy, nutrition, noodles, coffee), decades of pricing power, capital-light economics, near-zero debt. All figures are illustrative, modeled on ~FY24/FY25 scale. It's a teaching specimen, so don't read it as a recommendation, and pull the actual AR and screener.in. Market cap ≈ ₹2,30,000 cr, net cash ≈ ₹2,500 cr and 96.4 cr shares give a price ≈ ₹2,386. Revenue ≈ ₹20,000 cr, with an EBIT margin of 22% and tax at 25%, gives NOPAT ≈ ₹3,300 cr. So the market pays EV/NOPAT ≈ (2,30,000 − 2,500)/3,300 ≈ 69×. Your steady-state anchor says a steady state supports ~15×, so you already know a super-normal story is priced in. The reverse DCF tells you which story.

Step 1: hold the machine. The company is essentially all-equity, so WACC = r_e = 4.5% + 0.8 × 7.1% ≈ 10.2%, built the usual way for a low-beta staple. Margins are held at 22%, which grants permanent brand economics. Sales-to-capital is 3.0, so incremental ROIC ≈ 3.0 × 16.5% ≈ 50%. That's Nestlé-grade capital-lightness, and we grant it. In the terminal state, g = 4.5%, the full cap, granted. ROIC_T = 25% (WACC + ~15), an extraordinarily generous "brands are forever" grant. Every contested dial except the solve lever is set in the market's favor. Target: EV = ₹2,27,500 cr.

Step 2: the lever is near-term revenue growth, on the standard machine. That means g_hi for years 1 to 5, a linear fade to 4.5% by year 10, then terminal.

Step 3: bracket and iterate. Each row is one forward run of the cement-build engine. Do two of them by hand so you believe them:

The 10-year machine can't reach the price at all. Even 25%-a-year growth for five years, which triples revenue, recovers barely two-thirds of the market cap. A 4.5%-capped terminal state pins the far future, and 10.2% discounting crushes a decade. That's the first finding, and it's useful, because it decodes the price instead of ending the analysis. The price can't mainly be a bet on near-term growth, so it must be a bet on duration. Time to move the lever the market is actually using.

Step 3′, re-solve on duration with a 20-year machine. Hold g_hi for years 1 to 10 and fade it to 4.5% across years 11 to 20, which grants a two-decade competitive-advantage period:

Interpolate and refine, and you get implied g_hi ≈ 19% (19.06% to the model's precision).

Step 4: translate. Nineteen percent for a decade, then a decade-long glide. Revenue reaches ≈ ₹1,14,000 cr by year 10 (5.7× today) and ≈ ₹3,24,000 cr by year 20 (16×), with 22% EBIT margins and 25% steady-state ROIC throughout. For scale, ₹1,14,000 cr is larger than the entire current Indian packaged-foods market by most counts. One company, bigger than today's whole industry. Now try the other levers, for the three-lever diagnosis. Hold growth at the delivered ~11% and solve for the required return instead, and you get implied r ≈ 6.4%. So the price is also consistent with owners accepting a G-sec-minus return on a consumer franchise. The margin lever alone would need ~60% EBIT margins, which no consumer company in history has reached, so discard it.

Step 5: judge against base rates. Indian listed consumer staples' delivered 10-year revenue CAGRs cluster in the 8 to 13% band. Export the NSE-FMCG set from screener.in and count. Nominal GDP at ~10 to 11% is the center of gravity, and this company's own delivered decade ran at ~10 to 11%. How many companies sustained 19% for ten years from a ₹20,000 cr base? The club is approximately empty. At that scale, double-digit-teens growth for a decade has belonged to a handful of lenders and platform businesses. Staples haven't done it. Verdict: the expectations gap is against the buyer on every lever. The price requires growth ~2× the best base rate at this scale, or a two-decade zero-fade advantage period plus mid-teens growth, or a required return below the government bond. For calibration, the forward machine at delivered-history assumptions (11% for ten years, fading by year 15) values the share at ≈ ₹986 against ₹2,386. The price carries roughly 2.4× the value your evidence supports.

What this does and doesn't conclude. It doesn't say "short it". Bond-proxy bids, index flows and scarcity premia can hold such prices for years, and they have. Expectations analysis times nothing. What it does say, with numbers you can defend line by line, is that whoever buys at ₹2,386 is underwriting near-perfection for two decades or accepting a bond-like return. It also hands you a research agenda. Is there a categorically new growth leg, such as nutrition premiumization or a rural distribution step-change, that breaks the base rate? And it names the bull rebuttal you have to steelman. That's the Marcellus-school argument that elite Indian franchises have historically defied fade for 20+ years, which would make duration the right thing to pay for. It's the contested debate at the heart of India's premium-multiple regime. Be evenhanded, and go and look at the persistence data yourself in Phase 4. One tool, and a shouting match has become a measurable disagreement.

Worked example 2: Full reverse DCF, US: an Apple-scale devices-and-services giant

The specimen: the world's most valuable consumer-technology franchise, a hardware installed base feeding a high-margin services annuity. Illustrative, at ~FY24/FY25 scale, so pull the 10-K. Market cap ≈ $3,200bn, net cash ≈ $50bn and 15.0bn diluted shares give a price ≈ $213. Revenue ≈ $400bn, with an EBIT margin of 31.5% and tax at 20%, gives NOPAT ≈ $101bn. That puts EV/NOPAT ≈ 3,150/101 ≈ 31×, half the Indian specimen's and double the steady-state anchor.

Step 1: hold the machine. WACC ≈ 8.7%, built the usual way: Rf 4.2%, β ≈ 1.1, ERP 4.3% and a trivial debt weight. Margin held at 31.5%. Sales-to-capital is 4.0 because manufacturing is outsourced and working capital is negative, which implies incremental ROIC ≈ 100%. We grant it, because the ecosystem really is that capital-light. In the terminal state, g is 2.7% and ROIC_T is 30%, a monumental grant of an eternal ecosystem moat. Target: EV = $3,150bn.

Steps 2–3: solve for growth. The 10-year machine runs g_hi in years 1 to 5 and fades to 2.7% by year 10:

Implied g_hi ≈ 14.4%. That's a 10-year revenue CAGR ≈ 10.8%, taking revenue to ≈ $1,113bn by year 10, with 31.5% margins throughout. Unlike the Indian case, the 10-year machine can reach this price. The expectations are enormous, but arithmetic doesn't rule them out.

Step 4: translate, on all three levers. Start with growth. $1.1 trillion of revenue is ~1.5× the largest revenue base any company reports (Amazon's ~$717bn for 2025, just ahead of Walmart's ~$713bn, both at margins far below 31.5%; as of February 2026 filings). This specimen's delivered history is a ~5 to 6% five-year CAGR. Next, margin alone. Hold growth at the ~5% you can underwrite and the implied EBIT margin is ≈ 52%. No hardware-anchored company has ever printed that, so discard margin as a single lever. Then the required return. At 5% growth and 31.5% margins, the price is consistent with r ≈ 6.3%. The marginal buyer of the world's favorite mega-cap may simply be charging ~2.4 points less than your CAPM does (index flows, safety premium, buyback bid). The joint readings are the realistic frontier. These pairs reproduce the price exactly:

Now you can read the price as a menu of stories. Today's price = "services/AI mix lifts margins to ~40% AND revenue compounds ~10% near-term," or "margins hold and growth doubles its delivered rate," or "a 6.3% discount rate." You can test each one, and none of them comes free.

Step 5: base rates and verdict. Pull the PDF and look at The Base Rate Book's large-cap tables. Among companies starting above ~$50bn revenue, 10-year sales CAGRs above 10% happened in roughly one case in twenty, and no case began from $400bn. The base-rate table literally runs out of precedents. You're being asked to underwrite an event outside the recorded distribution. That isn't impossible, since this company has broken base rates before, but you have to name it as such. The margin-mix story is the strongest leg. Services gross margins are real and growing, so the 40%-margin column isn't fantasy, but it still needs ~10% growth beside it. Verdict: the expectations gap is against the buyer at a CAPM-grade required return, and the price is best read as quality-duration + margin-mix + a structurally lower r. The research agenda writes itself. Track services mix and margin each quarter against the 40%-column trajectory. The price has told you which KPI to watch. Compare the two examples. The Indian franchise's price was mostly duration. The American one's is mostly margin mix plus required return. Same tool, different decoded messages, and that specificity is what makes it the signature instrument.

Worked example 3: Terminal-value surgery: repairing a broken TV step by step

A sell-side intern hands you this terminal value for the UltraTech-style company, at WACC 11%: "FCFF₁₀ is ₹14,284 cr; grow at 6% India-forever growth: TV = 14,284 × 1.06 / (0.11 − 0.06) = ₹3,02,821 cr." That's a PV of 3,02,821/1.11¹⁰ = ₹1,06,650 cr, against the ₹70,901 cr from our own cement build. Repair it one violation at a time, and put a number on each:

1. Mistake #1 (g above the cap). 6% > the 4.5% INR cap. Re-run at g = 4.5%: TV = 14,284 × 1.045/0.065 = ₹2,29,643 cr. Those illegal 1.5 points of g had manufactured ₹73,178 cr out of nothing. That's a quarter of the printed terminal value (24.2%) and a third of the repaired one (31.9%).
2. Mistake #2 (free growth). The numerator grew FCFF with reinvestment frozen at year-10's fade-depressed level. Rebuild from NOPAT: FCFF₁₁ = 20,937 × (1 − 0.045/0.12) = ₹13,086 cr, so TV = 13,086/0.065 = ₹2,01,317 cr. Free growth was worth another ₹28,000 cr of fiction.
3. Mistake #4 check (steady state). Run the year-10 audit. Growth of 4.5% equals terminal g. The margin is flat, tax is at the marginal rate, and reinvestment is now back at g/ROIC_T. It passes.
4. Cross-check against the exit multiple. Implied exit = 2,01,317/26,714 = 7.5× EBIT, a plausible year-10 private-market multiple for a mature cement leader. The intern's original implied 11.3× while assuming 6% perpetual growth. That multiple and that growth rate don't describe the same company.

Repaired TV: ₹2,01,317 cr, PV ₹70,901 cr. The two five-second errors had overstated the company's terminal value by 50% and its enterprise value by roughly a third. The repair took four minutes, and they're the highest-paid four minutes in analysis. Practice P2 has you run this surgery on your own.

Worked example 4: Gordon vs. exit multiple, reconciled (INR)

The specimen: Nashik Components (synthetic), a mid-cap Indian auto-ancillary at the end of its 10-year forecast. You have to set the terminal value and stress it against the exit multiple a sell-side model would paste on. All figures are synthetic, and the arithmetic is the lesson. Terminal-year NOPAT₁₀ = ₹750 cr and EBIT₁₀ = ₹1,000 cr (tax 25%). WACC is 11%, terminal g 4.5% (the full INR cap), and ROIC_T 14%, which is WACC + 3 for a graded tooling-switching-cost moat.

Step 1: the Gordon TV, by the rules. NOPAT₁₁ = 750 × 1.045 = ₹783.75 cr. RR = g/ROIC_T = 4.5/14 = 32.14%. FCFF₁₁ = 783.75 × (1 − 0.3214) = ₹531.83 cr. TV₁₀ = 531.83/(0.11 − 0.045) = ₹8,182 cr.

Step 2: read its implied exit multiple, forward direction. Divide the Gordon TV by the year-10 metric a comp is quoted on: 8,182/1,000 = 8.18× EV/EBIT. Would a rational year-10 buyer pay ~8× EBIT for a mature auto-components business? Yes. Steady industrial cash-compounders change hands at about that level. The cross-check passes, and your TV describes a company someone would actually buy.

Step 3: invert a banker's multiple, reverse direction. The sell-side model pastes on a "conservative-sounding" 13× EV/EBIT instead. That gives TV = 13 × 1,000 = ₹13,000 cr, 59% above your Gordon figure. Now solve for the perpetual growth 13× smuggles in, 750 × (1+g) × (1 − g/0.14)/(0.11 − g) = 13,000, by bracketing. At g = 8.5% you get ₹12,788 cr, just low. At g = 9% you get ₹14,598 cr, too high. Interpolate and refine, and you land on g ≈ 8.6% forever.

The verdict. 8.6% perpetual growth is nearly double the 4.5% INR cap. Somebody has assumed, without saying so, that an auto-parts maker will outgrow the whole Indian economy until the end of time. It's a g-cap violation in a trench coat. The multiple felt moderate ("only 13×"), and the assumption inside it can't be defended. Keep the Gordon ₹8,182 cr as primary and log the 8.18× as the cross-check. If comps really do trade at 13× today, that's a fact about the current market, not about year-10 steady state.

Worked example 5: Reverse DCF on a mid-cap: solving the growth a price implies (USD)

The specimen: Cedar Ridge Instruments (synthetic), an unglamorous US precision-instruments mid-cap. A case like this gets the most out of a reverse DCF, because there's no famous narrative to anchor on. All figures are synthetic. Revenue is $12.0bn, the EBIT margin 20% and tax 25%, so current NOPAT is $1.8bn. Sales-to-capital is 2.5, WACC 8.5% and terminal g 2.7% (cap 4.2%). ROIC_T is 16%, a named metrology moat faded from a higher current print. The market values the enterprise at 22× current NOPAT = $39.6bn. Your steady-state anchor of ~15× for a mature business already whispers "super-normal period priced." The reverse DCF tells you how much.

Step 1: hold the machine. Freeze WACC, tax, margin, sales-to-capital and all three terminal rules. The only free dial is near-term revenue growth: g_hi for years 1 to 5, fading linearly to 2.7% by year 10, the same engine as the cement build. Discounting is mid-year, the default from here on, and it applies to the terminal value as well. Target output: EV = $39.6bn.

Steps 2–3: bracket and interpolate. Each row is one forward run of the standard machine. The end-of-year EV is shown alongside so you can see what the convention is worth:

The 10% row now overshoots by more than four percent. Interpolate down from it and you get implied g_hi ≈ 9.1%. Read the fourth column before you move on, because it's a lesson in itself. On end-of-year discounting, the same price would read as 9.9% of implied growth. The convention is worth 0.8 points of the answer. That's more than most of the judgments an analyst agonizes over, and it's the reason you state the convention instead of assuming the reader shares it.

Step 4: translate onto the three levers. Growth first. 9.1% for five years, then a fade, takes revenue to $18.6bn by year 5 (1.55× today) and $24.0bn by year 10 (2.0× today), at a 20% margin throughout. That's about 1.8× the ~5% delivered. Then required return. Hold growth at the delivered 5% and solve for the discount rate instead, and the price is just as consistent with WACC ≈ 7.4%. That's a marginal buyer charging ~1.1 points under your CAPM. The margin is already a healthy 20%, so growth and required return are the live readings.

Step 5: judge against base rates. A US instruments mid-cap sustaining ~9% revenue growth for a decade is uncommon, but it isn't extraordinary. It's in the upper third of the base-rate distribution, well short of its end. Compare the Indian FMCG in the first worked example, whose price was arithmetically unreachable on any 10-year growth rate. Verdict: the expectations gap is mildly against the buyer. The price wants about 1.8× the delivered growth, or a modestly cheaper required return, but it's inside the plausible band. This is the reverse DCF's most common and most valuable output. It won't shout "wildly mispriced". It says "here is what you must believe; it's a stretch rather than a fantasy, so size the position accordingly."

Practice set

Work each problem all the way through before you open its solution (working agreement #3). Currency is ₹ cr unless stated. "Machine" always means the same thing: FCFF = NOPAT − reinvestment, reinvestment = ΔRevenue ÷ sales-to-capital, terminal FCFF from NOPAT with RR = g/ROIC_T, and end-of-year discounting unless you're told otherwise. Problems 1 to 4 are guided, 5 to 9 independent, 10 to 12 exam-condition (P12 timed).

P1 (guided: the legal numerator). Meridian Foods (synthetic) ends its 10-year forecast with NOPAT₁₀ = ₹400 cr. WACC 10%, terminal g 4%, ROIC_T 16%. Compute the terminal reinvestment rate, FCFF₁₁, TV₁₀, and PV(TV).

Solution. RR = g/ROIC_T = 4/16 = 25%. NOPAT₁₁ = 400 × 1.04 = 416, so FCFF₁₁ = 416 × 0.75 = ₹312 cr. TV₁₀ = 312/(0.10 − 0.04) = ₹5,200 cr. PV = 5,200/1.10¹⁰ = 5,200/2.5937 = ₹2,005 cr. Get the order into your fingers: grow NOPAT first, then charge reinvestment, then capitalize.

P2 (guided: quantify the free-growth error). An analyst's terminal year shows NOPAT₁₀ = ₹500 cr and (because the fade suppressed reinvestment) FCFF₁₀ = ₹500 cr. He computes TV = 500 × 1.04/(0.10 − 0.04). The defensible ROIC_T is 12.5%. Compute his TV, the correct TV, and the percentage overstatement.

Solution. His TV = 520/0.06 = ₹8,667 cr. The correct one: RR = 4/12.5 = 32%, FCFF₁₁ = 520 × 0.68 = 353.6, and TV = 353.6/0.06 = ₹5,893 cr. Overstatement = 8,667/5,893 − 1 = +47%. At ROIC_T = 12.5%, growth's capital bill claims nearly a third of every future rupee of NOPAT. Ignore the bill and the perpetuity inflates by almost half.

P3 (guided: feel the hyperbola, respect the cap). FCFF₁₁ = ₹100 cr, WACC 9%, USD-style Rf = 4.2%. Compute TV at g = 3%, at g = 4.2%, and at g = 6%; then state which are admissible and why.

Solution. g 3%: 100/0.06 = ₹1,667 cr. g 4.2%: 100/0.048 = ₹2,083 cr. g 6%: 100/0.03 = ₹3,333 cr. Admissible: 3% (below cap) and 4.2% (at cap, the ceiling case). But 6% is inadmissible, because it asserts the company outgrows its economy forever. Notice how lopsided the temptation is. The illegal move from 4.2% to 6% adds 60% to TV, so the cap has to be a rule. If it were only a preference, it would lose that argument every time.

P4 (guided: the exit-multiple two-step). Kaveri Industrials (synthetic): NOPAT₁₀ = ₹240 cr, EBIT₁₀ = ₹320 cr, WACC 9.5%, terminal g 3.5%, ROIC_T 14%. (a) Compute the Gordon TV and its implied exit EV/EBIT. (b) A banker instead asserts a 15× EV/EBIT exit. What perpetual growth does 15× smuggle in (solve to the nearest 0.1%)? Judge it.

Solution. (a) RR = 3.5/14 = 25%. FCFF₁₁ = 240 × 1.035 × 0.75 = 186.3, so TV = 186.3/0.06 = ₹3,105 cr and the implied exit = 3,105/320 = 9.7× EBIT. (b) Solve 240(1+g)(1 − g/0.14)/(0.095 − g) = 4,800. Try g = 6%: 240 × 1.06 × 0.5714/0.035 = 4,153, too low. Try 7%: 240 × 1.07 × 0.50/0.025 = 5,136, too high. Try 6.7%: 240 × 1.067 × 0.5214/0.028 = 4,768 ≈ target, so g ≈ 6.7%. That's nearly double a typical cap. The "market-standard multiple" was a 6.7%-forever growth assumption in a trench coat. Verdict: reject it as the primary TV. If comps really trade there, that's a fact about today's market mood, not year-10 steady state.

P5 (independent: mid-year and the bridge). A DCF gives end-of-year-convention EV = ₹46,000 cr at WACC 8.6%. Cash and liquid investments ₹3,000 cr, debt ₹9,000 cr, minority interest ₹500 cr, 120 cr diluted shares. Apply the mid-year convention and compute value per share.

Solution. Factor = 1.086^0.5 = 1.0421, so EV = 46,000 × 1.0421 = ₹47,937 cr. Equity = 47,937 + 3,000 − 9,000 − 500 = ₹41,437 cr. Per share = 41,437/120 = ≈ ₹345. The bridge items are today's stocks, so they take no mid-year adjustment.

P6 (independent: build three grid cells). Deccan Cables (synthetic): revenue ₹1,000 cr; EBIT margin 15%; tax 25%; sales-to-capital 2.0; growth 8% (yrs 1–5) fading to terminal g by year 10; ROIC_T 13%; cash ₹50 cr; debt ₹150 cr; 10 cr shares; mid-year convention. Compute per-share value for (WACC, g) = (10%, 3.5%), (11%, 4.0%), (12%, 4.5%), the grid's diagonal.

Solution. Run the machine, and do at least one column fully by hand. The year-1 row is revenue 1,080, EBIT 162, NOPAT 121.5, reinvest 40, FCFF 81.5. The mid-year EVs = ₹1,753 / 1,540 / 1,362 cr, equity = EV − 100, and so per share = ₹165 / ₹144 / ₹126. Look at what that says. Two points of WACC across the diagonal cost ~24% of value. The range is the answer, and a single point picked off this grid is theater.

P7 (independent: scenario weighting done properly). Your three coherent scenarios for a company: bear ₹60/share (p = 30%), base ₹105 (p = 50%), bull ₹170 (p = 20%). Market price ₹128. Compute the probability-weighted value, then state the two-sentence plain read.

Solution. EV = 0.3 × 60 + 0.5 × 105 + 0.2 × 170 = 18 + 52.5 + 34 = ₹104.5. Read: "Weighted value ≈ ₹105 against a ₹128 price means the market is paying above my probability-weighted estimate and close to my bull-adjacent territory, so owning it requires believing my bull case is materially more likely than 20%, or that my scenarios are mis-specified. The spread (60 to 170) matters more than the 104.5, because the price is in the top third of my distribution." If your instinct was "104.5 < 128, therefore sell" and you never looked at the distribution, go back and reread the section on scenarios and Monte Carlo.

P8 (independent: Monte-Carlo interpretation). A Monte-Carlo run (10,000 draws, drivers: growth, margin, terminal g; constraints enforced per draw) on a $354 stock returns: P10 $247, P25 $266, median $289, P75 $313, P90 $334, mean $290, P(value > $354) = 4%. Answer: (a) What does the 4% mean? (b) Why is "the stock is worth $290" the wrong sentence? (c) Name two input sins that would make these bands untrustworthy.

Solution. (a) Under my stated input distributions and model, only 4% of internally consistent scenario draws produce a value above today's price. Buying here means believing something my distributions give ~4% weight. It's a statement about what my model implies, and it isn't an objective probability of mispricing. (b) The mean is one summary of a distribution the exercise exists to display. Reporting it as "the value" brings back the false point-precision Monte Carlo was built to kill, and it hides asymmetry. Here the band is roughly symmetric, but in skewed cases the mean isn't anywhere interesting. (c) First sin: independent draws of correlated drivers. Growth and margin both fall in recessions, so independence understates the true value spread, making the bands too narrow and the 4% too confident. Second sin: distributions with no evidential basis, such as a hopeful mean or an arbitrarily tight sigma, which just formalizes wishful thinking. Also acceptable: unconstrained draws (g exceeding the cap inside draws) or too many free drivers.

P9 (independent: reverse DCF by hand-iteration). Sagar Textiles (synthetic): revenue ₹2,000 cr; EBIT margin 12%; tax 25%; sales-to-capital 2.5; WACC 11.5%; terminal g 4.0%; ROIC_T 12%; standard 10-year machine (g_hi yrs 1–5, fade to 4% by yr 10). The market prices the enterprise at ₹2,300 cr. Given two anchor runs (g_hi = 6% → EV ₹2,106 cr; g_hi = 9% → EV ₹2,335 cr), solve the implied g_hi to the nearest 0.5%, translate it, and judge it against a delivered history of 7% growth.

Solution. Interpolate: (2,300 − 2,106)/(2,335 − 2,106) = 194/229 ≈ 0.85 of the way from 6% to 9%, or ≈ 8.5%. Check it. At g_hi = 8.5%, EV ≈ ₹2,294 cr, a whisker low. At 9% it's 2,335, high. So implied g_hi ≈ 8.5 to 8.6%. In operating terms, revenue must run at ~8.5% for five years (≈ ₹3,010 cr by year 5) before fading, roughly 1.2× the delivered 7%. Judgment: the expectations gap is mildly against the buyer but inside the plausible band. On the evidence this price is roughly fair, with no edge either way. Most reverse DCFs end right here. The tool's daily job is telling you where you have nothing to add. Readings like the first two worked examples are rare, and worth a lot when you find one.

P10 (exam-condition: H-model). Trivandrum Utilities (synthetic) paid D₀ = ₹25/share. Growth is 15% today, fading linearly to 4.5% over 12 years (so H = 6); cost of equity 12%. Compute the H-model value, and state the direction of the approximation's bias.

Solution. V = [25 × 1.045 + 25 × 6 × (0.15 − 0.045)]/(0.12 − 0.045) = [26.125 + 15.75]/0.075 = 41.875/0.075 = ≈ ₹558. Bias: the H-model slightly overstates against the exact linear-fade computation. It credits average-rate growth on a straight line, while actual compounding front-loads less, and the error grows with (g_S − g_L) and H. This is a long, steep fade, so expect the exact value to be a few percent lower. The H-model is fine as a cross-check. Don't use it as the sole method.

P11 (exam-condition: bank preview). (a) In one sentence each, give two reasons FCFF-DCF fails for a bank. (b) A mature bank: ROE 16%, cost of equity 12%, terminal growth 9%… wait. First check whether g = 9% is even available to a bank paying out 60% of earnings. Then value P/B with a legal assumption set of your choosing and defend it.

Solution. (a) Debt and deposits are a bank's raw material instead of its financing, so operations and financing can't be separated and "FCFF to all claimholders" has no meaning. And capex and working capital, the reinvestment concepts FCFF needs, don't describe how a bank grows, which it does by retaining capital against a regulatory ratio. (b) Retention check: g = ROE × (1 − payout) = 16% × 0.40 = 6.4%, so 9% growth with a 60% payout is arithmetically impossible. It's mistake-family #2 in bank clothing, growth someone forgot to fund. A legal set: mature g = 4.5% (the cap), so required retention = 4.5/16 = 28%, a payout of 72%, which is fine. P/B = (ROE − g)/(r − g) = (16 − 4.5)/(12 − 4.5) = 11.5/7.5 = ≈ 1.53× book. The defense: ROE 16% > r 12% persists on a franchise and deposit moat (to be proven in Phase 5's CASA/NIM terms), growth runs at the economy's ceiling, and everything is funded. A 2.5× market P/B would then be a duration bet on that 4-point spread. You can read it, and now it's yours to research.

P12 (TIMED: 25 minutes, full mini-DCF to per share). Arvind Consumer (synthetic): revenue ₹5,000 cr; EBIT margin 18% flat; tax 25%; sales-to-capital 2.0; growth 9% flat for years 1–5, terminal thereafter; WACC 12%; terminal g 4.5%; ROIC_T 15%; mid-year convention; cash ₹400 cr; debt ₹900 cr; 50 cr diluted shares. Deliver: the 5-row engine table, TV, EV, equity, per share, and TV as % of EV.

Solution. Engine (revenue, NOPAT = rev × 0.135, reinvest = ΔRev/2, FCFF, PV at 12%):

• yr1: 5,450 / 735.8 / 225.0 / 510.8 / 456.0
• yr2: 5,940 / 802.0 / 245.2 / 556.7 / 443.8
• yr3: 6,475 / 874.1 / 267.3 / 606.8 / 431.9
• yr4: 7,058 / 952.8 / 291.4 / 661.4 / 420.4
• yr5: 7,693 / 1,038.6 / 317.6 / 721.0 / 409.1

ΣPV explicit = ₹2,161 cr. Terminal: NOPAT₆ = 1,038.6 × 1.045 = 1,085.3. RR = 4.5/15 = 30%. FCFF₆ = 759.7. TV₅ = 759.7/0.075 = ₹10,130 cr. PV(TV) = 10,130/1.12⁵ = ₹5,748 cr. EV end-year = 7,909, and × 1.12^0.5 (=1.0583) gives ₹8,370 cr. Equity = 8,370 + 400 − 900 = ₹7,870 cr, so per share is ≈ ₹157. TV share = 5,748/7,909 = 73%, high-normal for a 5-year horizon. The horizon table in the first section told you to expect ~75 to 80%. If the steady-state audit of year 5 troubles you (9% growth jumping to 4.5% with no fade), extend the horizon. That's the right instinct, and it's why 10-year machines are the default here. (Scoring: ≥90% of the numeric cells within ±2% and a correct TV-share comment = pass-grade work.)

Bonus (the other direction: value above price). Kanpur Forgings (synthetic) trades at EV/forward-NOPAT = 8× with WACC 12%, ROIC_T 14%. Show what growth is priced in.

Solution. No-growth value = NOPAT₁/0.12 = 8.33×, which is above the 8× price. So the market is pricing negative perpetual growth. Solving (1 − g/0.14)/(0.12 − g) = 8 gives g ≈ −4.7%/yr, a business melting ~5% a year, forever. The negative-RR term means capital is released as it shrinks. If your work says plain stability is likely, the expectations gap is in your favor before you count any growth at all. That's what statistically cheap looks like in reverse-DCF language, and it's why the tool cuts both ways.

P13 (guided: Gordon vs. exit, both directions). Godavari Steel Tubes (synthetic): terminal-year NOPAT₁₀ = ₹600 cr, EBIT₁₀ = ₹800 cr, WACC 10.5%, terminal g 4%, ROIC_T 14%. (a) Compute the Gordon TV and its implied exit EV/EBIT. (b) A banker instead pastes a 12× EV/EBIT exit. What perpetual growth does 12× smuggle in (to the nearest 0.1%)? Judge it.

Solution. (a) RR = 4/14 = 28.57%. NOPAT₁₁ = 600 × 1.04 = 624. FCFF₁₁ = 624 × 0.7143 = 445.7. TV = 445.7/(0.105 − 0.04) = ₹6,857 cr, and implied exit = 6,857/800 = 8.57× EBIT, plausible for a mature tube-maker. (b) Solve 600(1+g)(1 − g/0.14)/(0.105 − g) = 9,600. At g = 7% you get ₹9,171 cr, too low. At g = 8% you get ₹11,109 cr, too high. Interpolate, and g ≈ 7.3%, nearly double the 4.5% cap. The "market-standard" 12× is a 7.3%-forever growth assumption wearing a multiple. Reject it as the primary TV. If comps really trade there, it's a fact about today's mood, not year-10 steady state.

P14 (guided: the sixty-second multiple inversion). An Indian FMCG (synthetic) trades at EV/forward-NOPAT = 45×. WACC 10.5%; grant a generous ROIC_T = 22% forever. Solve the implied perpetual growth from (1 − g/ROIC_T)/(WACC − g) = 45, and state whether it clears the 4.5% cap.

Solution. (1 − g/0.22)/(0.105 − g) = 45 → 1 − 4.545g = 45(0.105 − g) = 4.725 − 45g → 40.455g = 3.725 → g ≈ 9.2%. That's roughly double the 4.5% INR cap, forever, even after granting Nestlé-grade capital-lightness. In one breath: 45× can't be a steady state, so the price must embed a long super-normal phase. Now go and measure how long, and test it against base rates. Watch the trap. Dropping the reinvestment term and solving 1/(WACC − g) = 45 gives g ≈ 8.3%. That's still illegal, but it understates the embedded growth by hiding the capital bill.

P15 (independent: legal TV, mid-year, and the bridge, USD). Lakeside Devices (synthetic): terminal-year NOPAT₁₀ = $4.0bn, WACC 8%, terminal g 2.7%, ROIC_T 15%; the explicit-period PVs already sum to $28.0bn. Bridge: cash $3bn, debt (incl. leases) $12bn, minority interest $0.5bn, 0.6bn diluted shares; mid-year convention. Compute FCFF₁₁, TV, PV(TV), TV as % of EV, mid-year EV, equity value, and value per share.

Solution. RR = 2.7/15 = 18%. NOPAT₁₁ = 4.0 × 1.027 = 4.108. FCFF₁₁ = 4.108 × 0.82 = $3.369bn. TV = 3.369/(0.08 − 0.027) = $63.56bn. PV(TV) = 63.56/1.08¹⁰ = $29.44bn. EV(end) = 28.0 + 29.44 = $57.44bn, so TV share = 29.44/57.44 = 51%. That's below the 55 to 75% norm, but it's no red flag. This is a long-duration machine with modest terminal growth. Mid-year: EV = 57.44 × 1.08^0.5 = $59.69bn. Equity = 59.69 + 3 − 12 − 0.5 = $50.19bn, and per share = 50.19/0.6 = ≈ $83.7. The bridge items take no mid-year factor.

P16 (independent: reverse DCF by hand, judge the gap). Vidarbha Agro (synthetic): revenue ₹3,000 cr; EBIT margin 14%; tax 25%; sales-to-capital 2.2; WACC 11%; terminal g 4.5%; ROIC_T 13%; standard 10-year machine (g_hi yrs 1–5, fade to 4.5% by yr 10). The market prices the enterprise at ₹4,200 cr; the company has delivered ~8% revenue growth. Given two anchor runs (g_hi = 7% → EV ₹4,030 cr; g_hi = 10% → EV ₹4,517 cr), solve the implied g_hi to the nearest 0.5%, translate to year-5 revenue, and judge.

Solution. Interpolate: (4,200 − 4,030)/(4,517 − 4,030) = 170/487 ≈ 0.35 of the way from 7% to 10%, so implied g_hi ≈ 8.0%. Translated, revenue runs at ~8% for five years, reaching ≈ ₹4,428 cr by year 5 (1.48× today) before fading. Judgment: 8% implied against 8% delivered, so the price is essentially fair on the evidence, with no edge either way. That's the tool's everyday verdict, as in P9. Readings as dramatic as the first and fifth worked examples are rare. Recognizing "nothing to add here" is the output.

P17 (independent: bank P/B–ROE, USD). Summit Financial (synthetic), a mature US bank: ROE 13%, cost of equity 10%. (a) What is the maximum growth it can fund at a 50% dividend payout? (b) Choosing a legal terminal g = 2.7% (US cap 4.2%), compute the required retention, the implied P/B from (ROE − g)/(r − g), and interpret what a market P/B of 2× would then assert.

Solution. (a) g_max = ROE × (1 − payout) = 13% × 0.50 = 6.5%, so a 50%-payout bank can't grow book faster than 6.5% without raising equity. That's the retention identity at work. (b) At g = 2.7%, retention = g/ROE = 2.7/13 = 20.8%, a payout of 79.2%, which is comfortable. P/B = (13 − 2.7)/(10 − 2.7) = 10.3/7.3 = ≈ 1.41×. Interpretation: ROE (13%) > r (10%), so the bank is worth a premium to book. A market P/B of 2× would be a duration bet. The market would be asserting that the ~3-point ROE-minus-r spread (or a higher ROE/growth) persists far longer than this steady state grants. It's the FMCG-at-45× duration bet wearing a banking license.

P18 (TIMED: 20 minutes, full mini-DCF to per share, INR). Konkan Ceramics (synthetic): revenue ₹4,000 cr; EBIT margin 16% flat; tax 25%; sales-to-capital 1.8; growth 8% flat for years 1–5, terminal thereafter (a 5-year machine); WACC 11.5%; terminal g 4.5%; ROIC_T 13%; mid-year convention; cash ₹200 cr; debt ₹1,100 cr; 25 cr diluted shares. Deliver: the 5-row engine table, TV, EV, equity, per share, and TV as % of EV.

Solution. Engine (revenue, NOPAT = rev × 0.12, reinvest = ΔRev/1.8, FCFF, PV at 11.5%):

• yr1: 4,320.0 / 518.4 / 177.8 / 340.6 / 305.5
• yr2: 4,665.6 / 559.9 / 192.0 / 367.9 / 295.9
• yr3: 5,038.8 / 604.7 / 207.4 / 397.3 / 286.6
• yr4: 5,442.0 / 653.0 / 224.0 / 429.1 / 277.6
• yr5: 5,877.3 / 705.3 / 241.9 / 463.4 / 268.9

ΣPV explicit = ₹1,434.5 cr. Terminal: NOPAT₆ = 705.3 × 1.045 = 737.0. RR = 4.5/13 = 34.6%. FCFF₆ = 481.9. TV₅ = 481.9/(0.115 − 0.045) = ₹6,884 cr. PV(TV) = 6,884/1.115⁵ = ₹3,995 cr. EV(end) = 1,434.5 + 3,995 = ₹5,429 cr, and × 1.115^0.5 (= 1.0559) gives ₹5,733 cr. Equity = 5,733 + 200 − 1,100 = ₹4,833 cr, so per share is ≈ ₹193. TV share = 3,995/5,429 = 74%, high-normal for a 5-year horizon. The year-5 growth (8%) jumps to a 4.5% perpetuity with no fade, and that's the steady-state-audit flag that argues for a 10-year machine. (Scoring: ≥90% of numeric cells within ±2% and the TV-share comment = pass-grade work.)

Applied mini-project

Assemble, bound, and invert a full valuation, twice. Pick one Indian and one US company you've already spread in Phases 1 and 2. Each should be non-financial and profitable, with a history of ≥10 years. If you followed the path's defaults, you have both on file. If not, good pairs are an IT-services major + a US beverage/staples major, or a paints leader + a US big-box retailer. Financials are barred until you reach the special cases, and you now know why.

Data pulls (exact sources): India: last 5 annual reports (company site/BSE), screener.in export for the 10-year history, RBI/CCIL for the G-sec, Damodaran's ctryprem page (January vintage) for spread/CRP. US: 10-K from SEC EDGAR, FRED DGS10, Damodaran's home page for the ERP and industry tables. Shares: diluted count from the latest filing. Price and market cap come from the exchange, on the same date.

Deliverables (one workbook + a two-page memo per company):

1. The forward DCF: a 10-year forecast built with your sales-to-capital method. State your sales-to-capital and show the g = RR × ROIC consistency check for year 1 and year 10. Then a terminal value that obeys all three constraints, with a one-paragraph written defense of ROIC_T that names the moat or its absence. Use the mid-year convention and the full bridge, and finish at a per-share value.
2. The bounds: a WACC × g grid (±1 pt × the legal g band) AND either three coherent scenarios (with written one-sentence worlds and weights) or a 100-draw spreadsheet Monte Carlo (state distributions and their evidential basis). Report: defensible range, base value, PV(TV)/EV, implied exit multiple.
3. The reverse DCF: solve for implied growth at the market price on your held machine, and show ≥3 bracketing iterations. Translate it to year-10 revenue and an operational claim. Judge it against a base rate you compute yourself from screener.in or EDGAR, meaning the 10-year revenue CAGR distribution of ≥30 comparable-scale companies, with the histogram included. Then write the three-lever diagnosis.
4. The verdict page: where the price falls in your range, and what the market believes that you don't, or the other way round. Add the ONE variant question you'd research next, which feeds your Phase-8 file. And include the sentence "this is analysis of a teaching specimen, not investment advice". Write it and mean it.

Scoring rubric (self-grade 0/1/2 per row; ≥80% = pass; log score + top miss in your journal):

Case Lab: run case CL-3.2 (Priced for Perfection: decoding a premium multiple) in the app's Cases tab for this module: a build-bound-and-invert on a fresh synthetic pair (one ₹, one $), auto-graded against the rubric above, with the terminal-value legality, steady-state audit, and three-lever reverse-DCF diagnosis scored line by line.

Reading & resources

Core (do these this week):

• Damodaran, Investment Valuation (3rd ed.): Ch. 12 "Closure in Valuation: Estimating Terminal Value" (the constraints, from the source) and Ch. 13 (DDM). [Paid] [Intermediate]
• Damodaran's free NYU Valuation course (spring session on his site/YouTube): the terminal-value and "what's priced in" lectures; download fcffsimpleginzu.xls and trace every terminal-value cell formula against the reinvestment and fade constraints above. [Free] [Intermediate]
• Mauboussin & Rappaport, Expectations Investing (rev. ed., 2021). Ch. 1–5: the price-implied-expectations framework behind the reverse DCF. [Paid] [Intermediate]
• Mauboussin, The Base Rate Book (Credit Suisse, 2016): free PDF; the sales-growth and margin base-rate tables are your Step-5 reference forever. [Free] [Intermediate]

Reinforcement:

• McKinsey, Valuation (8th ed.): Ch. 14 "Estimating Continuing Value" (their RONIC/key-value-driver treatment mirrors the reinvestment constraint; note they say "RONIC" where we say ROIC_terminal). [Paid] [Advanced]
• Damodaran, The Little Book of Valuation: Ch. 3–4 for a gentle re-derivation if the perpetuity algebra wobbles. [Paid] [Beginner]
• Damodaran, Narrative and Numbers: the story-to-numbers audit behind mistake #15. [Paid] [Intermediate]
• Zerodha Varsity, Fundamental Analysis module: the DCF chapters, for an India-grounded quick second pass. [Free] [Beginner]

Data (live, forever):

• Damodaran Data page (updated each January): implied ERP, country premiums, industry cost-of-capital and sales-to-capital tables, your machine's calibration set. [Free]
• screener.in (India) / stockanalysis.com or TIKR (US): 10-year histories for base-rate building and reverse-DCF inputs. [Free/freemium]
• SEC EDGAR and BSE/NSE filings: the primary documents every number must trace to (working agreement #7). [Free]

Optional depth: Mauboussin, "The Math of Value and Growth" (Morgan Stanley/Counterpoint, free PDF; the growth/ROIC/value triangle in 20 pages); Damodaran's blog posts on his January data updates (he re-derives the g-cap and implied-ERP logic annually, so read one full cycle). [Free] [Advanced]

Going deeper:

• Going deeper: Mauboussin & Callahan, "Everything Is a DCF Model" (Morgan Stanley Investment Management / Counterpoint Global, free PDF) [Free] [Intermediate]. The cleanest argument that every valuation method, multiples included, is an implicit discounted-cash-flow statement. It hardens the exit-multiple reflex ("a multiple is a Gordon formula in disguise") and the reverse-DCF way of seeing prices. A natural read for the week you finish the reverse DCF.
• Going deeper: Stephen Penman, Accounting for Value (Columbia Business School Publishing) [Paid] [Advanced]. It anchors valuation on what the accounts already tell you (book value, current earnings) and treats growth as the speculative residual, something to challenge instead of forecast. It's the accounting-side complement to this week's habit of interrogating the growth a price demands.

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": "Terminal value under all three constraints, the exit-multiple cross-check and its circularity, the mid-year convention, the enterprise-to-equity bridge, scenarios and Monte Carlo, and the reverse DCF",
  "use": "The reverse DCF makes this the module where AI is most useful and most dangerous at the same time, because the tool is excellent at generating stories and the whole method depends on judging stories against base rates. Use it on the story side, never on the base-rate side. (1) Scenario DRAFTING: give it your base case and ask for the bear and bull as *worlds* — which drivers move together, in what order, with what lag, and the one sentence that names each world. §4.11 forbids 'base minus 10%', and a tool that has read a thousand downturns is a good generator of the linkage list you then edit down to what you can defend. (2) TRANSLATION, which is step 4 of the reverse DCF: hand it your implied growth path and ask it to convert it into countable units — year-10 revenue, market share, plant capacity, store count, subscribers — and to find the industry-size figure that makes the claim testable. (3) The pre-mortem: ask what would have to break for the bear case to be too optimistic. (4) Ask it to state your own model's implied exit multiple and implied competitive-advantage period back to you, since both are outputs you did not type and should always look at.",
  "tools": ["Chat assistants — Claude, ChatGPT for scenario drafting as linked worlds, unit translation and the pre-mortem", "Document/PDF AI — for capacity, industry-size and market-share disclosures that make the implied growth testable", "Coding assistants — for the constrained Monte-Carlo engine, the variance decomposition and the reverse-DCF solver below"],
  "prompt": "Here is my base-case driver set for <COMPANY>: <PASTE>. Write a bear world and a bull world as *stories*: one sentence naming each world, then the full driver set for each, with an explicit note on which drivers move together and why. Every column must survive the question 'tell me the world in which all of these hold at once.' Do not adjust my base case, do not attach probabilities, and do not compute any value — I am doing that. Separately: my reverse DCF implies 10-year revenue of <X>; find the published current industry size and tell me what share that implies, citing the source.",
  "verify": "Two verifications, and the second is the one people skip. First, the arithmetic: a drafted scenario is only a driver set until you have run the steady-state audit on it — growth at or below the cap, reinvestment equal to g/ROIC_T, flat margins, marginal tax rate, defensible incremental ROIC. A scenario that fails its own audit is not a scenario, it is a mood. Second, the base rates: step 5 of the reverse DCF is the whole method, and a language model is the wrong instrument for it — it will produce a confident, fluent, unsourced frequency. Pull the base-rate tables yourself (Mauboussin's *Base Rate Book* for the US; build your own NSE-500 ten-year revenue-CAGR distribution for India) and count. And run the reverse DCF BEFORE you form a view, so that you cannot tune toward a target you have already decoded. A model asked whether a growth path is plausible will always say something; only the distribution of what companies actually delivered can answer (Guardrail 1-5).",
  "diy": "The gate is unaided: you build terminal value under all three constraints, derive the ROIC = WACC collapse, invert a multiple to its implied growth on paper, and solve a reverse DCF by bracketing and interpolation with a calculator. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You derived the g-cap from the compounding absurdity and built the terminal value from NOPAT instead of growing FCFF. You proved that g vanishes when ROIC equals WACC, inverted a 14× exit multiple to its implied 10% perpetual growth, and solved a reverse DCF by bracketing. Every one of those has to stay in your hands, the bracketing solve most of all, because it's what makes you independent of a goal-seek. Speed isn't the point of the code here. It turns the constraints from things you remember into things the model can't violate. And it turns the output from a number into a distribution.

Today's workflow. Write the valuation as a function of a driver vector, and put the three constraints inside the function:

• gT = min(gT, Rf), the g-cap enforced in the model's own currency. It uses the same risk-free rate as the WACC stack, so the cap adjusts itself across currencies the way the g-cap derivation requires.
• RR = g / ROIC applied in both the explicit period and the terminal year, so growth is always paid for and free growth can't even be represented.
• ROIC_T supplied as a faded input, never as today's ROIC.

Written that way, a Monte Carlo can't cheat. The fifth Monte Carlo habit, "constraints still bind inside every draw", holds automatically, because no code path exists in which they don't. Three things then become available that a grid can't give you. First, a distribution instead of a point, reported as percentile bands and probability-of-price statements instead of a mean. Second, a variance decomposition. It's the tornado chart's rigorous cousin, and it answers a question the tornado only gestures at: which driver owns the uncertainty in the answer? Third, the reverse DCF as a root-find, which lets you invert for whichever lever you want to interrogate. That could be near-term growth or margin. For a premium franchise, the most useful one is the duration of excess returns. The reverse-DCF section named it as the usual content of the market's bet, and it's painful to solve by hand.

Tools & sources (IN + US). In code, use numpy for the vectorised engine and the draws, scipy.optimize.brentq for every inversion, pandas for the base-rate panel, matplotlib for the value distribution and the fan chart. India: the consolidated annual report gives you the base year and the reinvestment lines. Capacity and volume disclosures plus concall transcripts make the implied growth countable. RBI and MOSPI nominal-GDP series describe the economy the g-cap refers to. CCIL and FBIL G-sec yields give the risk-free rate that is the cap. Screener.in exports of NSE-500 ten-year revenue CAGRs let you build the Indian base-rate distribution yourself. US: EDGAR XBRL companyfacts has the ten-year NOPAT and invested-capital history, and FRED (DGS10, GDP) covers the cap and the economy. Damodaran's sector ROIC, reinvestment and cost-of-capital tables are at pages.stern.nyu.edu. Mauboussin's Base Rate Book (free) has the growth-persistence distributions that step 5 of the reverse DCF is judged against.

# Constrained DCF: the two constraints live INSIDE the function — R3, recomputed in-session
import numpy as np
RF, W, T, N0 = 0.045, 0.11, 10, 20035.0 / 1.06 ** 10        # module's yr-10 NOPAT, back-cast
def dcf(gr, mar, gT, roicT, cap_g=True, tie_rr=True):
    gr, mar, gT, roicT = map(np.atleast_1d, (gr, mar, gT, roicT))
    gT = np.minimum(gT, RF) if cap_g else gT                 # C1: g <= economy (Rf proxy)
    t = np.arange(1, T + 1)
    nop = (N0 * mar)[:, None] * (1 + gr[:, None]) ** t
    rr_ex = np.clip(gr / roicT, 0, 1)[:, None] if tie_rr else 0.0     # C2: explicit period
    rr_T = np.clip(gT / roicT, 0, 1) if tie_rr else 0.0               # C2: terminal year
    return ((nop * (1 - rr_ex) / (1 + W) ** t).sum(1)
            + nop[:, -1] * (1 + gT) * (1 - rr_T) / (W - gT) / (1 + W) ** T)
b = (0.060, 1.00, 0.045, 0.12)
print(f"{dcf(*b)[0]:,.0f}",                                            # 114,692  both on
      f"{dcf(0.060,1.0,0.070,0.12,cap_g=False)[0]:,.0f}",              # 122,438  +6.8%
      f"{dcf(*b, tie_rr=False)[0]:,.0f}",                              # 201,025  +75.3%
      f"{dcf(0.060,1.0,0.070,0.12,False,False)[0]:,.0f}")              # 276,335  +140.9%

Verify. Before the engine goes anywhere near a real company, prove it against the worked figures above. It should reproduce all of these:

• year-11 NOPAT of ₹20,937 crore
• terminal values of ₹1,90,332 / ₹2,01,313 / ₹2,25,471 / ₹2,49,628 crore at terminal ROICs of 11 / 12 / 15 / 20%
• the free-growth ₹3,22,101 crore, which overstates the 12% line by exactly 60.0%
• the ROIC = WACC identity (₹1,90,332 crore = ₹20,937 / 0.11, with g gone)
• the sub-WACC case at 8% giving ₹1,40,919 crore, below the no-growth value, as the key-value-driver formula demands
• the Gordon terminal value's implied 7.54× EBIT, and a 14× exit's implied perpetual growth of 10.03%
• steady-state multiples of 15.03× in rupees and 15.32× in dollars

Keep one convention straight when you check. The steady-state multiple from the exit-multiple section is quoted on this year's NOPAT, so it carries a (1+g) factor. The quick inversion in the reverse-DCF section is quoted on NOPAT₁ and doesn't. They're the same formula with denominators one year apart, so 60× of next-year NOPAT implies 9.1% while 60× of this-year NOPAT implies 9.04%. Both are right. A code implementation must say which denominator it used, because the two differ by exactly one year of growth.

Then check what the model tells you back. The implied exit multiple and the implied competitive-advantage period are outputs you never typed, and looking at them is the cheapest sanity check in valuation. On worked example 5, mid-year throughout, you should see enterprise values of 32.4 / 37.5 / 41.4 / 45.7 billion dollars at trial growth rates of 5, 8, 10 and 12 percent, against 31.1 / 36.0 / 39.7 / 43.9 on year-end discounting. The implied high growth should be 9.1% instead of 9.9%, and the implied discount rate 7.4% on the delivered-growth route. The convention is worth 0.8 points of implied growth on that build, so you state it instead of assuming it.

Quantitative lens

Once the valuation is a constrained function, you can answer three questions that paper can raise but can't settle. Every figure below comes from the engine above, run in-session on the cement-scale figures from the worked build. They're properties of that engine and its stated inputs. None of them measures any company.

The two constraints aren't equal partners, and they interact. Start from the fully constrained base value of ₹1,14,692 crore and break one constraint at a time. Lifting the terminal growth rate from the 4.5% cap to 7%, the "aggressive but arguable" move, raises the value by 6.8%. Removing the reinvestment tie instead, so growth is free, raises it by 75.3%. Removing both raises it by 140.9%. The two effects don't simply add up: 1.068 × 1.753 = 1.872, against an actual 2.409, so the interaction alone contributes another 54 percentage points. Hold on to the mechanism. With the reinvestment tie enforced, a higher terminal growth rate pays for itself. It raises the numerator, and at the same time it raises the reinvestment g/ROIC that funds it, so the net effect is modest. Break the tie and the same higher g arrives free. So the reinvestment constraint is doing roughly eleven times the work of the growth cap. An analyst who polices the g-cap while leaving FCFF-grown terminal values in the model has guarded the wrong door. Earlier you read that terminal reinvestment is the error that inflates more DCFs than any other. This is the ratio behind that claim.

Independent draws lie about the range, and by a measurable amount. You saw the warning earlier: draw correlated drivers independently and the output narrows misleadingly. Here's the size of it. With growth and the margin level drawn independently, the 10th-to-90th-percentile width of value is ₹35,279 crore. Correlate them at 0.6 and the width is ₹38,069 crore. That's the realistic case, because operating leverage brings a growth shortfall and a margin shortfall together. So the independent version understates the range by 7.3%. At a correlation of 0.8 it understates by 9.8%. That isn't catastrophic, but it's systematically in the flattering direction. A narrower distribution makes a "probability that value exceeds price" statement look more confident than it is. The proper fix is to draw jointly. The minimum acceptable one is to say the spread is understated.

Your distributions are probably on the wrong drivers. This is the finding that should change how you work. Decompose the variance of value across the four drivers and the answer is lopsided. Terminal ROIC owns 76.5% of it, the margin level 20.5%, ten-year growth just 2.2% and terminal growth 0.8%. Your instinct is to put a distribution on next year's revenue growth, because that's the number everyone argues about and the one guidance is given on. It's nearly irrelevant to the answer. This company's value depends on what returns the business earns on capital once it is mature, and that's a moat question, graded in Phase 4. The terminal growth rate, the input the whole g-cap discussion is about, contributes under one percent of the variance, and the reason is that the reinvestment tie neutralises it. Two earlier observations turn out to be the same fact seen from two directions. One is that terminal value swings ₹60,000 crore on the choice of ROIC_terminal. The other is the reverse DCF's deepest lesson, that the market's implied bet on a premium franchise is about duration and quality, and not about next year's growth. In practice, spend your uncertainty budget on ROIC_terminal and on the fade, and stop simulating the revenue line.

And the reverse DCF becomes an inversion for any lever. On this engine, a price 1.60 times the base value requires ten-year NOPAT growth of 16.58% against a 6.00% base case. That implies year-10 NOPAT of 2.59× the base case's, a claim you can translate into capacity, share and units and take to the base-rate tables. Solving for growth is the standard move. For a quality name, solving for the competitive-advantage period at a granted terminal ROIC tells you more. Once the machine is a function, both are the same one-line root-find.

Now the limits, at some length, because a Monte Carlo is the easiest place in finance to manufacture false confidence. The distributions above are assumed, and the output's spread is a direct function of assumptions the model can't check. Garbage distributions in, confident-looking garbage out, as the Monte Carlo rules warned. The variance shares depend on the ranges chosen. Widen the growth distribution and growth's share rises, so re-run the decomposition with your own ranges instead of quoting 76.5%. The engine treats ROIC_terminal as a scalar when it's really the endpoint of a fade path. It also holds WACC fixed, when a real downside world would raise the cost of capital at the same moment it lowers the cash flows. The model leaves that correlation out, and including it would widen the left tail. The base-year NOPAT here is back-cast from the cement build's year-10 figure at a constant 6%. That's a construction for teaching, and it isn't any company's history. Most important of all, none of this makes the output a probability distribution over the world. It's a distribution over the model's own inputs. The probability-of-price statement it supports is "under my stated uncertainty, this fraction of coherent worlds justifies the price". That's a sharp and useful sentence, and it's a different sentence from "there is an n percent chance the stock is cheap."

Do it in code: build the constrained engine and assert both constraints inside it. Run the draws jointly, and report percentile bands and a probability-of-price line instead of a mean. Decompose the variance to find out where your uncertainty actually is, and invert for growth and for duration. Then check the implied exit multiple against the ~15× steady-state anchor. See G1/QM1.03, G2/QD2.04 and G1/QM1.01.

Where this goes next: galaxy cross-links

• G1 · QM1.03 Time Series, Monte Carlo & ML. Sampling, joint distributions, copulas and convergence: the machinery behind the value distribution above, including how to draw correlated drivers properly instead of apologising for not having.
• G1 · QM1.01 Probability, Distributions & Estimation: variance decomposition, percentile reporting and the difference between a distribution over a model's inputs and a probability about the world.
• G2 · QD2.04 Reproducible Research & Deployment: a valuation whose every figure can be regenerated from its sources months later, which is the difference between a model that is auditable and one that is merely persuasive.
• G2 · CS2.06 Software Engineering & AI-Assisted Development: the constraints as assertions the engine cannot run without, scenarios as named objects, and a test suite that fails when a driver set violates the steady-state audit.
• G3 · CN1.02 Hypothesis-Driven Problem Solving: the reverse DCF is a hypothesis test on the market's implied claim, and this is the habit of stating the claim, naming the evidence that would settle it, and going to get that evidence.

Teach it back & journal

Feynman prompt. A smart 15-year-old asks: "If nobody can predict anything past three years, isn't a 'terminal value' just a made-up number that's 70% of the answer?" Write one page, with no formulas beyond arithmetic. Cover three things. First, why most of a business's value really does belong to the far future, which is duration, admitted plainly. Second, how the three constraints stop the far future from saying impossible things: nothing outgrows the economy forever, growth must be paid for, and competition eats extraordinary returns. Third, why the right output is a range plus a decoded message from the price. Use one of the real specimens you've worked with here (the ₹2,386 FMCG share or the $354 retailer) as your example. If your page makes the terminal value sound more precise than the near-term forecast, rewrite it.

Journal reflection. Open your decision journal and answer in ink: Before you learned to read a price this way, when you saw a famous company at a breathtaking multiple, what story did you tell yourself: "it's obviously overvalued," or "quality is worth anything"? Now you can decode a price into growth × duration ÷ required return. Which of the three levers are you personally most tempted to over-grant, and in which direction? Most learners find they either grant duration too freely to companies they admire, or refuse it too stubbornly to companies they missed. Write the sentence you'll re-read before every reverse DCF: "When the price disagrees with my machine, the first hypothesis I will test is ___." Date it. Phase 6 will ask whether you kept the promise.

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