DCF I: Free Cash Flows & Forecasting

Phase 3 · Week 29 · ~20 hours

Where it sits. Fourth module of the valuation phase, and the one where valuation finally happens to a company instead of to a formula. The first module built the discounting machine; the second priced risk; the third assembled the discount rates: an 11.0% INR WACC for an Indian cement leader and a 7.8% USD WACC for a US home-improvement retailer, each defensible line by line. Those rates have been waiting for cash flows. Here they get them. What follows defines exactly whose cash flow you are entitled to discount (FCFF vs FCFE), drills the matching rule that decides which discount rate each one is allowed to meet, and then teaches the craft that separates a real analyst from a spreadsheet operator. That craft is forecasting: revenue builds that respect base rates, margin paths that respect competition, and reinvestment that respects the iron tie between growth and returns on capital, g = reinvestment rate × ROIC. By the end you will have built a complete ten-year FCFF forecast for the same two companies whose WACC you constructed last week, numerator finally meeting denominator. The next module finishes the job (terminal value, sensitivity, and reading what the market's price implies). On the five-questions spine (from the master map), this is the engine room of question 5 (what is the price implying, and is that beatable?) and the place where question 2's answer (is ROIC durably above the cost of capital?) gets converted from a judgment into a cash-flow path.

Suggested hour budget. Core content: 8h · worked examples: 2h · practice set: 4h · mini-project: 4h · flashcards + mastery check: 2h.

App link. This module powers the Intrinsic Valuation branch's foundation nodes: the FCFF-builder (assemble free cash flow line by line from a mock filing, graded per line), the matching-rule reflex deck (cash flow ↔ rate ↔ value pairings against the clock), the growth-consistency checker in the valuation sandbox (it flags any forecast where g, reinvestment, and ROIC disagree: the exact audit taught below), and the forecast-build challenges graded against defensible bands rather than a single "right" number.

Learning objectives

By the end you can:

1. Define and compute FCFF three ways from real statements: from EBIT (FCFF = EBIT(1−t) + D&A − capex − ΔWC), from net income, and from CFO. Reconcile all three to the same number, and handle the classification traps (interest in CFO vs CFF under US GAAP / IFRS / Ind AS, stock-based compensation, leases, acquisitions) without double-counting.
2. Define and compute FCFE (FCFE = FCFF − interest(1−t) + net borrowing), explain what net borrowing is doing there, and state when FCFE is the better tool and when it is a trap.
3. Apply the matching rule without exception: FCFF ↔ WACC ↔ enterprise value; FCFE ↔ cost of equity ↔ equity value. Quantify, in percent of value, what each of the two classic mismatches does to a valuation.
4. Build a revenue forecast three ways (market size × share, capacity × utilization × price, units × price), choose the right architecture for the business, and discipline every growth number against base rates and the law of large numbers.
5. Forecast margins from drivers: separate gross-margin economics from operating leverage, anchor the path to sector gravity, and attach a mechanism to every point of assumed improvement.
6. Forecast reinvestment both ways: explicit capex + D&A + working capital, and Damodaran's sales-to-capital shortcut (reinvestment = Δrevenue ÷ sales-to-capital). Reconcile the two.
7. Derive and police the growth-quality tie g = reinvestment rate × ROIC: given any two of growth, reinvestment, and ROIC, solve for the third; audit any forecast (yours or a broker's) for the internal inconsistency that is the single most common amateur DCF error; and run the implied-ROIC check on every model you build.
8. Construct base, bull, and bear cases as coherent stories, scenarios in which every driver moves together for a stated economic reason, and explain why "±10% on revenue" is not a scenario.

Prerequisites & connections

Builds on. M3.03 directly. You built two discount rates there, and they arrive here as finished inputs. The cement leader came out at ≈ 11.0% INR WACC and 11.4% cost of equity, the US home-improvement retailer at ≈ 7.8% USD WACC and 8.4% cost of equity. Both rest on an illustrative mid-2026 stack: US Rf 4.2%, clean INR Rf 4.5%, mature ERP 4.3%, India CRP 2.8% (re-pull it live before real use). The currency rule from that week, INR cash flows with INR rates, now gets applied to the numerator. From M3.01 you need PV mechanics and the Fisher logic linking real and nominal, which is what makes an Indian revenue forecast carry Indian inflation. M2.03 to M2.05 gave you ROIC, NOPAT, invested capital, non-cash working capital and cash-conversion analysis, and here all of it runs forward in time instead of backward. M2.06 to M2.08 taught base-year cleaning (one-offs, aggressive revenue, capitalization games), because you can't forecast from a polluted base. M1.05/M1.08 covered D&A mechanics, lease accounting across the three standards, and the classification differences in the cash-flow statement that decide how you extract FCFF from a filing.

Feeds forward. M3.05 completes this DCF with terminal value, which is where most of the value computed here will turn out to live, plus sensitivity, scenario weighting and the reverse DCF. In M3.06 you'll see that every multiple is a compressed version of the forecasting you do here. P/E and EV/EBITDA are DCFs with the forecasting hidden. In M3.07, g = RR × ROIC becomes the key-value-driver formula and the capital-allocation report card. In M3.08 to M3.09 the model stack automates what you do by hand here, and the working agreement still stands: by hand first. In Phase 8, the Day-2 model of the 2 to 3 day deep dive is this same forecast, run at speed.

The short version. A DCF forces your story about a company through arithmetic that won't let the story cheat. It won't tell you what a company is worth. Free cash flow is what the story pays out, and the growth-reinvestment-ROIC tie is where the cheating gets caught.

Core teaching content

4.1 What a DCF actually is: a structured argument, not an oracle

Strip every valuation method on earth to its chassis and one idea remains: the value of any asset is the present value of the cash flows you expect it to deliver, discounted at a rate that reflects their risk. A bond is the clean case: promised coupons, contractual dates, observable rate. A business is the hard case: the cash flows are not promised, their size depends on competition and management and macro weather, and they stretch beyond any horizon you can see. The DCF's answer:

Value of operating assets = Σ  E[FCF_t] / (1+r)^t     for t = 1 … n
                          + Terminal value_n / (1+r)^n

E[FCF_t] = expected free cash flow in year t (this module)
r        = the matched discount rate (M3.03's WACC or cost of equity)
n        = your explicit forecast horizon (typically 5–10 years)
Terminal value = the value of everything after year n (M3.05's job)

Before a single line of mechanics, get your posture toward this tool right. More DCF damage comes from posture than from arithmetic.

First, a DCF is a structured argument. Every DCF tells a story about a business: how fast it grows, what it earns on sales, what growth costs in capital. That story gets translated into numbers that have to obey accounting identities and economic gravity, so the output can't be any better than the story. Damodaran calls this narrative and numbers. A valuation without a story is numerology, because you can make a spreadsheet say anything. A story without numbers is a fairy tale, because you can make yourself believe anything. The DCF is the bridge that keeps each side in check. Most of the work ahead happens at the crossings where stories cheat: growth without reinvestment, margins without mechanisms, share gains without victims.

Second, if a DCF looks falsely precise, that's on you. The machine will happily print ₹4,132.87 per share. Read properly, any DCF gives you a range. Your revenue path carries error and so does your margin path, and your WACC is ±1 point at best, as you saw when you built it. Expected cash flows are expectations, probability-weighted averages over futures you can't enumerate. So the point estimate is the middle of a distribution and never a fact. In the next module every DCF you build goes into sensitivity tables and scenario ranges. The point estimate is there to anchor the range.

Third, you still can't skip the DCF. The fashionable dismissal goes "DCFs are garbage in, garbage out, just use multiples," and it misses what the tool is for. A multiple is a DCF with the assumptions hidden, which the multiples module proves algebraically. Using one doesn't remove the assumptions. It only removes your awareness of them. What a DCF really produces is a forced confession. To finish one you have to write down, in public, what you believe about growth, profitability, reinvestment, and risk, and every one of those beliefs becomes checkable, debatable, and gradeable. The deep-dive playbook in Phase 8 builds one for every company for that reason. The number isn't reliable, but the argument can be inspected. So when your DCF and the market price disagree, you haven't discovered mispricing. You've found a specific, named disagreement between your story and the market's. Your next job, with the reverse DCF in the next module, is to decide whose story is better.

Every DCF answers four questions:

1. What are the cash flows from existing assets?
2. How fast will they grow, and for how long?
3. What does that growth cost in reinvestment?
4. How risky is the stream?

You answered question 4 last week, with the cost of capital. Questions 1 to 3 are the work ahead. Question 3 is the one amateurs skip, and the growth-quality audit later on exists to catch them.

4.6 Forecasting I: clean the base year before you grow it

Every forecast is base year × story. Corrupt the base and the story compounds the corruption for ten years. So before you project anything, run the base-year audit. It takes fifteen minutes and uses Phase 2 skills, and you don't skip it:

1. Strip one-offs: asset-sale gains, one-time settlements, restructuring charges, COVID-type distortions, "exceptional items." That last one is an Indian annual-report favorite. Check the note, not just the label. You're forecasting the ongoing engine. A base-year EBIT flattered by a ₹900 cr land sale grows that land sale forever.
2. Check the base against history. Is this year's margin near the 5-year average, or is it a cyclical peak or trough? For cyclical businesses a single-year base is a landmine. Think cement, metals, autos, chemicals. Base a DCF on peak margins and you've capitalized the top of the cycle. The full mid-cycle normalization toolkit arrives in the phase capstone, but start the reflex now: "is my year-zero normal?"
3. Re-verify revenue quality. The earnings-quality red flags matter double here: channel stuffing, ballooning contract assets, receivables outgrowing sales. Aggressive revenue recognition steals from the very future years you're about to forecast.
4. Rebuild the reinvestment lines. Pull capex from the cash-flow statement instead of relying on "additions" in the PP&E note alone. Average acquisition spend over 3 to 5 years. Compute non-cash WC from the balance sheet yourself. Then compute the base-year reinvestment rate and ROIC, because you'll need both as anchors when you forecast reinvestment and run the growth audit.
5. Fix the accounting seams. Expense SBC, check lease consistency as described earlier, and undo any capitalized-vs-expensed distortions.

Only now does forecasting begin. A forecast has only three levers: revenue, margins and reinvestment, with the tax rate as supporting cast. Everything else in a DCF is derived. We'll take the levers one at a time, then wire them together with the growth-quality tie.

4.7 Forecasting II: revenue: architectures, base rates, and the arithmetic of humility

Revenue is the lever with the widest error bars and the strongest gravitational pull on value, so it gets the most structure. Two principles first, then three architectures.

Principle 1 is granular near, aggregate far. For years 1 to 3 you can actually see drivers: announced capacity, store-opening pipelines, order books, pricing actions. Model them explicitly. By year 6 that visibility is gone, and pretending otherwise is theater. From there, forecast aggregate growth rates that glide toward the economy's speed. A good ten-year revenue path is a fade, built from drivers at the front and from economics at the back.

Principle 2 is that the currency carries the inflation. An INR forecast rides Indian inflation of ~4 to 4.5% expected (illustrative), which is last week's Fisher machinery at work. A USD forecast rides ~2 to 2.3%. So the same real business performance shows up as "11% growth" in rupees and "8%" in dollars, and the benchmarks shift with it. India's nominal GDP has grown ~10 to 11% a year on the long-run average, which breaks down as ≈6.5% real plus ~4.5% inflation (illustrative). In FY2025-26 it grew 8.9% with 7.7% real growth, a year of unusually low inflation (MOSPI provisional estimates, June 2026, on the new 2022-23 base). Pull current MOSPI/RBI values. US nominal GDP runs ~4 to 4.5%. A 10% INR revenue forecast for an Indian company is par, while a 10% USD forecast for a mature US company is a big claim. Analysts who move between markets without re-anchoring their sense of "fast" misjudge both. It also means you must never compare an INR growth number with a USD growth number raw, just as you never rank rates across currencies.

Architecture 1 is market size × share: Revenue = TAM × market share. It's the right tool when the market is measurable and share is the battleground: paints, adhesives, quick-commerce, cloud infrastructure. It needs three habits from you:

• Size the serviceable market, and leave the press-release TAM out of it.
• Make share gains zero-sum. For every point you award your company, name the loser and explain why they lose it. If your model has HUL, Dabur, and Marico all gaining share, you've built three fictions.
• Beware the "1% of China" fallacy. "The market is $500bn, surely they get 1%" isn't analysis. It's a raffle ticket.

Share forecasts must trace to capability: distribution, cost position, product, moat. Phase 4 gives you the full toolkit.

Architecture 2 is capacity × utilization × price: Revenue = capacity × utilization × realization per unit. Use it when physical assets gate output: cement, steel, power, hotels, hospitals, refineries. Its great virtue is that it keeps growth realistic by construction. The company can't sell tonnes it can't make. So every unit of growth beyond current utilization headroom arrives with a visible capex bill and a lead time, and a cement line takes ~3 years from announcement to commissioning. Check forecast volumes against announced capacity additions in the annual report and investor deck. When the model needs capacity nobody has announced, it's writing cheques the company hasn't. Build 1, the cement forecast later on, runs this architecture end to end.

Architecture 3 is units × price: Revenue = volume × average realization, decomposed however the business actually sells. That might be stores × revenue per store (retail, as in Build 2), customers × ARPU (telecom, SaaS), vehicles × ASP (autos) or patients × ARPOB (hospitals). The decomposition forces the right question. How much of the growth is volume, which is usually durable, and how much is price/mix, which lasts only with pricing power (the moat evidence in Phase 4)? FMCG analysts live on this split. Value growth on flat volumes is a warning, and Phase 5 drills the master brief's thresholds sector by sector.

Then there are base rates, the outside view that keeps your inside view honest. Whatever architecture you use, the finished growth path has to survive contact with the historical record of what companies actually achieve. Mauboussin's Base Rate Book tabulates decades of US outcomes. It's a free PDF on the reading list, so pull it. The shape is brutal:

• Sustained hyper-growth is rare. Across the full universe only a small minority of companies, on the order of 1 in 10, sustain a 10-year sales CAGR above ~20%. The odds collapse as size rises.
• The law of large numbers is undefeated. For companies already above roughly $25 to 50bn of revenue, double-digit 10-year growth is exceptional and 20%+ is essentially unobserved. Big bases make big percentages arithmetically monstrous. A $160bn retailer growing 15%/yr must add a Fortune-100 company every couple of years.
• Growth mean-reverts fast. The correlation between one decade's growth ranking and the next's is weak. Extrapolating a hot five years is how every bubble-era model was built.

In practice, once you've built your path, compute the implied 10-year revenue CAGR and ask two questions. First, what fraction of comparable companies at the same starting size ever achieved this? If the answer is "under 10%," your base case is secretly a bull case. You can keep it only with named, specific evidence that this company belongs in that tail. Second, what does terminal-year revenue imply physically? Think market share of the then-market, stores per city, tonnes against national demand. Models that give a company 70% share of its industry by year 10 are common, and always wrong.

One footnote for India. Apply base rates in real terms or against Indian nominal benchmarks. A 13% INR CAGR is a ~8.5% real claim, ambitious but not out in the tail the way 13% USD would be.

4.8 Forecasting III: margins: drivers, operating leverage, and sector gravity

Revenue says how big the story is, and margin says how profitable it is. Forecast the path of the operating (EBIT) margin, built from three layers:

Layer 1 is gross margin, the economics of the product. It's price minus direct cost, and competitive position sets it: pricing power, input-cost pass-through, mix. Gross margin moves slowly in stable businesses. At HUL or Colgate-India it barely wanders across years. It moves violently where inputs are commodities. Think of cement's energy bill or a chemical firm's feedstock. The forecast question is what's normal, and how far is the current figure from normal? Phase 2's common-size work gives you the history. Mean-revert the deviations unless you can name a reason not to.

Layer 2 is operating leverage, the arithmetic of fixed costs. Below gross profit are costs that don't scale one-for-one with revenue. The consequence is mechanical, and a mini-example burns it in. Take a firm with revenue 100, variable costs 60 (60% of sales) and fixed costs 25. EBIT is 15, a 15% margin. Now let revenue grow 10% with the cost structure unchanged. Revenue is 110, variable costs 66, fixed costs still 25, so EBIT is 19. That's a 26.7% EBIT jump on 10% revenue growth, with the margin going from 15% to 17.3%. Reverse the sign and a 10% revenue fall crushes EBIT 26.7% to 11, a margin of 12.2%.

Operating leverage is a margin elevator on the way up and a trapdoor on the way down. Growing fixed-cost businesses earn expanding margins without heroics for that reason: exchanges, platforms, hospitals filling beds, cement at rising utilization. The same arithmetic is why cyclical downturns vaporize their profits. Your margin path should encode this. If you forecast rising utilization or scale, some margin expansion is earned by arithmetic. Flat volume with assumed margin expansion needs a different mechanism.

Layer 3 is sector gravity, which is competition acting as mean reversion. High margins invite entry, price attacks, and imitation. Sector economics pull on any one company's margin like gravity. Two anchors keep your path inside physics:

• The first is the sector distribution. Damodaran's data page publishes operating and net margins by industry each January, in US, global, emerging-markets/India datasets. The recent shape is illustrative, so verify it live: software/IT services EBIT margins ~20 to 25%, FMCG ~18 to 25% EBITDA, cement mid-cycle EBIT ~12 to 18%, airlines low single digits on a good day, grocery retail 2 to 4%. If your forecast walks a company from sector-median to top-decile margin, the story must contain the mechanism that top-decile peers used. It also needs an argument for why incumbents and entrants permit it.
• The second is the company's own history across a cycle. The best predictor of a firm's normal margin is its own 5 to 10 year record, read across the cycle. Forecasts that exceed the company's best-ever margin and hold it forever are bull cases in base-case clothing.

Improving-margin stories need a convergence rule. Young or under-earning companies legitimately have margin-expansion runways (scale, mix, cost programs). Model them as a glide to a stated target margin by a stated year. Don't let the margin ratchet up forever. For example, "8% today, rising to 15% by year 6, because contribution margin is already 32% and fixed costs grow at half the rate of revenue." Every point of expansion needs a named mechanism: price, mix, scale on fixed costs, input normalization, a specific cost program with a rupee figure. "Efficiencies" isn't a mechanism. It's a hope with a spreadsheet cell.

A word on tax. Apply the marginal statutory rate to EBIT to get NOPAT. That's 25% for India's new regime and ~25% blended for the US (illustrative). Sometimes the effective rate is temporarily lower and you can see why, say SEZ holidays winding down or loss carryforwards with an expiry. Then glide from effective to marginal across the forecast. The terminal year should almost always be at marginal. Never grow a tax holiday forever.

4.11 Scenarios as stories, not knobs

A single path pretends to knowledge you don't have. The professional fix is to write two or three more coherent stories and value each one. Don't blur the single path with a ±10% fudge.

So why isn't "±10% on revenue" a scenario? Because drivers are causally linked. A real downside is a state of the world, and "everything as before, minus 10%" isn't one. In a cement downside, volumes stall and realizations soften from oversupply and operating leverage crushes margins. And, one mercy, growth capex gets deferred, which cushions FCFF. Knob-twiddling misses all four linkages. It produces a bear case with peak margins and a bull case that never paid for its growth. You can spot a knob model because its bull and bear cases have margins, reinvestment, and ROIC audits identical to base. In a story model, each case's audit is internally consistent and distinct. The bull reinvests more, because growth is bought. The bear's operating leverage bites.

Here's how to build them, step by step:

1. Write the base case as a paragraph before you touch any cells. "Industry volume grows with GDP. The company adds announced capacity and holds share and mid-cycle margins. It reinvests at historical S2C." Every number in the sheet must trace to a clause.
2. Make bull and bear different narratives, and not just different arithmetic. For the bull, name what goes right: consolidation firms pricing, a cost program lands, utilization runs hot. Then charge for it. Faster growth means more reinvestment, earlier capacity, maybe rising WC. For the bear, name what goes wrong: a supply glut, an energy spike, share loss. Then let the linkages fire. Margins fall more than revenue through operating leverage. Reinvestment falls too, as projects get shelved.
3. Audit each scenario separately with the policing workflow. A bull case has to survive its own implied-ROIC check. "Everything goes right" still can't mean ROIC of 45% for a cement company.
4. Pre-commit to signposts. For each scenario, write down what you would observe in the next 2 to 4 quarters if that world were unfolding (utilization prints, pricing surveys, competitor capex announcements). That turns scenarios from decoration into a monitoring plan. It also feeds the decision journal, the habit you started at the very beginning of the course and the system Phase 9 builds.
5. Weight only if you must, and always give the range. A probability-weighted value is fine, and 60/25/15 is reasonable shorthand. But the range itself is the primary output. "₹2,900 to 4,600, centered ~₹3,800" tells the truth, while "₹3,847" hides it. The formal sensitivity machinery comes in the next module.

Scenarios pay off most in a few places. In cyclicals the cases are simply cycle states. Binary-event businesses such as pharma approvals or litigation need them too. So does levered equity, where small EV swings become huge equity swings (FCFE users, take note). And so does any thesis whose bull case needs three independent good things to happen at once. Multiply the probabilities and watch your conviction deflate.

4.12 Build 1: ten years of FCFF for the Indian cement leader (INR)

This is the same UltraTech-style specimen whose WACC (≈11.0% INR) you built last week. Every figure is illustrative, modeled on ~FY24/FY25 scale. It's a teaching specimen and not a recommendation. Pull the actual annual report, investor presentation, and screener.in history before any real use, and expect different numbers. The architecture is capacity × utilization × price for years 1 to 5, fading to aggregate growth for 6 to 10. Margins glide with utilization. Reinvestment runs by sales-to-capital, cross-checked against the announced capacity pipeline.

Write the story before the numbers, as the scenario steps say. Housing and the infrastructure push keep India's cement demand growing a little above GDP. The leader grows ~2 points faster than the industry through announced brownfield additions and recent acquisitions, taking modest share. Pricing stays soft at ~2%/yr, below CPI, because the industry is fragmented with regional pockets of oversupply. Margins lift ~1.5 points over five years on scale, cost programs (green power, logistics), and rising utilization, then hold at mid-cycle. Reinvestment continues at the company's historical marginal sales-to-capital of ~1.2. By year 10, growth fades toward India's nominal GDP speed.

The base year has already been cleaned. Revenue is ₹80,000 cr, from volumes of 128 mt × realization of ₹6,250/t, with capacity of 160 mtpa at 80% utilization. EBIT is ₹12,000 cr (15.0%), close to the 5-year mid-cycle average, so there's no peak-basing. Tax is 25%. Invested capital is ₹60,000 cr, so base ROIC = 9,000/60,000 = 15.0%. Base reinvestment is ₹2,400 cr, so RR is 27%.

For years 1 to 5 the drivers are granular. With price up 2%/yr, volume growth = 1.10/1.02 − 1 ≈ 7.8% in years 1 to 3, then 6.9% and 5.9%. That takes volumes from 138 to 182 mt by year 5. At a ceiling utilization of ~84%, year-5 volumes need ~216 mtpa of capacity. That's +56 mtpa over five years, matching the announced pipeline (illustrative). Now a sanity check. 56 mtpa at ~₹6,400 per tonne of annual capacity comes to ≈ ₹35,800 cr of growth capex. That cost is brownfield-weighted. Greenfield runs ~₹8,000 to 9,000/t, so verify current costs. Add WC build of ~5% of cumulative ΔRevenue (≈ ₹2,300 cr) and you get ≈ ₹38,100 cr. Hold that number, because the S2C route has to reproduce it.

Here's the forecast, in ₹ cr with S2C = 1.2 throughout. Every cell can be reproduced from the previous column:

PV of explicit-period FCFF ≈ ₹43,350 cr. The discount factors are 1.11^t. For year 5, that's 7,774 ÷ 1.6851 = 4,614.

Now reconcile the two reinvestment routes, as promised. Cumulative reinvestment by the S2C route over years 1 to 5 is 6,667 + 7,333 + 8,067 + 7,986 + 7,738 = ₹37,791 cr. The capacity-math estimate was ≈₹38,100 cr of growth capex plus WC build, with maintenance capex ≈ D&A. They're within ~1%. The shortcut and the engineering agree, and that's the check that catches a broken S2C assumption. When they disagree by 25%+, believe the engineering and re-fit S2C.

Then run the growth-quality audit on our own model. Never skip it just because the model is yours:

Now read the audit for what it says. Growth is being paid for. An RR of 50 to 66% in the heavy-build years is the real price of 8 to 10% growth at ~15% returns. Two-thirds of NOPAT disappears into kilns and inventory. That's why near-year FCFF looks thin against ₹10,000+ cr of NOPAT. But the implied ROIC has drifted up from the base 15.0% to ~17%, because the margin expansion lifted returns on the whole capital base. Is 17% defensible for a cement company? It's top quartile across a cycle for the sector (Damodaran EM construction-materials tables and the leader's own best years). We accept it here because the story names mechanisms: scale, cost programs, consolidation pricing. But flag it plainly. This base case already leans optimistic. The bear case's first cut is that margin line. The terminal-value work in the next module will force ROIC back toward WACC in perpetuity. The audit didn't decorate the model. It showed us where our own thumb was on the scale.

Also note the year-10 exit. g = 5% with RR = 32% implies forward ROIC ≈ 5/0.32 ≈ 15.6%. That's consistent, and it glides toward India's nominal-GDP-ish speed, which is where the terminal value needs to pick it up.

So what's this worth so far? ₹43,350 cr of PV, against last week's enterprise value of ≈ ₹3,40,000 cr (market cap 3,20,000 + debt 20,000). The explicit decade covers only ~13% of what the market is paying. ~87% of this company's market value is beyond year 10. Most of what the market is paying for lies past the right-hand edge of your spreadsheet. Nothing is wrong with the company or the model here. That's the normal anatomy of a reinvesting business. It's also why terminal value gets its own module and its own checks. A ten-year forecast is the approach road. Treat anyone's "the DCF says" with suspicion until you've seen their terminal assumptions.

4.13 Build 2: ten years of FCFF for the US home-improvement retailer (USD)

This is the Home Depot-scale specimen from last week's WACC build, with WACC ≈ 7.8% USD. It's illustrative, modeled on ~FY24/FY25 scale, so pull the 10-K from SEC EDGAR before any real use. The architecture is stores × revenue-per-store, with explicit-line reinvestment (Route A) this time. It's a deliberate mirror image of Build 1.

The story is a mature, saturated franchise. It has ~2,350 big boxes. The store count is nearly flat (+10 to 15/yr), so growth comes from comps, meaning transactions and ticket on existing boxes. Comps run ~3.5 to 4% nominal, tracking US nominal GDP (housing turnover, pro-customer share, digital attach). There's no margin gift. The 13.4% EBIT margin is held flat, because competition from one formidable direct rival and from e-commerce, plus mix, offsets scale. For a franchise this mature, margin stability is the bull case. On reinvestment, capex is ~2.6% of revenue. That's above the ~2.2% maintenance run-rate, for supply chain, pro fulfillment and refresh. D&A is ~2.2%. Working capital is ~7.5% of incremental revenue, since the business is inventory-heavy and payables-financed. Tax is 25%.

EBIT is lease-adjusted under the consistency rule from earlier. Last week's build counted the ~$8bn of lease liabilities as debt. So the imputed interest inside operating-lease rent (~$0.5bn, disclosed in the 10-K lease note) is treated as financing, and SBC (~$0.4bn) stays expensed.

In the base year, revenue is $160.0bn (2,350 stores × ~$68m). EBIT is $21.5bn (13.4%) and NOPAT $16.1bn. Invested capital is $55bn, so base ROIC ≈ 29%, the signature of a wide-moat franchise. Later in the phase you'll see what three decades of that spread built.

Here's the forecast, in $bn with explicit reinvestment lines:

PV of explicit-period FCFF ≈ $125bn (discount factors 1.078^t), against last week's enterprise value of ≈ $400bn. So the explicit decade covers ~31% of value here, against 13% for the cement builder. It's the same machine with the opposite anatomy. The low-reinvestment cash cow front-loads its value. The heavy reinvestor back-loads it.

Now the audit, and it throws up a puzzle. RR runs only 6.5 to 7.5%, against cement's 50 to 66%. Implied marginal S2C ≈ ΔRev/Reinv ≈ 6, against cement's 1.2. Implied incremental ROIC ≈ 0.75 × 13.4% × 6 ≈ 60%, and book ROIC drifts from 29% to ~36% by year 10. Run step 3 of the policing workflow on that. For almost any company, a model claiming 60% incremental returns would be the #1 amateur sin in a nicer suit. You'd cut growth or raise reinvestment.

Here, three observations save most of it.

1. The growth is mostly nominal comps, price and mix on existing boxes. That's the efficiency-flavored growth from the third consequence above. It drags WC and maintenance capex along but needs almost no new capacity.
2. The franchise's achieved ROIC has run ~30 to 45% for years (brand, scale density, negative-ish incremental capital needs). Verify that in the filings.
3. Store economics say a new box costs ~$40 to 60m and does ~$50 to 70m of mature sales, an S2C of ~1 on new stores. But almost no new stores are being built. The blended S2C of 6 comes from comps arithmetic. Construction isn't driving it.

So it's defensible for the explicit decade, but flagged. The year-10 exit (g 3.5%, RR 6.5%, so implied ROIC ~54%) isn't a forever-assumption anyone should sign. The terminal mechanics in the next module will re-set reinvestment off a sober terminal ROIC, using RR_terminal = g/ROIC. At g 4% and ROIC 30%, RR ≈ 13%, roughly double our year-10 charge.

Notice how differently the same audit resolved in the two builds. It softened a too-rosy margin story in cement and licensed a seemingly outrageous ROIC here. The audit asks questions. It doesn't hand down vetoes.

Both builds teach one lesson. The cement company earns 15 to 17% on capital and must spend two-thirds of NOPAT to grow 8 to 10% in INR. The retailer earns 30%+ and spends 7% of NOPAT to grow 4% in USD. Neither growth rate means anything without its ROIC. Neither FCFF profile means anything without its reinvestment. And neither currency's numbers compare raw with the other's.

Common mistakes & how experts think differently

Here are ten ways real DCFs go wrong at the cash-flow-and-forecast stage. Last week's list covered the rate errors, and the next module covers the terminal ones. Between the three lists you'll hold the full checklist of common DCF mistakes the master brief promised. The app's error-spotting deck draws from all three.

1. Free growth, the cardinal sin. Revenue grows 15% while reinvestment stays at maintenance levels, so implied ROIC climbs past 50% without anyone noticing. It's the most common defect in amateur models, promoter decks, and rushed sell-side work. Expert reflex: run implied ROIC = g / RR on every model within the first two minutes. Include your own, especially your own.

2. Crossing the streams. This means FCFE discounted at WACC, the common error because it flatters. It also means FCFF at the cost of equity, or FCFF at Re presented as equity value with no bridge. Worked example 2 prices these at ±30 to 40% of equity value. Expert reflex: say the mantra before you read any model's output. Whose cash flow, whose rate, whose value?

3. Counting an item twice, or zero times. Cash goes inside working capital and into the bridge. The interest tax shield goes in the cash flows and in WACC's Rd(1−t), whether it gets there through actual taxes or a pre-tax interest add-back. Lease liabilities go in debt while rent stays in EBIT. Expert reflex: draw the boundary once. Every rupee is operating or financing, never both. Audit the seams, not the middles.

4. Forecasting the cash-flow statement instead of building it. That's projecting "CFO grows 10%" or extrapolating last year's FCF directly. CFO and FCFF are outputs of revenue, margins, and reinvestment. Forecast the drivers and let cash flow emerge. Otherwise your model has no place for the story to live. Expert reflex: three levers, everything else derived.

5. Basing the forecast on a dirty base year. Think peak cyclical margins, an un-stripped one-off gain, channel-stuffed revenue, a tax holiday grown forever. Ten years of compounding turns a 5% base error into a 60% value error. Expert reflex: run the base-year audit before a single forecast cell, and ask "is my year zero normal, clean, and real?"

6. Hockey-stick margins with no mechanism. Margin expands 300bp because the spreadsheet drags right. Expert reflex: every point of expansion has a named mechanism, a comparable who achieved it, and a competitor who will contest it. Expansion beyond the company's own best-ever or the sector's top decile is a bull case, so label it as one.

7. Heroic growth against the base rates. A $160bn company growing 15% USD for a decade. An Indian mid-cap holding 25% INR growth to year 10. Three companies in the same model all gaining share. Expert reflex: compute the implied 10-year CAGR and terminal size, then look up the base rate in Mauboussin. Name the share losers. If you can't, cut the number.

8. Sales-to-capital fitted to one lucky year. S2C gets estimated from a single year in which capex paused, which makes growth look free. Or it comes from a year when a mega-project landed, which makes growth look ruinous. Expert reflex: fit S2C on 3 to 5 year sums. Cross-check it against the sector table and the engineering, meaning unit capacity costs. Then reconcile Route A with Route B as Build 1 did.

9. SBC added back and dilution ignored, so value is counted twice for the employees. This is "free cash flow" that adds back stock comp and divides it over today's share count. The company pays wages in equity, and the model pretends nobody pays them. Expert reflex: expense SBC in FCFF, which is the course standard. Use diluted shares in the bridge. Any add-back must come with explicit dilution modeling.

10. Scenario theater. Bull and bear run as ±10% knobs. A "conservative case" trims growth but keeps bull margins and bull reinvestment. Scenarios get built after the answer, to bracket it comfortably. Expert reflex: scenarios are stories with their own audits and their own signposts. If the bear case doesn't change what you'd watch next quarter, it's a decoration.

Underneath all ten is one difference between novices and experts. A novice treats the forecast as the future and defends every cell. An expert treats it as the argument and defends the structure: clean base, three levers, growth priced by g = RR × ROIC, scenarios that cohere, and an audit trail a colleague could attack line by line. Novices hide assumptions inside formulas. Experts bring them out, because an assumption you can see is one you can debate, monitor, and change your mind about. The best practitioners hold the tool with Damodaran's lightness. Your DCF will be wrong. The job is to be wrong for reasons anyone can inspect, and less wrong than the price.

Worked examples

Worked example 1: Three routes to FCFF, one answer (and the Ind AS trap), then FCFE

Kaveri Consumer Products Ltd is a synthetic Indian FMCG company, with numbers invented for the drill. Take the FY statements. EBIT is ₹1,200 cr. Interest expense is ₹80 cr, and the marginal tax rate is 25%. D&A is ₹150 cr, capex ₹260 cr. Non-cash working capital goes from ₹300 cr opening to ₹340 cr closing. Net new borrowing is ₹50 cr. The cash-flow statement (Ind AS) classifies interest paid under CFF and reports CFO of ₹1,030 cr.

Route 1, from EBIT:

NOPAT = 1,200 × 0.75              = 900
+ D&A                             = +150
− Capex                           = −260
− ΔWC   = −(340 − 300)            =  −40
FCFF                              = ₹750 cr

Route 2, from net income. NI = (1,200 − 80) × 0.75 = ₹840 cr.

FCFF = 840 + 80×(1−0.25) + 150 − 260 − 40 = 840 + 60 + 150 − 260 − 40 = ₹750 cr  ✓

Route 3, from CFO, minding the classification. First verify the reported CFO. With interest paid parked in CFF, CFO = NI + D&A − ΔWC + interest added back = 840 + 150 − 40 + 80 = ₹1,030 cr, which matches the filing. Interest isn't inside CFO, so you use the shield-stripping form:

FCFF = CFO − interest × t − capex = 1,030 − 80×0.25 − 260 = 1,030 − 20 − 260 = ₹750 cr  ✓

Had this been a US GAAP filer, with interest inside CFO, reported CFO would have been 1,030 − 80 = ₹950 cr. Then the add-back form applies: 950 + 80×0.75 − 260 = 950 + 60 − 260 = ₹750 cr. You land in the same place. Same company, same FCFF, two different CFO numbers on the page. The analyst who doesn't check the classification gets ₹830 cr or ₹670 cr and never knows why.

FCFE, both routes:

From FCFF:        FCFE = 750 − 80×0.75 + 50 = 750 − 60 + 50 = ₹740 cr
From net income:  FCFE = 840 + 150 − 260 − 40 + 50 = ₹740 cr  ✓

What does that mean? Of ₹750 cr generated for all financiers, lenders absorbed a net ₹10 cr. That's ₹60 cr of after-tax interest less ₹50 cr of new money, which leaves ₹740 cr that could reach shareholders. Whether it does, through dividends, buybacks, or a growing cash pile, is a capital-allocation question for later in the phase. It isn't a valuation input.

Worked example 2: The matching rule priced: what each mismatch costs

A synthetic firm, engineered to be fully consistent so every route has one true answer. Deccan Foods has FCFF next year of ₹800 cr, growing 4% forever. That's a Gordon setup like the ones you used to build the discounting machine. It's legitimate here because we need closed forms. Capital structure is held at 1/3 debt and 2/3 equity at market values. Cost of equity is 12%. Pre-tax cost of debt is 8%, and with tax at 25% the after-tax cost is 6%. There's no surplus cash. WACC = (2/3)×12% + (1/3)×6% = 10.0%.

Correct valuation, firm route:

EV     = 800 / (0.10 − 0.04) = ₹13,333 cr
Debt   = (1/3) × 13,333      = ₹4,444 cr
Equity = 13,333 − 4,444      = ₹8,889 cr

Correct valuation, equity route. Check the machine agrees. Interest = 4,444 × 8% = ₹356 cr, or ₹267 cr after tax. Debt grows with the firm at 4%, so net borrowing = 4% × 4,444 = ₹178 cr.

FCFE   = 800 − 266.7 + 177.8 = ₹711.1 cr
Equity = 711.1 / (0.12 − 0.04) = ₹8,889 cr  ✓  (exactly — consistency is possible, just rare)

Mismatch 1, FCFE at WACC (the flattering error): 711.1 / (0.10 − 0.04) = ₹11,852 cr. That overvalues equity by 2,963 cr, or +33%. A levered, residual claim was priced at the blended rate of the whole structure. The more debt, the bigger the free lunch the model hallucinates.

Mismatch 2, FCFF at Re, debt then subtracted: 800 / (0.12 − 0.04) = 10,000. Minus debt of 4,444, that's ₹5,556 cr, or −37%. All claimholders' cash flow was punished at the equity-only rate. Then the lenders were paid a second time in the bridge.

Mismatch 2b, the compounding amateur version: FCFF at Re with the bridge forgotten gives "equity = ₹10,000 cr," or +12.5%. That's wrong by two errors that happen to partially offset at this leverage and wouldn't at another. Errors that sometimes cancel are more dangerous than errors that don't, because they pass casual inspection.

So one synthetic firm, one true equity value, and three published-quality wrong answers spanning ₹5,556 cr to ₹11,852 cr. That 2.1× spread comes entirely from pairing errors, with not one forecast assumption changed. It's the reason the matching rule gets drilled as law.

Worked example 3: Policing a broker's forecast with g = RR × ROIC

The setup is synthetic, but the pattern is from life. A sell-side note on an Indian specialty-chemicals company (call it Sahyadri Chem) shows revenue and NOPAT growing 14%/yr for the next five years. NOPAT rises from ₹800 to ₹1,540 cr. Capex runs ≈ ₹280 cr/yr against D&A ≈ ₹200 cr/yr. ΔWC ≈ ₹60 cr/yr. Over ten years the company's achieved ROIC has been 15 to 18%, against a sector median of ~14%.

Step 1, reinvestment and RR. Reinvestment ≈ 280 − 200 + 60 = ₹140 cr against year-1 NOPAT ≈ ₹912 cr, so RR ≈ 15%. It falls from there, as NOPAT grows against flat reinvestment.

Step 2, implied ROIC. g = RR × ROIC ⇒ ROIC = 14% / 15.4% ≈ 91%. The note claims each rupee reinvested creates ~91 paise of annual NOPAT, forever after. Cross-check it at the margin. Year-2 ΔNOPAT ≈ 0.14 × 912 ≈ ₹128 cr on ₹140 cr reinvested, which is ≈ 91%. Confirmed.

Step 3, confront with evidence. Achieved ROIC is 15 to 18%, and the sector's is ~14%. The note names no mechanism. There's no idle capacity to fill, since the note itself says utilization is 88%. Nor is there a margin-expansion story beyond 40bp. So the forecast is internally false. At the company's own excellent 18% ROIC, 14% growth costs RR = 14/18 ≈ 78% of NOPAT.

Step 4, price the correction. Properly charged, year-1 reinvestment ≈ 0.78 × 912 ≈ ₹711 cr, so FCFF ≈ ₹201 cr against the note's ≈ ₹772 cr. The model overstates near-term free cash flow by roughly 3.8×, and every later year compounds it. Or keep the ₹140 cr reinvestment. Then g can only be ≈ 15% × 18% ≈ 2.7% instead of 14%. Either the growth is a fantasy or the free cash flow is. The note is charging the client for both. The fix has to pick ONE lever the story supports. Here capacity math says growth needs the capex, so keep g, slash FCFF, and watch the target price deflate.

That's ninety seconds of arithmetic and no proprietary data. It's astonishing how often this audit fails models that moved real money. And it's the skill the gate certifies.

Worked example 4: A five-year FCFF forecast built from scratch (US $, sales-to-capital)

Redwood Analytics Inc. (synthetic) is a US enterprise-software firm, and every figure is invented for the drill. It's the whole machine in miniature, on a horizon short enough to hold in your head. You go from revenue growth to the margin path. From there you move on to reinvestment and FCFF. Discounting and the audit finish it off. The base year is already cleaned. Revenue is $2,000m. An EBIT margin of 15% gives EBIT of $300m. Marginal tax of 25% leaves NOPAT of $225m. Invested capital is $750m, so base ROIC = 225 / 750 = 30%. The capital is people and code instead of factories, so it's asset-light. SBC is expensed inside EBIT, not added back. The matched USD WACC, from a build like last week's, is 9.0% (illustrative, re-derive live).

Write the story before the numbers. Redwood is a maturing category leader. Revenue growth fades from 20% to 10% as the base enlarges. Check that against the base rates. A $2bn software firm compounding ~15% for five years is plausible and not out in the tail. The implied 5-yr revenue CAGR is ~15.3%. The EBIT margin glides from 16% to 24% over five years. Sales-and-marketing and R&D grow slower than revenue. That operating leverage, plus a mix shift toward higher-margin renewals, drives the glide. Each point of expansion carries that named mechanism. Reinvestment is priced by sales-to-capital, S2C = 3.0, which fits Damodaran's asset-light software band.

Here's the forecast in $m, with Reinvestment = ΔRevenue ÷ 3.0. Every cell can be reproduced from the previous column:

PV of the explicit five-year FCFF ≈ $1,337m. The discount factors are 1.09^t, so year 5 is 611.3 ÷ 1.5386 = 397.3. There's no terminal value yet. This is the approach road, and the next module builds the destination.

Now run the growth-quality audit on our own model. RR = Reinvestment ÷ NOPAT falls from 46% to 17% as margins fatten and growth fades. Implied incremental ROIC = after-tax margin × S2C = 0.75 × margin × 3.0. It rises from 36% to 54% with the margin glide. Book ROIC (NOPAT_t ÷ IC_{t−1}) drifts from ~38% to ~56%. Confront it with step 3. Is 40 to 55% ROIC defensible? For asset-light software it's characteristic, since the leaders print 40 to 60%, so we acquit the explicit window. But flag what's driving it. The ROIC climbs because we assumed the margin glide. Every one of those eight points has to survive the mechanism test from the margins work. If they don't, the whole forecast is leaning on unaudited operating leverage. And the year-5 exit (g 10%, RR 17%, so forward ROIC ≈ 10 / 0.17 ≈ 59%) is emphatically not a forever-assumption. The terminal work in the next module re-bases reinvestment off a sober terminal ROIC. It's the same posture as with Build 2's retailer. The audit asks questions. It doesn't veto.

Worked example 5: Getting paid to grow: a negative-working-capital forecast (India ₹, explicit lines)

Ganga Naturals Ltd (synthetic) is an Indian branded-foods (FMCG) company, with figures invented. This time reinvestment is forecast the other way, by Route A with explicit lines. The point of the example is the line amateurs get the sign wrong on, working capital. In the cleaned base year, revenue is ₹6,000 cr. An EBIT margin of 18% gives EBIT of ₹1,080 cr. Tax of 25% leaves NOPAT of ₹810 cr. Invested capital is ₹2,700 cr, so base ROIC is 30%. The matched INR WACC is 11.0%.

The story is steady mid-teens nominal growth, fading from 14% to 10% as Indian FMCG tracks nominal GDP. The 18% EBIT margin holds flat. These are mature brands with a slow-moving gross margin, the stable case from the margins work. Capex is 7% of revenue and D&A 4% of revenue. The feature is non-cash working capital of −5% of incremental revenue. The company collects from distributors fast and stretches payables. So every rupee of new sales releases cash instead of absorbing it. It's the "delicious exception" from the FCFF build. Because ΔWC is negative, Reinvestment = capex − D&A + ΔWC is charged the growth capex but credited the working-capital release.

Here's the forecast in ₹ cr. A negative ΔWC is a cash inflow:

PV of the explicit five-year FCFF ≈ ₹3,461 cr. Now run the same forecast with a normal +5% working-capital drag instead of the −5% release. The PV falls to ≈ ₹3,126 cr. So the negative-WC engine is worth ≈ ₹335 cr, a ~11% uplift. Over five years it hands back ≈ ₹228 cr of cash, purely by growing. That's one quiet reason the market pays premium multiples for business models like this.

The audit comes with a caveat for negative working capital. Book ROIC (NOPAT_t ÷ IC_{t−1}) runs at a sector-characteristic 34% to 41%. Asset-light FMCG really does earn this (check against history and Damodaran's India tables). RR runs ~18%. But watch the marginal reading. g/RR in year 1 is 14 / 18 ≈ 79%, and read naively as "implied ROIC" it looks like the #1 amateur sin. It isn't. It's an artefact. The working-capital release depresses measured reinvestment, which inflates the g/RR ratio. So every negative-WC forecast has to carry the lesson: the marginal g/RR audit is distorted for these firms, so lean on book ROIC and the sector comparison instead. Remember too that the WC tailwind is finite. It scales with growth but can't fund perpetual terminal growth. So the terminal work re-bases terminal reinvestment off a sober terminal ROIC, as it would for any other business.

Practice set

There are ten problems, ramped from guided to independent to timed. All companies are synthetic unless noted. Write your full answer before you open any solution, as working agreement #3 asks. Currency units are as stated. Tax is 25% throughout unless you're told otherwise.

P1 (guided: FCFF from EBIT). Tristar Tools (US industrial) has EBIT of $400m, D&A of $90m and capex of $150m. Non-cash working capital is $210m opening and $240m closing. Compute FCFF, and state the reinvestment and the reinvestment rate.

Solution. NOPAT = 400 × 0.75 = $300m. ΔWC = 240 − 210 = +$30m, which consumes cash. FCFF = 300 + 90 − 150 − 30 = $210m. Reinvestment = capex − D&A + ΔWC = 150 − 90 + 30 = $90m, so RR = 90/300 = 30%. The compact form checks it: FCFF = NOPAT − reinvestment = 300 − 90 = 210.

P2 (guided: the NI route and FCFE). Same firm, with interest expense of $40m and net new borrowing of +$20m. Recompute FCFF from net income, then compute FCFE two ways.

Solution. NI = (400 − 40) × 0.75 = $270m. FCFF = NI + interest(1−t) + D&A − capex − ΔWC = 270 + 30 + 90 − 150 − 30 = $210m, which matches P1, as the routes must. FCFE from FCFF is 210 − 40×0.75 + 20 = 210 − 30 + 20 = $200m. FCFE from NI is 270 + 90 − 150 − 30 + 20 = $200m, the same. The lenders' net take is $30m of after-tax interest less $20m of new money, so $10m.

P3 (guided: the CFO route and the classification trap). Narmada Polymers is Indian and reports under Ind AS. EBIT is ₹600 cr. Interest is ₹100 cr, classified in CFF according to the policy note. D&A is ₹80 cr, ΔWC +₹40 cr, capex ₹180 cr. Reported CFO is ₹515 cr. (a) Verify the reported CFO. (b) Compute FCFF from CFO. (c) What CFO would an otherwise identical US GAAP filer report, and what is the FCFF formula from that CFO?

Solution. (a) NI = (600−100)×0.75 = ₹375 cr. With interest paid outside CFO, CFO = 375 + 80 − 40 + 100 = ₹515 cr, as reported. (b) Interest is in CFF, so strip only the tax shield: FCFF = CFO − interest×t − capex = 515 − 25 − 180 = ₹310 cr. The EBIT route checks it, 450 + 80 − 180 − 40 = 310. (c) US GAAP puts interest paid in CFO, so CFO = 515 − 100 = ₹415 cr. Then FCFF = CFO + interest(1−t) − capex = 415 + 75 − 180 = ₹310 cr, again. Same company, two CFO numbers ₹100 cr apart, one FCFF. Check the classification before you touch Route 3.

P4 (independent: matching-rule rapid fire). For each, write OK or VIOLATION, and for violations the direction of the equity-value error: (a) PV(FCFE @ Re) = equity value. (b) PV(FCFF @ WACC) + cash − debt − minority interest = equity value. (c) PV(FCFE @ WACC) = equity value. (d) PV(FCFF @ Re) − debt = equity value. (e) PV(FCFF @ WACC) = equity value, no bridge, for a firm with net debt. (f) PV(FCFE @ Re) − net debt = equity value.

Solution. (a) OK. Equity cash flow, equity rate, equity value. (b) OK. It's the full bridge, each item once. (c) VIOLATION, overvalues. A levered residual claim is priced at the blended, lower rate. (d) VIOLATION, undervalues. Firm cash flow is punished at the equity rate, and then lenders are paid again in the bridge. (e) VIOLATION, overvalues. The lenders' claim was never removed. (f) VIOLATION, undervalues. FCFE is already struck below debt service, so subtracting net debt charges shareholders for it twice.

P5 (independent: the growth-quality tie, three ways). (a) NOPAT is ₹2,400 cr, reinvestment ₹1,080 cr, incremental ROIC 16%. What NOPAT growth does this support? (b) A company wants g = 12% with ROIC of 15%. What reinvestment rate does that require, and what fraction of NOPAT is left as FCFF? (c) A firm reinvests 30% of NOPAT at 25% ROIC and management guides to margin expansion. What growth does the formula license, and how do you treat the margin claim?

Solution. (a) RR = 1,080/2,400 = 45%, so g = 0.45 × 16% = 7.2%. (b) RR = g/ROIC = 12/15 = 80%, which leaves FCFF at 20% of NOPAT. That's the price tag of fast growth at moderate returns. (c) Reinvestment-funded g = 0.30 × 25% = 7.5%. The margin expansion is efficiency growth on top. It's legitimate for a few years with a named mechanism, but it lifts implied ROIC on the whole base, so re-run the audit each year it persists. It also can't fund terminal growth, as the terminal-value work will show.

P6 (independent: sales-to-capital mini-build). Arjuna Foods (Indian FMCG) has base revenue of ₹5,000 cr, growing 12%, 12%, 10%, 8%, 6%. The EBIT margin is 18% flat, tax is 25%, and S2C = 2.5. Build five years of FCFF. Then compute the implied incremental ROIC and judge it against an FMCG lens.

Solution. (₹ cr, rounded)

Implied incremental ROIC ≈ after-tax margin × S2C = 0.75 × 18% × 2.5 = 33.75%. That's high but characteristic of the sector. Asset-light FMCG with distribution-led growth and modest WC, often negative, really does print 30%+ ROIC (check the company's history and Damodaran's India tables). The same number in cement would be an indictment. In FMCG it's a description. Sector context is part of the audit.

P7 (independent: indict or acquit). A model for Vindhya Wires shows NOPAT growing 18%/yr with reinvestment steady at 25% of NOPAT. The company's achieved ROIC over ten years is 20%, and the sector's best is 24%. (a) What ROIC is the model implying? (b) Indict or acquit, with the fix and its FCFF consequence.

Solution. (a) ROIC = g/RR = 18%/25% = 72%, three times the company's best-ever and the sector ceiling. (b) Indict. At the achieved 20% ROIC, 18% growth needs RR = 18/20 = 90% of NOPAT. That leaves FCFF of 10% of NOPAT against the model's 75%, so near-year free cash flow is overstated ~7.5×. Or keep RR = 25%, and the growth it can actually fund is 0.25 × 20% = 5%. Pick the lever the story supports. If capacity and announced capex justify 18% growth, FCFF collapses. If they don't, growth collapses. Either way the valuation falls sharply.

P8 (independent: clean the base first). Chambal Agro reports EBIT ₹1,500 cr, which includes a one-time gain of ₹300 cr on a land sale and a ₹100 cr impairment charge. The ongoing figures are D&A of ₹200 cr, capex of ₹350 cr, ΔWC of ₹50 cr. (a) Compute normalized base-year FCFF. (b) If an analyst forecasts 10% growth for a decade off the reported EBIT, roughly how large is the year-10 EBIT error?

Solution. (a) Clean EBIT = 1,500 − 300 + 100 = ₹1,300 cr. You strip the non-recurring gain and add back the non-recurring charge. NOPAT = 975, so FCFF = 975 + 200 − 350 − 50 = ₹775 cr. (b) Both bases grow 10%, so the ₹200 cr base error compounds untouched. Year-10 EBIT is overstated by 200 × 1.10^10 ≈ ₹519 cr. That's about 15% too high in every single year, and the terminal value inherits all of it. Dirty base, poisoned decade.

P9 (timed: 30 minutes, full mini-forecast). Meridian Infotech (Indian IT services, INR) has base revenue of ₹10,000 cr, growing 8%, 8%, 7%, 6%, 5%. The EBIT margin is 22% flat, tax is 25%, S2C = 3.0 and WACC is 12% (INR). Build the five-year FCFF table, discount it, and run the audit: implied incremental ROIC, the RR path, a one-line verdict.

Solution. (₹ cr)

PV of explicit FCFF = ≈ ₹6,334 cr (accept 6,320 to 6,345 for rounding paths). For the audit, implied incremental ROIC = 0.75 × 22% × 3.0 = 49.5%. A same-year check on year 2 gives ΔNOPAT 142.5 ÷ reinvestment 288 = 49.5%, because under the S2C convention year-t reinvestment delivers year-t revenue. RR ≈ 15%, falling to ~10%. Ultra-high ROIC and low RR are the signature of top-tier IT services. They're people-led and capital-light, and the leaders' filings show 40 to 60% ROIC. Acquit. But notice that these economics are why such firms return most of NOPAT as dividends/buybacks instead of reinvesting it. They can't deploy more capital at these returns, which is a question for the capital-allocation work later in the phase.

P10 (timed: 20 minutes, scenario coherence). Take the cement leader from the ten-year build. A colleague's "bull case" for it reads like this. Volumes grow +12%/yr and realization +5%/yr, so nominal growth is ≈17%. The EBIT margin reaches 19% by year 3. The capex plan is unchanged from base, and S2C is raised to 2.0 "for efficiency." (a) Find every incoherence. (b) Rebuild a coherent bull case (state your choices) and compare year-3 FCFF under both.

Solution. (a) There are three incoherences. (1) Volumes without capacity. At +12%/yr, volumes hit the 84% utilization ceiling inside two years. With capex unchanged, the model is selling cement from kilns nobody built. (2) S2C = 2.0 is asserted, not earned. Cement's engineering (₹6,000 to 9,000/tonne capacity) caps marginal S2C near 1.0 to 1.3. Doubling it is the free-growth sin in a bull costume, with implied incremental ROIC = 0.75 × 19% × 2.0 = 28.5%, far beyond any cement company's history. (3) Margins and pricing double-count the same good news. A 19% margin already requires the +5% realization, so check it isn't being layered in twice through a separate cost-decline line. (b) A coherent bull has growth of ~11% nominal, from volumes up to the utilization ceiling plus accelerated-but-announced capacity, with price +3%. The margin is peak-flavored at 18%. S2C is 1.3 (best-case brownfield mix). That gives implied ROIC = 0.75 × 18% × 1.3 ≈ 17.6%, at the company's historical best, which is what a bull case should claim. Now compare year 3. For the incoherent bull, revenue = 80,000 × 1.17³ ≈ 128,129. NOPAT = 128,129 × 0.19 × 0.75 ≈ 18,258. Reinvestment = ΔRev/2.0 = (128,129 − 109,512)/2 ≈ 9,308. So FCFF ≈ ₹8,950 cr. For the coherent bull, revenue = 80,000 × 1.11³ ≈ 109,410. NOPAT = 109,410 × 0.18 × 0.75 ≈ 14,770. Reinvestment = (109,410 − 98,568)/1.3 ≈ 8,340. So FCFF ≈ ₹6,430 cr. The incoherent case smuggles in ~39% more year-3 free cash flow and an impossible ROIC. Bull cases have to buy their growth like everyone else.

P11 (guided: a three-year FCFF forecast end to end). Solstice Devices Inc. (synthetic, US) has base revenue of $1,200m, growing 12%, 10%, 8%. The EBIT margin is 20% (flat), tax is 25%, sales-to-capital is 2.5 and WACC is 9%. Build three years of FCFF, discount them, and state the implied incremental ROIC.

Solution. ($m)

PV of explicit FCFF = ≈ $421.9m. Implied incremental ROIC = 0.75 × 20% × 2.5 = 37.5%, constant because margin and S2C are flat. Here are the mechanics in brief. NOPAT = revenue × 20% × 0.75. Reinvestment = ΔRevenue ÷ 2.5. FCFF = NOPAT − reinvestment, and you discount at 1.09^t.

P12 (guided: the working-capital sign trap). Konkan Foods (synthetic, Indian FMCG) has EBIT of ₹900 cr, with tax at 25%. D&A is ₹120 cr and capex ₹180 cr. Non-cash working capital moves from −₹400 cr to −₹470 cr, so it becomes more negative as sales grow. (a) Compute reinvestment and FCFF. (b) What would FCFF have been if working capital had instead risen by ₹70 cr (a normal drag)?

Solution. (a) NOPAT = 900 × 0.75 = ₹675 cr. ΔWC = (−470) − (−400) = −₹70 cr. Non-cash WC fell (it grew more negative), which is a cash inflow. Reinvestment = capex − D&A + ΔWC = 180 − 120 + (−70) = −₹10 cr. Net, the operating base released cash while it grew. FCFF = NOPAT − reinvestment = 675 − (−10) = ₹685 cr. (b) With ΔWC = +₹70 cr, reinvestment = 180 − 120 + 70 = ₹130 cr. So FCFF = 675 − 130 = ₹545 cr. One line's sign swings FCFF by ₹140 cr. Get the sign of ΔWC wrong and the whole forecast tilts the wrong way.

P13 (independent: a margin-glide story, audited). Aurora Cloud Inc. (synthetic, US SaaS) has base revenue of $800m, growing 15%/yr for four years. The EBIT margin glides 12%, 15%, 18%, 20%. Tax is 25% and sales-to-capital is 3.0. (a) Build four years of FCFF. (b) Compute the implied incremental ROIC each year and state what the model is assuming without saying so.

Solution. (a) ($m)

(b) Implied incremental ROIC = 0.75 × margin × 3.0, rising from 27% to 45% as the margin glides. Without saying so, the model is asserting that returns on new capital nearly double. That's defensible only if each point of margin expansion has a named mechanism and the sector supports 40%+ returns. If neither holds, the rising ROIC line is the tell that the margin story is doing unaudited work. Note the leverage too. FCFF more than triples, from 42.8 to 149.0, while revenue grows only ~52%, from 920 to 1,399. That's operating leverage flowing straight to owner cash.

P14 (independent: the equity bridge). A completed FCFF valuation of Deccan Cements (synthetic) gives an operating-asset value, or enterprise value, of ₹52,000 cr. The balance sheet shows surplus cash & non-operating investments of ₹6,000 cr. Market value of debt, including leases, is ₹18,000 cr. Minority interest is ₹2,000 cr. Diluted shares outstanding are 40 cr. (a) Compute equity value per share. (b) Name the earlier boundary decision that places each non-EV item where it sits.

Solution. (a) Equity value = EV + cash & non-operating assets − debt − minority interest = 52,000 + 6,000 − 18,000 − 2,000 = ₹38,000 cr. Per share, that's 38,000 ÷ 40 = ₹950. (b) Go back to the bridge in the matching rule. Cash was excluded from non-cash WC, so it re-enters here as a +. Lease liabilities were counted as debt in WACC's weights, so they exit here as a −. FCFF was pre-interest, so all debt exits at market value. Minority interest is the slice of consolidated value that isn't the parent's. Each item crosses once and only once. Forget the +cash and equity reads ₹32,000 cr, which is ₹800/share. That's a ₹150/share error from one skipped line.

P15 (independent: reconcile the two reinvestment routes). For Vidarbha Steel (synthetic), next year's revenue rises ₹3,000 cr. Under Route A, with explicit lines, growth capex is ₹900 cr and working-capital build is ₹300 cr. Maintenance capex ≈ D&A, so they cancel in reinvestment. (a) Compute Route-A reinvestment and the sales-to-capital ratio it implies. (b) An analyst instead assumes S2C = 3.0. How much reinvestment does that imply, and how do you resolve the gap?

Solution. (a) Route-A reinvestment = growth capex + ΔWC (maintenance capex − D&A = 0) = 900 + 300 = ₹1,200 cr. The implied S2C = ΔRevenue ÷ reinvestment = 3,000 ÷ 1,200 = 2.5. (b) At S2C = 3.0, reinvestment = 3,000 ÷ 3.0 = ₹1,000 cr. That's ₹200 cr below the engineering, a 17% gap. It's inside the ~25% reconciliation band, so it isn't alarming. But the capex plan is physical. A steel plant's per-tonne cost is engineering, not assumption. So the rule from the reinvestment work applies: believe the engineering and re-fit S2C to ~2.5. The analyst's rounder 3.0 was making growth ₹200 cr/yr cheaper than the plant actually costs.

P16 (timed/stretch, 25 minutes: forecast + audit + value anatomy). Kaveri Cements (synthetic, Indian) has base revenue of ₹20,000 cr, growing 9%, 9%, 8%, 7%, 6%. The EBIT margin glides 14.0%, 14.5%, 15.0%, 15.5%, 16.0% on scale and cost programs. Tax is 25%, sales-to-capital 1.5, WACC 12% (INR). (a) Build the five-year FCFF table and discount it. (b) Run the audit (implied incremental ROIC and the RR path) and give a verdict. (c) A full DCF with terminal value puts enterprise value at ₹30,000 cr. What fraction of value falls inside your explicit five years, and what does that teach?

Solution. (a) (₹ cr)

PV of explicit FCFF = ≈ ₹5,769 cr. (b) Implied incremental ROIC = 0.75 × margin × 1.5, rising from 15.8% to 18.0% with the glide. RR falls from 52% (1,200 / 2,289) to 31% (1,098 / 3,493) as growth fades and margins lift. The verdict is that a 16 to 18% ROIC for a cement leader is defensible, but look at the input doing the work. S2C = 1.5 exceeds even the brownfield best case (~1.3, the coherent bull for the cement leader). So the model leans optimistic on capital efficiency. Re-fit toward 1.3 and FCFF falls. Flag the margin glide and the S2C together, because both are thumbs on the scale. (c) Explicit five-year PV ₹5,769 cr ÷ EV ₹30,000 cr ≈ 19%. Roughly four-fifths of the company's value is beyond the explicit window, in the terminal value. So a five-year forecast is the approach road. Even a ten-year one is never the destination. Distrust anyone's "the DCF says ₹X" until you've interrogated their terminal assumptions.

Applied mini-project

The task (≈4 hours): build a real ten-year FCFF forecast, twice. Pick one Indian company from Dabur, Havells, Pidilite, or Trent, and one US company from Starbucks, Nike, Costco, or Lowe's. They're deliberately non-financial, non-deep-cyclical names with readable segment data. Every one is a teaching specimen and not a recommendation.

Data pull (just this, nothing exotic):

1. For India, get the last 3 annual reports from the company website or BSE/NSE filings. Take 10-year financials from screener.in (export to a spreadsheet). For the US, get the latest 10-K from SEC EDGAR, reading Items 7 to 8 for the cash-flow statement, the lease note, the segment note. Take 10-year history from stockanalysis.com or TIKR (free tier).
2. From Damodaran's data page, take the sector's operating margin, sales-to-capital, and ROIC tables (US + emerging-markets datasets, latest January vintage).
3. Use last week's mini-project WACC if you built one for these names. Otherwise spend 30 minutes assembling a quick WACC from last week's stations, and document your sources.

Build, per company:

1. Base-year sheet: extract and clean the base year. Show reported figures, then adjustments, then normalized figures, with one line of justification per adjustment. Compute base FCFF by two routes (EBIT and CFO), noting the interest classification you found.
2. Driver sheet: revenue architecture (state which of the three and why), margin path with named mechanisms, reinvestment by S2C and a Route-A cross-check. Add a tax glide if relevant.
3. The ten-year table (Build 1/2 format) ending in PV of explicit FCFF at your WACC.
4. Audit panel: RR path, invested-capital line, implied book and marginal ROIC, g = RR × ROIC check at year 10, base-rate sanity (implied 10-yr CAGR vs Mauboussin's tables and vs India/US nominal GDP).
5. A one-page story memo (Minto: conclusion first): the base-case narrative, the two biggest assumptions, what the bear breaks, what you'd monitor.

Scoring rubric (self-score 0/1/2 per row; pass ≥ 12/16):

Park both models. The next module bolts terminal value, sensitivity, and a reverse-DCF onto these same sheets. The phase capstone then reuses the whole approach end to end.

Reading & resources

Core (do these):

• Damodaran, Investment Valuation (3rd ed.): Ch. 10 (From Earnings to Cash Flows), Ch. 11 (Estimating Growth, the g = RR × ROIC engine), Ch. 14 to 15 (FCFE and FCFF models). It's the canonical treatment, and this week compresses it. [Paid] [Intermediate]
• Damodaran, The Little Book of Valuation (2nd ed.): Ch. 2 to 4 for the same machinery in 40 pages, and the ideal pre-read before the big book. [Paid] [Beginner]
• Damodaran's free NYU Valuation course (YouTube/his site, refreshed annually): the sessions on estimating cash flows, growth, and the sales-to-capital method (roughly sessions 9–13 in recent vintages), plus the fcffsimpleginzu.xls spreadsheet: download it, trace every formula, then rebuild it blank per working agreement #2. [Free] [Intermediate]
• Damodaran's data page (pages.stern.nyu.edu/~adamodar): margins by sector, sales-to-capital by sector, ROIC by sector, US/global/emerging datasets, each January. Your forecast anchors live here. [Free] [All levels]
• Mauboussin & Callahan, The Base Rate Book (Credit Suisse, 2016): the sales-growth and margin base-rate tables behind the revenue base rates you use here. It's a free PDF, so keep it beside every forecast you ever build. Pair it with Mauboussin & Rappaport, Expectations Investing (rev. ed., 2021), Ch. 1 to 3 now, for the price-implied mindset M3.05 completes. [Free / Paid] [Intermediate]
• McKinsey (Koller, Goedhart & Wessels), Valuation (8th ed.): the chapter arc from "Reorganizing the Financial Statements" through "Analyzing Performance" to "Forecasting Performance." That's chs. 9 to 11 in older editions and one to two chapters later in the 8th, so use the TOC. It's the corporate-finance-department view of the same forecasting method, heavier on NOPAT/IC mechanics. [Paid] [Advanced]

Reinforcement & India grounding:

• Zerodha Varsity, Fundamental Analysis module, the DCF chapters (free, India-flavored, gentle). Read it afterwards and notice what it simplifies away. [Free] [Beginner]
• Two real filings this week: one Indian annual report (any mini-project name: MD&A, segment note, cash-flow statement, borrowings note) and one 10-K (cash-flow statement + lease note), extracting base-year FCFF by two routes from each. The filings are on BSE/NSE or the company site, and on SEC EDGAR. [Free] [Intermediate]
• CFA Program curriculum, Level II equity reading on free-cash-flow valuation: dry but airtight on FCFF/FCFE adjustments, and a good drill source if you have access. [Paid] [Intermediate]
• Penman, Financial Statement Analysis and Security Valuation (5th ed.), Part II, the accounting-anchored dissent: forecasting residual income instead of cash flows. Read a chapter so you know the argument exists. It sharpens what you learn here without replacing it. [Paid] [Advanced]

Going deeper:

• Aswath Damodaran, "Ups and Downs: Valuing Cyclical and Commodity Companies" (working paper, free on pages.stern.nyu.edu / SSRN), on how to normalize a mid-cycle base and forecast through a cycle instead of off a single year. It treats the peak-basing landmine from the base-year audit in full, ahead of M3.10. Read it the first time a cement, metals, or chemicals name turns up in your mini-project. [Free] [Advanced]
• CFA Institute, Equity Asset Valuation (Pinto et al.), the FCFF/FCFE chapters: a second rigorous pass on the same adjustments (leases, stock-based comp, non-operating assets, the equity bridge) with fully worked problems and end-of-chapter drills. It's the ideal drill source if Damodaran's treatment feels terse and you want reps before the mastery check. [Paid] [Intermediate]

The Modern Analyst's Addendum

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

AI-Augment this skill

{
  "skill": "Building FCFF line by line, the three routes and their classification traps, FCFE, the matching rule, cleaning a base year, forecasting revenue, margins and reinvestment, the g = RR × ROIC audit, and scenarios as stories",
  "use": "A DCF is a structured argument, and a tool is genuinely good at *stress-testing an argument* and genuinely bad at supplying its premises. Split the work on exactly that line. (1) Hand it your completed forecast (the driver table, not the output) and ask it to state, for each year, the implied incremental ROIC and how that compares with the company's own history and its sector distribution. That is §4.10's audit, and having a second party compute it catches the sin faster than re-reading your own model. (2) Ask it to write the *downside story* rather than a downside number: which drivers move together, in what order, with what lag. §4.11's whole point is that a bear case with peak margins is not a bear case, and a tool that has read a thousand cyclical downturns is a decent generator of the linkage list you then edit. (3) Ask it to build the revenue architecture options (top-down market share, bottom-up units and price, cohort build) and to name the observable each one commits you to; the architecture choice is the forecast's biggest single decision and it should be made deliberately. (4) Ask it what would have to be true for your growth path to hold, and then go check those things in the filings and the industry data.",
  "tools": ["Chat assistants: Claude, ChatGPT for the ROIC audit as a second opinion, the downside-linkage list and revenue-architecture options", "Document/PDF AI: for segment disclosures, capacity notes and management guidance across several years of filings and calls", "Coding assistants: for the FCFF build, the audit as an assertion, and the driver surface below"],
  "prompt": "Here is a ten-year forecast driver table for <COMPANY>: revenue growth, EBIT margin, tax rate, capex, D&A and change in working capital by year, plus the historical invested capital and NOPAT. Compute, for each forecast year, the reinvestment rate, the implied incremental ROIC (ΔNOPAT ÷ prior-year reinvestment) and the implied book ROIC, and present them as a table with the company's own historical ROIC beside them. Then list what would have to be true about the business for the implied ROIC path to hold. Do not adjust my drivers, do not compute a valuation, and do not tell me whether the forecast is reasonable — I am judging that.",
  "verify": "The audit is the verification, and it must be run on the model rather than on the narrative: read g and the reinvestment rate off the built model, compute the implied ROIC both ways, and confront it with the company's achieved history and the sector distribution. If the implied ROIC is indefensible, fix one lever the story supports — never average your way out. Then check the matching rule mechanically: FCFF meets WACC and enterprise value, FCFE meets the cost of equity and equity value, and a mismatch is not a small error. Every input to the base-year clean-up traces to a filing line, and the three routes to FCFF must reconcile to the rupee before any of them is forecast. A model asked to 'build a DCF for X' will produce a complete, internally smooth, entirely unfounded forecast, and its smoothness is the danger (Guardrail 1-5).",
  "diy": "The gate is unaided: you build FCFF from a filing by all three routes, clean a base year, construct a ten-year driver forecast, and run the g = RR × ROIC audit on paper. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You built FCFF three ways and reconciled them, cleaned the base year, forecast ten years of drivers, and ran the audit with a calculator. Keep every step of it. The audit works because you know which line each number came from. What code adds is that the audit stops being something you remember to do and becomes something the model cannot run without.

Today's workflow. Write the forecast as a driver dictionary and a function, and three things change.

First, the audit becomes an assertion. g = RR × ROIC is an identity, so the model can assert it every time it runs. It can refuse to produce a valuation when the implied incremental ROIC exceeds a stated ceiling. That turns the ninety-second check into a guarantee instead of a habit. The demonstration above is the test case. A firm with ₹900 crore of NOPAT, 15% growth and reinvestment of only ₹150 crore is implicitly claiming a 90% incremental ROIC on those figures (the rounded reinvestment rate of ~15% used earlier makes it 100%). Either way that's some six times its achieved 16%. Reinvestment at its achieved 16% would require 93.75% of NOPAT, leaving free cash flow of ₹56.2 crore instead of ₹750 crore. That's an overstatement of 13.3×, and an assertion catches it where a spreadsheet doesn't.

Second, scenarios become objects instead of columns. A scenario is a named set of linked drivers, so bull, base and bear are three dictionaries and the audit runs on each independently. That mechanically prevents the knob model, because a bear case that kept base-case margins and base-case reinvestment will fail its own audit.

Third, base rates become data. The arithmetic of humility wants to know how often companies of a given size actually sustain a given growth rate. You answer that from a panel, once, and then carry it as a prior into every forecast.

Tools & sources (IN + US). Use pandas for the driver table and the panel, numpy for the surface and pytest for the audit assertions. In India, the consolidated annual report gives you the base-year clean-up and the reinvestment lines. The segment note supports a business-by-business revenue architecture. Capacity and volume disclosures plus concall transcripts feed the unit build. Screener.in has the ten-year history you reconcile against filings, and RBI and MOSPI series give the nominal GDP the growth path must eventually respect. In the US, EDGAR XBRL companyfacts has the ten-year NOPAT and invested-capital series. The 10-K's segment and property notes cover capacity, and earnings-call transcripts give guidance. Damodaran's sector ROIC and reinvestment-rate tables at pages.stern.nyu.edu serve the confrontation step, and FRED (GDP, GDPPOT) gives the economy-wide ceiling.

# The growth-quality identity as an audit — R3: recomputed in-session, module's own figures
nopat, g, roic_hist, reinvest = 900.0, 0.15, 0.16, 150.0
rr_actual = reinvest / nopat                       # 0.1667 (the module rounds this to ~15%)
implied_roic = g / rr_actual                       # 0.90 -> a 90% incremental ROIC claim
rr_honest = g / roic_hist                          # 0.9375 -> growth at the ACHIEVED return
print(round(implied_roic, 2), round(rr_honest, 4),
      round(nopat * (1 - rr_honest), 1), round(nopat - reinvest, 1))   # 56.2 vs 750.0
assert nopat * (1 - rr_honest) < nopat - reinvest   # the free-growth model overstates FCFF
print(round((nopat - reinvest) / (nopat * (1 - rr_honest)), 1))        # 13.3x overstatement
w = 0.10                                           # value per ₹1 of NOPAT = (1 - g/ROIC)/(w - g)
print([round((1 - 0.06 / r) / (w - 0.06), 2) for r in (0.08, 0.10, 0.12, 0.16, 0.25)])
#   [6.25, 10.0, 12.5, 15.62, 19.0]  -> at ROIC = WACC the value is 10.00 at EVERY g

Verify. Reproduce the arithmetic above before you trust a pipeline. ₹400 crore reinvested at 15% adds ₹60 crore of NOPAT, which is 6% growth and equals 40% × 15%. 9% growth costs 75% of NOPAT at a 12% ROIC and 30% at a 30% ROIC. The specialty-chemicals model overstates near-year free cash flow 13.3×. Then move on from the arithmetic and verify the forecast's premises. A revenue architecture commits you to an observable (a market size, a share, a unit count, a price), and each of those has a source and a date. Reconcile the three FCFF routes before forecasting any of them. A model whose historical FCFF doesn't agree with itself won't become trustworthy by being extended ten years.

Quantitative lens

The identity g = RR × ROIC is usually taught as a check. Turned into a surface, it becomes something stronger: a map of which forecasts are possible, drawn before any judgment is applied.

Rearranged, free cash flow is NOPAT × (1 − g/ROIC). So every pair of growth rate and return has one payout ratio and only one, and the table is worth building once and remembering. At a 16% ROIC, growing 4% leaves 75.0% of NOPAT free, 8% leaves 50.0%, 12% leaves 25.0%, and 15% leaves 6.2%. At a 12% ROIC, 12% growth leaves exactly zero. The entire operating profit is consumed by the growth it's buying. At an 8% ROIC, 10% growth leaves −25%. The company has to raise external capital every year just to grow, and it's doing so to earn 8% on it. Nothing in that table is an opinion. It's arithmetic. It means a forecast pairing high growth with high free cash flow is making a claim about ROIC whether the analyst intended one or not. That's the growth-quality point from earlier, now visible as a whole surface instead of one example.

Push it one step further into value and the surface says something the check can't. At a 10% cost of capital, the value of ₹1 of NOPAT in perpetuity is (1 − g/ROIC)/(WACC − g). Read the resulting table across a row and the celebrated result appears by itself. In the ROIC = 10% column, value is exactly 10.00 at every growth rate: 0%, 2%, 4%, 6%, 8%, all identical. Growth at the cost of capital is worth nothing at all, because the reinvestment it requires uses up all the value it creates. In the ROIC = 8% column, value falls as growth rises: 10.00 at zero growth, 9.38 at 2%, 8.33 at 4%, 6.25 at 6%, and 0.00 at 8%. In the 25% column it rises steeply: 10.00, 11.50, 14.00, 19.00, 34.00. That's three different signs of the same derivative, from one input. It's the value-driver argument you'll meet later in the phase, available here as a by-product of the identity you just derived. And it's the reason the ROIC audit is more than a bookkeeping formality. It decides whether the growth in your model is an asset, a wash, or a liability.

The surface has limits, and they matter for how you use it. The identity assumes returns on existing capital are unchanged and that all growth is bought with new capital. But efficiency growth, the third consequence from earlier, is real, comes free, and belongs in the early years. So a strict audit applied to year one will fail a legitimate forecast. The audit becomes non-negotiable at the terminal boundary, and that's where the terminal-value module enforces it. The surface also treats ROIC as a single constant when it's a path. The identity needs the marginal return on new capital, while book ROIC is what you can measure. The two diverge most for the fast-growing companies where the stakes are highest. Finally, the value table is a perpetuity, so it exaggerates. A real company earns a high ROIC for a competitive advantage period and then fades, which caps how much of that 34.00 anyone should believe. Treat the surface as a feasibility map. It isn't a valuation.

Do it in code: build the (g, ROIC) payout and value surfaces once, mark your forecast's implied point on them, and make the audit an assertion the model runs instead of a step you remember. Then check where your terminal year falls on the map. See G1/QM1.01 and G2/CS2.06.

Where this goes next: galaxy cross-links

• G2 · CS2.06 Software Engineering & AI-Assisted Development, where a forecast is software: drivers as data, the ROIC audit as a test that fails loudly, scenarios as named objects, and version control so last month's model is still runnable.
• G2 · DA1.02 Pandas Data Wrangling: the ten-year driver panel and the base-rate work behind the arithmetic of humility in revenue forecasting, where "how often does a company this size sustain this growth" becomes a query.
• G1 · QM1.01 Probability, Distributions & Estimation: priors, base rates and what a distribution of outcomes does to a forecast built as a single path.
• G3 · CN2.02 Market Sizing & Guesstimates: the consulting craft behind the revenue architectures: top-down, bottom-up, and the observable each one commits you to defending.

Teach it back & journal

Feynman prompt. Explain to a smart 15-year-old why a company can grow fast for ten years and still hand its owners almost no cash, and why a slower-growing company might hand them a lot. Use a chain of lemonade stands. Each new stand costs ₹20,000 to build and earns ₹3,000 a year, which is 15%, decent but not magic. Set that against a friend's single famous stand that earns ₹6,000 on ₹10,000 invested (60%) and has no room for a second stand nearby. Walk the money. Where does each year's profit go, who decides, and when does the cash finally reach the owner's pocket? End with the one-line rule: growth costs reinvestment, and the return on reinvestment decides whether growth is worth it. Then write it up as a one-page explainer with a small table of both stands over five years. If the table doesn't force the conclusion by itself, the explainer isn't done.

Journal reflection. Find a forecast you personally made and believed, in any domain. It might be a hiring plan that assumed a team could double output without new people, a project timeline that assumed zero rework, a salary trajectory, the course's own schedule. Then write down three things:

1. What the "free growth" assumption was, meaning the output you projected without pricing its input.
2. What the real reinvestment line would have been.
3. What an implied-ROIC-style audit (output claimed ÷ input budgeted, checked against your own base rates) would have flagged in advance.

Then commit in writing to the standing rule for every model you build from this week on: no growth number survives review until I have computed what it costs, and checked what return that cost implies.

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