Returns on Capital Done Right

Phase 2 · Week 20 · ~20 hours

Where it sits. You're about to build the single most important number in the program from raw statements, with your own hands. It's return on invested capital. The master map's five-questions spine (section 2.1) puts it plainly. Question 2 of the five is "Is it a good business — is ROIC durably above the cost of capital?", and you'll answer it properly here. You already have the margin stack and DuPont, plus liquidity, solvency and the cash conversion cycle. Everything after leans on what you build here. Free cash flow shares its operating/financing separation, and the cost-of-capital work builds the WACC you'll compare it against. The value-driver work proves growth destroys value when ROIC < WACC. The moat weeks (M4.04–4.06) use ROIC persistence as the numbers-test for moats, and the deep-dive workbook has an incremental-ROIC row you'll fill in for every company you ever study. Next comes cash flow analysis and free cash flow.

Suggested hour budget. Core content 8h, worked examples 2h, practice set 5h, mini-project 3h, and flashcards plus the mastery check 2h.

App link. What you learn here powers the Quality & Red Flags branch's ratio-builder drills ("build NOPAT from this P&L," "sort these balance-sheet lines into operating vs financing") and the case engine's "is it a good business?" scoring. The valuation sandbox later enforces the g = reinvestment × ROIC consistency you learn here.

Learning objectives

By the end you can:

1. Build a clean NOPAT from any income statement: identify the right EBIT, strip one-offs and non-operating income, and choose (and defend) the tax rate you apply.
2. Build invested capital both ways, the operating approach (working capital + fixed assets + other operating assets) and the financing approach (debt + equity − non-operating assets), then reconcile the two to the last rupee or dollar on any balance sheet.
3. Explain the operating–financing separation: why the balance sheet must be re-sorted before it can tell you whether the business (as opposed to its funding) is any good.
4. Compute and correctly deploy ROIC, ROCE (Indian convention), ROE, and ROA: state what question each answers, when each misleads, and convert between pre-tax and post-tax hurdles.
5. Execute the three adjustments that most often change the verdict: operating-lease capitalization (pre- and post-ASC 842/IFRS 16/Ind AS 116), capitalized R&D per Damodaran, and the goodwill-in/goodwill-out double view. Say when each matters.
6. Compute incremental ROIC (ROIIC) over multi-year windows, link it to growth via g = reinvestment rate × ROIC, and use it as the compounder test.
7. Judge a return spread against a cost-of-capital hurdle (~12% post-tax for India, ~8–9% for the US, as rules of thumb until M3.03) and apply the persistence/mean-reversion base rates to decide whether the spread is believable going forward.
8. Run a full blank-sheet ROIC build on a never-seen company in ≤25 minutes: the gate for this module and a permanent row in your deep-dive workbook.

Prerequisites & connections

Six earlier modules feed this one. M1.03 taught you to classify any balance-sheet line cold, which matters because the whole sheet gets re-sorted here. M1.05 gave you the linkage, so you know where every number comes from and what it ties to. M1.07 covered goodwill mechanics and R&D expensing vs capitalization. M1.08 covered leases under ASC 842 vs IFRS 16/Ind AS 116, and that comparability trap returns here with money attached. M1.09 covered minority interest, which re-appears in the financing approach. M2.01's DuPont told you ROE mixes operations with leverage. Here you do the surgery that separates them.

It feeds forward everywhere. M2.04 uses the same operating/financing separation on the cash flow statement, and M2.05 benchmarks ROIC across peers and time. M3.03 supplies WACC, the other side of the spread, and in M3.04–3.05 the DCF reinvestment logic turns g = RR × ROIC into a forecasting rule. M3.07 covers value drivers, EVA and the key-value-driver formula. In M4.04–4.06, a moat that doesn't show up as persistent ROIC > WACC is just a story. Every M5 sector playbook carries a returns row, and there you'll learn where ROIC does not apply. M8.03's deep-dive history spreadsheet computes 10 years of ROIC and 5-year ROIIC for every target.

A business is a machine that turns capital into profit, and ROIC is that machine's measured efficiency. You cannot read it off any statement. You must build it, because the accountants who built the statements were answering a different question.

Core teaching content

4.6 The ratio family: four returns, four questions

All four ratios come from the same statements. They differ in whose profit goes on top and whose capital goes below. Keep the claimholders consistent, and know which question you're asking.

India runs on ROCE, so pin it down. By convention it's pre-tax: ROCE = EBIT ÷ capital employed. Screener-style implementations typically compute EBIT as PBT + finance costs, which slips other income in. They compute capital employed as total assets − current liabilities, which includes the cash pile and excludes short-term borrowings. Check any tool's own definition page before you trust it. On Kaveri, screener-style ROCE = (376 + 38) ÷ (1,900 − 435) = 414 ÷ 1,465 = 28.3%, against a clean pre-tax ROIC of 360 ÷ 1,190 = 30.3%. They're close here. They diverge badly when cash piles, exceptional items or short-term debt are large. Use convention-ROCE for screening and for reading Indian history fast, and rebuild clean ROIC before you judge.

Never cross the streams. WACC is a post-tax hurdle and ROCE is pre-tax. Compare post-tax with post-tax (ROIC ≥ 12%), or gross the hurdle up: pre-tax hurdle = WACC ÷ (1 − t) = 12% ÷ 0.75 = 16% at Indian tax rates. You'll hear Indian analysts say sustained ROCE above ~20% is a strong business. That rule of thumb is this hurdle plus a margin of safety.

Each lens has its moment. Use ROIC for business quality, moat tests, comparisons across borders and capital structures, and all value-driver math. Use ROCE when you're speaking the Indian market's language or screening ten years of history at a glance. Use ROE when you're an equity holder asking about equity. It's also the primary lens for banks and financials. There, borrowing money is the business, and separating operating from financing means nothing. A bank's deposits are its raw material and ROIC doesn't compute, so use ROE vs cost of equity and ROA, as the banking playbook does. ROA almost never makes a headline, except for banks (>1% good, >1.5% excellent) and quick asset-intensity comparisons.

Kaveri gives you one company and four verdicts. PAT attributable is 276 (PBT 376 − tax 94 = 282, less minority 6). ROE = 276 ÷ 985 = 28.0%. ROA = 282 ÷ 1,900 = 14.8%. Convention-ROCE = 28.3%. Clean ROIC = 22.7%. That's four numbers for one business. The 28% ROE flatters, thanks to some leverage plus non-operating gains in PAT. A fair answer, neutral to capital structure, is 22.7% post-tax, and it's still excellent. Quote one of these without knowing which question it answers and you're reading tea leaves.

4.9 Adjustment 3: goodwill in vs out: the double view, not a choice

Goodwill is the premium paid over fair value in acquisitions. It's capital that left the shareholders' pockets but bought no separable asset. So do you include it? Wrong question. Compute both, because they answer different questions:

• ROIC with goodwill: "did the company's actual deployment of investor money, acquisition premia included, earn its keep?" This tests management and its M&A record. Judge a serial acquirer with goodwill in, because the premium was real cash (or real shares).
• ROIC without goodwill: "how good is the underlying business itself — and roughly what does a marginal, organic rupee of capital earn?" This tests the franchise and its moat. It's also the better guide to the returns future reinvestment can achieve, since new plants don't come with acquisition premia.

Memorize the pattern. A big gap, with-goodwill far below without-goodwill, means a fine business whose management overpaid to buy growth. With-goodwill ROIC ≈ WACC means acquisitions merely returned the cost of capital, which is treadmill M&A. Both views strong means a compounder. Both weak means a bad business, fairly reported.

Watch how impairment plays into this. Writing off goodwill shrinks the denominator and flatters subsequent with-goodwill ROIC. A company can destroy a billion, impair it, then show "improved returns." So keep a memory, and use gross goodwill (pre-impairment) when you audit a management's lifetime record. Worked Example 1 runs the double view on Microsoft, where the Activision premium makes it vivid.

4.10 Incremental ROIC: the compounder test

Headline ROIC is the average return on all capital ever invested, and much of that capital is decades old. Investors eat the future, and the future is built by new capital. So ask a sharper question. What is each incremental rupee earning?

Incremental ROIC (ROIIC) = ΔNOPAT ÷ Δinvested capital     (over a multi-year window)
Reinvestment rate (RR)   = Δinvested capital ÷ cumulative NOPAT over the window
Growth ≈ RR × ROIIC      (the engine equation — reappears as forecasting law in M3.04)

That growth identity is the deepest line in the whole build. A company can only grow NOPAT by reinvesting at some return, so growth is never free and never independent of returns. It also shows you the two ingredients of a great compounder, high ROIIC with a real runway to reinvest. And it shows the tragedies. One is high ROIC with nowhere to reinvest. That's a tollbooth. It's fine, but the cash must come back to you. Another is heavy reinvestment at sub-WACC returns, which is growth that actively destroys value (section 4.11).

Here's the 5-year worked example, on Avenue Supermarts (DMart). Figures are illustrative, modeled on ~FY20–FY25 scale, so pull the actual filings and screener.in history. ₹ crore, t = 25%:

Five-year incremental return:

ΔNOPAT  (FY20 → FY25) = 3,450 − 1,500 =  1,950
ΔIC     (FY20 → FY25) = 20,600 − 9,500 = 11,100
ROIIC (5-yr)          = 1,950 ÷ 11,100 = 17.6%
RR = 11,100 ÷ (1,275 + 1,800 + 2,700 + 3,075 + 3,450 = 12,300) = 90%
Growth check: RR × ROIIC = 0.90 × 17.6% ≈ 15.8% vs actual NOPAT CAGR (3,450/1,500)^(1/5) − 1 = 18.1%

Now read it like an analyst. (1) DMart reinvested nearly everything it earned (RR ≈ 90%) at ~17–18%, comfortably above the 12% hurdle. It's a true compounder, and the identity roughly closes. The gap between 15.8% and 18.1% is timing and margin drift. Treat the identity as a consistency check and not a law of single years. (2) Never compute ROIIC on one year. FY21 alone gives ΔNOPAT −225 ÷ ΔIC 1,500 = −15%. That reads as "value destruction", and it's really COVID hitting stores that were still being built. Single-year ROIIC is noise. Use 3–5-year windows, and longer ones for lumpy investors. (3) Refine for timing. Capital spent this year produces profit next year and after. So for fast growers a lagged version, with ΔNOPAT measured one year behind ΔIC, is fairer. ΔNOPAT from FY22 to FY25 ÷ ΔIC from FY21 to FY24 = 1,650 ÷ 6,800 = 24.3%.

State it as a range: "incremental returns in the high-teens to low-twenties". That's precise enough for the only question that matters. Is the marginal rupee earning well above the hurdle, and is there runway to keep deploying? For DMart-the-illustration it's yes and yes, the compounder verdict. In your deep-dive workbook, this table is a mandatory row for every company you study.

4.11 The hurdle, the spread, and how long spreads survive

The spread is the point: ROIC in isolation is a speedometer with no speed limit posted. The economic verdict is ROIC − WACC:

• Spread > 0 and reinvesting: every rupee retained creates more than a rupee of value. Growth is good.
• Spread ≈ 0: growth is a treadmill. A bigger company, with the same owner value. (This previews the conservation of value you'll meet in the value-driver work. Paying up for growth at zero spread buys nothing.)
• Spread < 0 and reinvesting: growth is a shredder. A crude perpetuity shows the size of it. ₹100 reinvested at 8% against a 12% hurdle creates 100 × 8/12 ≈ ₹67 of value, so each rupee of "investment in the future" destroys a third of itself. That's how capital-hungry, low-return empire building became the classic Indian value trap. It's also why the value-driver work will make you grade a decade of every management's capital allocation.

Until you build real WACCs, use ~12% post-tax for India and ~8–9% for the US. India's is higher because of its higher risk-free rate and equity premium. These are rules of thumb as of mid-2026, so verify them against Damodaran's country pages each January. The pre-tax Indian screen is ~16%, hence the folk rule "ROCE 20%+ is quality."

How long do spreads last? Mauboussin's evidence gives the base rates. Sort all US non-financial firms into ROIC quintiles and watch them for a decade. (These are Mauboussin & Callahan's transition matrices, built on 1990s–2010s data. See the 2014 and 2022 Counterpoint Global/Credit Suisse papers.) Every quintile drifts toward the middle. Competition works, and mean reversion toward the cost of capital is the default fate. But the drift is incomplete at the extremes. Roughly half of top-quintile firms are still in the top two quintiles ten years later (approximate, so read the paper's exhibits). Bottom-quintile firms are nearly as sticky, because bad industries stay bad.

Carry two implications forever. First, the burden of proof is yours. Shown a 35% ROIC, your default forecast is fade. You may override the base rate only with an identified moat (Phase 4's four lenses), stable market shares and a reinvestment runway. "Great numbers" are the question, never the answer. Second, durability is worth more than level. A 20% ROIC defensible for 15 years beats a 40% ROIC that arbitrage erases in three. That defensible period is Mauboussin's Competitive Advantage Period (CAP). You'll estimate CAPs in the moat weeks, and the market's implied CAP in reverse-DCFs.

Asian Paints is a live Indian example. It compounded at 25–30%+ ROCE for decades, and that spread is what finally summoned a giant new entrant (Birla Opus, 2024) and squeezed margins. High spreads advertise themselves. The capital cycle is how they erode.

When you meet a durable spread, name the barrier. When you can't name one, believe the base rate.

One more caveat. ROIC is also sector-shaped. Software and branded consumer run structurally high because their denominators are light on booked intangibles (section 4.8). Utilities and commodities run structurally low. Cyclicals swing with the cycle, so judge them on mid-cycle ROIC and never on the peak. Always benchmark within the sector and against the hurdle, over 5–10 years.

4.12 Blank-sheet build 1 (India): Asian Paints

Your gate is a blank-sheet build, so here's the full worked pattern. Numbers are illustrative, modeled on ~FY25 scale, so pull the actual consolidated filing. You'll find the annual report on the company site or BSE, and ten-year history on screener.in. ₹ crore.

Step 1 is clean NOPAT. Reported revenue is 34,000. PBT is 5,980, and that's after three things. Other income of 900 comes from interest/MTM on a ~₹5,000 cr investment book. There's an exceptional impairment of −100. And finance costs of 220 are mostly lease interest, already below EBIT under Ind AS 116. Clean operating EBIT = 5,980 − 900 + 100 + 220 = 5,400. The ETR is ≈ 25.2%, in line with statutory, with no one-offs. So NOPAT = 5,400 × 0.748 ≈ 4,040.

Step 2 is invested capital by the operating approach:

Step 3 is the financing approach and the tie-out. Borrowings 700 + lease liabilities 1,400 + equity 19,100 + minority interest 400 + net DTL 400 = 22,000. Take off cash 800, current investments 2,700 and non-current financial investments 1,500, and 22,000 − 5,000 = 17,000. It ties.

Step 4 is the ratio family. ROIC = 4,040 ÷ 17,000 = 23.8% (ex-goodwill 24.5%). ROE = 4,400 ÷ 19,100 = 23.0%. Convention-ROCE = (5,980 + 220) ÷ (28,000 − 6,500) = 6,200 ÷ 21,500 = 28.8%. That's higher than clean post-tax ROIC because it's pre-tax. Yet it's lower than clean pre-tax ROIC (5,400 ÷ 17,000 = 31.8%), because the ₹5,000 cr investment book sits idle in its denominator. You can now explain every wedge between the four numbers, and that's the skill.

Step 5 is the verdict. The spread is ≈ +12 points over the 12% hurdle, which makes this an elite franchise. Notice that ROE < ROIC. That's the cash-drag signature of a net-cash company, where the investment book earns treasury returns and dilutes the equity number. And notice the forward question the spread itself raises. With Birla Opus attacking, is the CAP shortening? Phase 4 answers that. Your job here was to measure fairly. (Worked Example 1 is the matching US blank-sheet build: Microsoft, with the goodwill double view and the R&D adjustment.)

4.13 The build ritual (your permanent checklist)

Every ROIC build, forever, takes nine moves. Moves (1) and (2) make the numerator. Clean EBIT first. Other income comes out, one-offs come out, and leases follow the standard. Then choose and state t to get NOPAT. Moves (3), (4) and (5) make the denominator. Sort the balance sheet so every line has one home. Operating IC is NWC + fixed + other operating. Financing IC is debt + leases + equity + MI + quasi-equity − non-operating assets. Move (6) reconciles to the rupee, and if it doesn't tie, you don't go on. Moves (7) and (8) compute ROIC on average capital (ending if you have only one sheet), with and without goodwill, then compare like-for-like against the hurdle, post-tax vs post-tax. For any multi-year view, move (9) adds the ROIIC table plus the g = RR × ROIC sanity check. Once you've practiced, it takes twenty-five minutes on a clean company, and the mastery check holds you to it.

Common mistakes & how experts think differently

1. Comparing a pre-tax return to a post-tax hurdle. "ROCE 18% against a 12% WACC — six points of spread." On equal terms it's 18% × 0.75 = 13.5% against 12%, barely one point. Label every return pre- or post-tax before you compare. Then either tax the return or gross up the hurdle (WACC ÷ (1 − t)). Half of all bad spread arguments die right here.

2. One-sided adjustments break the matching rule. Deduct the cash pile from invested capital while leaving interest income in NOPAT, and you flatter the ratio twice. Include ROU assets in IC for a US GAAP company while EBIT still carries full rent, and you punish it twice. Experts close every build with the matching-rule sweep. For each numerator item, where's its asset? For each denominator item, where's its income?

3. Trusting "capital employed" off a screen means trusting somebody else's sort. Total assets − current liabilities silently includes idle cash and excludes short-term borrowings. And the usual numerator, PBT + finance costs, silently includes other income. For a company with a big treasury book or heavy working-capital debt, convention-ROCE can land several points away from clean ROIC, in either direction. Experts read the definition page of every tool once, then rebuild before judging.

4. Ending capital in a growth or deal year distorts. Say a company doubles its invested capital in November through an acquisition. Full-year NOPAT ÷ year-end capital punishes it, because the deal never had time to produce a rupee. Use average capital as the default. For large mid-year deals, time-weight or go pro-forma. A mirror error comes right after a big write-down, when ending capital flatters (see mistake 6).

5. Single-year ROIIC, and peak-cycle ROIC, are the two timing errors. One year of ΔNOPAT ÷ ΔIC is noise. DMart's COVID year read −15% while stores under construction were about to earn well. And a cyclical at the top of its cycle prints a 30% ROIC that's really a 12% mid-cycle business in costume. Experts use 3–5-year ROIIC windows, with longer ones for lumpy investors. They judge cyclicals only on mid-cycle returns.

6. Reading denominator shrinkage as improvement is a trap. Goodwill impairments, asset write-downs and buyback-shrunken equity all raise later measured returns with zero operating change. Home Depot's equity is negative after two decades of buybacks, so its ROE literally can't be computed, while its ROIC remains ordinary arithmetic. Experts keep a gross-goodwill memory for judging management's lifetime record. When returns rose because capital fell, they treat it as a question and never as an achievement.

7. Extrapolating the level, ignoring the base rate, turns the burden of proof upside down. A 35% ROIC isn't a fact about the future. It's an invitation competitors have already received. The default forecast is fade toward the cost of capital, and you may override it only with a named barrier, stable shares and a reinvestment runway. Assuming everything mean-reverts is the opposite error. All of Phase 4 is about identifying the minority that resists.

8. Applying ROIC where the separation is meaningless begins with lenders. For banks, NBFCs and insurers, deposits and float are raw material. They aren't financing, so the operating/financing sort can't be performed. Use ROE vs cost of equity and ROA. Invested capital can also sit near zero or below, as at franchisors and some platform businesses whose suppliers and customers fund everything. Then the ratio explodes or flips sign. Switch to absolute NOPAT growth plus the durability of the float, because capital isn't the binding constraint.

9. "Great ROIC, therefore buy the stock" confuses two verdicts. ROIC answers question 2 of the five (is it a good business?). Question 5 (is the price beatable?) is a different matter. A 30% ROIC franchise at 70× earnings can be a poor investment, and a 13% ROIC business at 8× can be a fine one. Keep the three verdicts separate (good business, good management, good price). Never let one pass itself off as another.

So how do experts think differently? They think in spreads, always against WACC and always like-for-like on tax. A level on its own tells them little. They think in increments and ask what the marginal rupee is earning. They think in ranges. Placement judgments move the number a point or two, so they state them, stay consistent and never let a verdict hang on one. And they read the denominator as history. Every line of invested capital is an archived management decision, so a ROIC build is an audit of a decade of capital allocation.

Above all, they build the number themselves. The reflex to install for good is "whose number is this, and built how?"

Worked examples

Worked example 1: Microsoft: the US blank-sheet build, both signature adjustments included

All figures are illustrative, modeled on ~FY25 scale, so pull the actual 10-K from SEC EDGAR (June year-end). $bn. This is the US twin of the Asian Paints build in section 4.12, and it's where the 105 and 320 used in section 4.8 come from.

Step 1 is clean NOPAT. Revenue is ~280 and reported operating income is 128. Run two checks before you accept that as clean EBIT. The first is geography. US GAAP parks "other income (expense), net", meaning interest on the treasury portfolio, investment gains and interest expense, below operating income. So unlike an Indian Schedule III P&L, there's usually nothing to strip out of the top. Still, scan the notes for impairments or restructuring buried inside operating expenses. Assume none are material this year. Next, check leases. Microsoft reports under ASC 842, so straight-line rent on operating leases (offices, some datacenter capacity) stays inside EBIT. Operating lease liabilities are ≈ 27 at ~4% implicit, so imputed interest is ≈ 1.1, under 1% of EBIT. Note it and call it immaterial. Then move on. What matters is checking every time. You won't need to adjust every time. Then tax. The ETR is ≈ 18%, persistently below the ~25% combined US statutory rate. Its tax note attributes the gap to the FDII deduction, R&D credits and foreign-rate mix. All of those are structural and durable, so use the ETR for the current-year build. NOPAT = 128 × 0.82 ≈ 105.

Step 2 is invested capital by the operating approach:

Two lines deserve a second look. Take the −55 of working capital first. Customers prepay for cloud and licenses (unearned revenue) and suppliers wait, so other people fund $55bn of the operation. That's a structural return-raiser, the same as the Indian FMCG pattern in section 4.4. Then the 205 of PP&E. The AI datacenter build-out has made "capital-light software" a period piece. Capital intensity is rising sharply, and that's why the returns arithmetic on new capital (ROIIC) now matters more than the glorious average.

Step 3 is the financing approach and the tie-out. Borrowings incl. finance leases 60 + operating lease liabilities 27 + shareholders' equity 345 = 432. Take off cash & short-term investments 95 and equity and other non-operating investments 17, and 432 − 112 = 320. It ties. One wrinkle is worth knowing. The real 10-K carries a net deferred tax asset, and not a liability. The simple convention treats a net DTA as the mirror image of a net DTL. Deduct it on the financing side like a non-operating asset and keep it out of operating assets. In this illustration it nets to ≈ 0.

Step 4 brings the double view and the ratio family. ROIC with goodwill = 105 ÷ 320 = 32.8%, and ex-goodwill = 105 ÷ 200 = 52.5%. The 120 of goodwill is dominated by Activision Blizzard (closed 2023, ~$69bn, mostly premium over separable assets), plus LinkedIn, Nuance and GitHub. Read it per section 4.9. The franchise earns ~52%, which is elite. The premium-included record earns ~33%. That's still nearly four times the ~9% US hurdle, so even after paying real acquisition premia, management's total deployment cleared the bar with room to spare. The 20-point gap is the measured price of buying growth instead of building it. ROE ≈ 97 ÷ 345 ≈ 28%, below ROIC. It's the same net-cash cash-drag signature you saw on Asian Paints in section 4.12. Two very different companies get one diagnosis. When ROE < ROIC, look for the treasury pile.

Step 5 is the R&D overlay and the verdict. Layer on section 4.8's Damodaran adjustment with a 3-year life. Your research asset comes to ≈ 60. Adjusted NOPAT becomes 110 on adjusted IC of 380, which gives 28.9% with goodwill. So report a fair range: high-20s to low-30s post-tax against a ~9% hurdle, on $300bn+ of capital. That's an elite spread at extraordinary scale. Base rates say fade. Moat evidence argues for a slow fade, though Phase 4 will make you prove both the switching costs and the network effects. And the rising capital intensity tells you to watch ROIIC more closely than ROIC.

With the 10-K open, the whole build takes about 20 minutes. That's the gate standard.

Worked example 2: The leverage illusion: one machine, two ROEs, and the surgery that separates them

Meet two synthetic twins (₹ crore, t = 25%), with no leases and no non-operating assets, so that only one force is at work. Vindhya Cables and Malwa Cables run identical operations, with clean EBIT of 160 on invested capital of 1,200. NOPAT = 160 × 0.75 = 120, so ROIC = 120 ÷ 1,200 = 10.0% for both companies, by construction. That's below the 12% Indian hurdle. You're looking at a mediocre business, twice.

The funding differs. Malwa is all-equity: 1,200 of equity. Vindhya funds with 900 of debt at 9% and 300 of equity (D/E = 3).

It's the same machine, yet ROE says 10% for one and 19.75% for the other. The wedge is pure financing, and it obeys an exact identity (book values, no non-operating assets or minorities):

ROE = ROIC + (ROIC − rd(1−t)) × D/E
    = 10% + (10% − 9%×0.75) × 3 = 10% + (10% − 6.75%) × 3 = 19.75%  ✓

Leverage adds (spread over after-tax debt cost) × (debt per rupee of equity). It manufactures ROE whenever ROIC > after-tax cost of debt, even when ROIC is below WACC, as it is here. Vindhya's 19.75% ROE dresses up a value-destroying business, because equity capital costs more than 6.75% debt.

Now run a downturn. EBIT falls 30% to 112, and ROIC drops to 84 ÷ 1,200 = 7.0% for both twins. Malwa's ROE eases from 10.0% to 7.0%. Vindhya's collapses from 19.75% to (112 − 81) × 0.75 ÷ 300 = 7.75%, a twelve-point swing. Meanwhile its interest cover falls from 160/81 ≈ 2.0× to 112/81 ≈ 1.4×, which is solvency alarm territory. The amplifier works both ways.

That's DuPont's warning, made surgical. ROE blends the machine with its funding, and ROIC isolates the machine. You need both numbers, and you must never let one impersonate the other.

Worked example 3: The empire builder: growing profits, shrinking value

Here's the classic Indian capital-hungry conglomerate, Bharat InfraCorp (synthetic; ₹ crore; t = 25%). Management's slide deck says "NOPAT up 44% in five years." That's true. Here's the full table you'd build instead:

ΔNOPAT (FY20→FY25) = 1,150 − 800  =    350
ΔIC    (FY20→FY25) = 13,000 − 8,000 = 5,000
ROIIC (5-yr)       = 350 ÷ 5,000   =   7.0%
Cumulative NOPAT FY21–25 = 860 + 925 + 1,000 + 1,075 + 1,150 = 5,010
RR = 5,000 ÷ 5,010 ≈ 100%
Check: g = RR × ROIIC = 1.00 × 7.0% = 7.0% vs actual NOPAT CAGR (1,150/800)^(1/5) − 1 = 7.5%  ✓ ties
Lagged variant: ΔNOPAT FY21→FY25 = 290 ÷ ΔIC FY20→FY24 = 4,000 → 7.25%  (robustly ~7%)

It retained all of them, RR ≈ 100%, and every retained rupee went to work at ~7% against a 12% hurdle. Use the crude perpetuity from section 4.11 to size the damage. ₹5,000 crore deployed at 7/12 of its cost ≈ ₹2,900 crore of value, so roughly ₹2,100 crore destroyed. That's about 42 paise vaporized per rupee reinvested, while profits grew every single year. Headline ROIC only drifted from 10.1% to 9.2%, because five years of bad marginal capital is diluted by decades of old capital.

So ROIIC is the compounder test, and the direction of the drift is a tell in itself.

What would fix it? Stop growing. Harvest, repay debt and return the cash, and NOPAT would flatten while value would rise. Growth is a multiplier on the spread, and the spread is negative. Put it next to section 4.10's DMart table, where RR ≈ 90% at ~18–24% ROIIC. Same reinvestment appetite, opposite verdict. Appetite was never the question. What it earned was. You'll formalize this as economic profit and the key-value-driver formula in the valuation phase. For now it's enough to see profits up and value down in one table.

Worked example 4: Ending or average? One machine, two ROICs, and the economic-profit rupees (India)

Konkan Ceramics Ltd (synthetic; ₹ crore; t = 25%) grew hard this year, so for once you have two balance sheets. That lets you do what the single-sheet builds in section 4.5 and section 4.12 couldn't. You can compute ROIC on average capital, the default of the build ritual (section 4.13, move 7), and watch how much the choice moves the number.

Step 1 is clean NOPAT for FY25. Reported operating profit of 340 includes ₹30 of interest income on the mutual-fund book and a ₹10 one-time gain on scrapping old kilns. Both leave the numerator (section 4.3). Clean EBIT = 340 − 30 − 10 = 300, and NOPAT = 300 × 0.75 = 225.

Step 2 is invested capital, both years and both ways. Sort each sheet with the section 4.2 rule and reconcile, so operating must equal financing:

Both years tie to the rupee. That's the built-in error check of section 4.2, run twice. Goodwill is unchanged because there was no acquisition this year, so it cancels out of every incremental comparison. It's a first hint of why the double view and the average are separate choices.

Step 3 is ROIC four ways, and why ending capital lies in a growth year. Invested capital grew +22.4% during FY25, from ₹1,250 to ₹1,530 with goodwill. The year's ₹225 of NOPAT was earned on capital that climbed all year. Divide by the year-end figure and you charge the business for money it didn't have for most of the period:

The gap between ending and average is ~1.5 points, and the growing denominator is the only thing producing it. Make the average of opening and closing IC your default. Reach for ending only when a single balance sheet is all you have (section 4.13, mistake #4). Read the goodwill columns per section 4.9. The franchise earns ~18% (average, ex-goodwill). The full deployment including the ₹150 of acquisition premium earns ~16%. Both are comfortably above the 12% hurdle, so the acquisition that created the goodwill cleared the bar.

Management's money was well spent.

Step 4 is the economic-profit rupees. ROIC states the machine's efficiency. Economic profit (EVA) states the rupees of value it threw off above its capital charge. That's the master idea's (ROIC − hurdle) × capital, in the form the valuation phase will formalize:

Economic profit = NOPAT − hurdle × average invested capital
   with goodwill: 225 − 12% × 1,390 = 225 − 166.8 = +₹58.2 cr/yr
   ex-goodwill:   225 − 12% × 1,240 = 225 − 148.8 = +₹76.2 cr/yr

After paying 12% for every rupee investors tied up, Konkan still generated ~₹58 crore of real surplus this year. On the operating franchise alone, before charging for the acquisition premium, it was ~₹76 crore. A positive spread became positive rupees. Had ROIC been below 12%, this line would have turned negative while net profit stayed comfortably positive. That's section 4.11's warning, made concrete in currency.

Worked example 5: The serial acquirer's two verdicts: franchise economic profit vs M&A economic profit (US)

Cascadia Diagnostics Inc. (synthetic; $m; US GAAP; t = 25%) rolled up dozens of independent labs over a decade, so $1,800m of its balance sheet is acquisition goodwill. It owns its labs, so operating leases are immaterial (checked and set aside per section 4.7). It reports no one-offs either. So the build is clean, and the whole lesson comes from the goodwill double view (section 4.9), expressed here for the first time in economic-profit dollars.

Step 1 is NOPAT, with a bottom-up cross-check. Reported EBIT is 640, so NOPAT = 640 × 0.75 = 480. Its P&L is simple: interest expense 88, interest income on the $300m treasury 12, PBT 564, tax 141, net income 423. That means you can verify NOPAT from the bottom up, and it's a habit worth keeping:

NOPAT = net income + after-tax interest expense − after-tax non-operating income
      = 423 + 88×0.75 − 12×0.75 = 423 + 66 − 9 = 480  ✓  (ties to EBIT × (1−t))

Two roads lead to the same 480. When they disagree, a stray item is misclassified.

Step 2 is invested capital, both approaches with the tie-out. On the operating side, NWC = (520 + 90 + 60) − (240 + 180 + 80) = 170. Add net PP&E 1,400 and operating intangibles 300 for 1,870 ex-goodwill, then 1,800 goodwill for 3,670. On the financing side, (60 + 1,400) debt + 2,410 equity + 100 net DTL − 200 cash − 100 short-term investments = 3,670. It ties.

Step 3 gives the two ROICs and the two EVAs, against the ~9% US hurdle:

The franchise is elite at 25.7%, throwing off ~$312m of economic profit on the capital actually working in the labs. But management paid $1,800m of premium to assemble it. Once that real cash sits in the denominator, the full-deployment return is 13.1% and the economic profit roughly halves to ~$150m. Both verdicts are true, and they answer different questions (section 4.9). As a business, it's superb. As an acquisition record, it's merely good. It added value, since with-goodwill EVA is still positive and the roll-up beat its cost of capital. But it was expensive, with roughly half the franchise's economic profit paid away as premium. Compare the failure mode in Practice P15. There the premium is so large that with-goodwill EVA turns negative, and a great business was assembled in a way that destroyed value.

Step 4 is the family, in one line. ROE = 423 ÷ 2,410 = 17.6%, above the 13.1% with-goodwill ROIC. The $1.46bn of debt costs 4.5% after tax while operations earn far more, so leverage lifts the owners' return (the WE2 identity, directionally). ROA = 423 ÷ 4,470 = 9.5%. It's structurally mismatched (equity profit over everyone's assets). Report it and set it aside. That's four numbers for one business, with every wedge explained. It's section 4.6's claimholder check again, now with the M&A premium as the star witness.

Practice set

Work each problem on paper before you open its solution. P1–P4 are guided and mirror the section noted. P5–P8 are independent. P9–P10 are timed blank-sheet builds at 25 minutes each, a rehearsal for the gate, and P11 is a 5-minute judgment set. All companies are synthetic unless noted, and t = 25% throughout unless stated.

P1 (guided; section 4.3). NOPAT surgery. Deccan Agro Ltd (₹ crore): revenue 2,000; reported operating profit 320, which includes other income of 50 (interest on fixed deposits 40 + export incentives on operating sales 10) and is stated after an exceptional flood-damage loss of 30. The tax note shows a cleaned ETR of 25%. Compute clean NOPAT, and the number a hasty analyst gets from the reported line.

Solution. Strip the non-operating slice of other income but keep the export incentives, which arise from operating sales and recur (section 4.3's Schedule III warning). Add back the one-off flood loss, since one-offs are stripped in both directions. Clean EBIT = 320 − 40 + 30 = 310. NOPAT = 310 × 0.75 = ₹232.5 cr. The hasty answer is 320 × 0.75 = 240. That's only ~3% off here, but only because the two errors partly cancelled (−40 hidden income, +30 hidden loss). Luck won't repeat. The process will.

P2 (guided; section 4.2). The bucket sort. Assign each of these fifteen lines exactly one home: operating asset (OA), non-operating asset (NOA), operating liability (OL), or financing (F). Here are the lines: trade receivables, liquid mutual funds, capital work-in-progress, deferred revenue (customer advances), short-term borrowings, trade payables, right-of-use assets, current lease liabilities, investment property, minority interest, net deferred tax liability, accrued wages, goodwill, capital advances, and cash in the tills of a retail chain.

Solution.

People miss two of these most often. Short-term borrowings and current lease liabilities are never operating liabilities. Maturity doesn't change who supplied the capital.

P3 (guided; sections 4.4–4.5). Small build, both ways, tie-out. Sundari Textiles (₹ crore). Assets: cash 100, liquid mutual funds 80, receivables 150, inventory 200, prepaid expenses 30, net PP&E 500, ROU assets 60, goodwill 40 (total 1,160). Liabilities & equity: trade payables 170, accrued expenses 50, short-term borrowings 90, current lease liabilities 10, long-term borrowings 250, long-term lease liabilities 50, net DTL 30, equity 500, minority interest 10 (total 1,160). Clean EBIT is 152. Compute invested capital both ways, prove the tie, and compute ROIC (ending capital, with goodwill).

Solution. Operating side first. NWC = (150 + 200 + 30) − (170 + 50) = 380 − 220 = 160. Fixed & other operating = 500 + 60 = 560. IC ex-goodwill = 720, and with goodwill it's 760. On the financing side, debt (90 + 250) + leases (10 + 50) + equity 500 + MI 10 + DTL 30 = 940. Less non-operating (cash 100 + funds 80), that's 940 − 180 = 760, so it ties. NOPAT = 152 × 0.75 = 114. ROIC = 114 ÷ 760 = 15.0% (ex-goodwill 114 ÷ 720 = 15.8%). Against the 12% hurdle that's a modest positive spread. It's a decent business, and not yet a moat claim.

P4 (guided; section 4.6). The ratio family and its wedges. Sundari (P3) also reports: other income 6 (on the mutual funds), finance costs 30, PBT 128, tax 32, PAT 96, of which minority share 1. Compute ROE, ROA, convention-ROCE (screener style: (PBT + finance costs) ÷ (total assets − current liabilities)), and clean pre-tax ROIC. Explain each wedge in one line.

Solution. ROE = 95 ÷ 500 = 19.0%. ROA = 96 ÷ 1,160 = 8.3%. Current liabilities = 170 + 50 + 90 + 10 = 320, so CE = 1,160 − 320 = 840, and ROCE = (128 + 30) ÷ 840 = 158 ÷ 840 = 18.8%. Clean pre-tax ROIC = 152 ÷ 760 = 20.0%.

Now the wedges. ROE (19.0%) > post-tax ROIC (15.0%) because debt and leases cost ~5.6% after tax (30 × 0.75 ÷ 400). That's below the 15% ROIC, so leverage lifts the equity return (WE2's identity). ROCE (18.8%) < clean pre-tax ROIC (20.0%) because CE carries the idle 180 of cash/funds and excludes the 100 of short-term debt/leases, a net +80 of extra denominator. Its numerator also holds the 6 of other income. ROA is structurally mismatched (equity profit ÷ everyone's assets). Note it and set it aside. Four numbers, one business, every gap explained. That's the skill section 4.6 teaches.

P5 (independent; section 4.7). The rent-hidden capital. Coastline Coffee Inc. (US GAAP; $m) reports EBIT of 240, stated after operating-lease rent of 180. Invested capital ignoring leases is 1,200. The operating lease liability, printed on the balance sheet, is 1,500 at 6% implicit, and t = 25%. Compute naïve and lease-adjusted ROIC, and state what an IFRS 16 twin would have reported as EBIT.

Solution. Naïve first. NOPAT = 240 × 0.75 = 180, and ROIC = 180 ÷ 1,200 = 15.0%. Now adjust. Imputed interest = 6% × 1,500 = 90, so adjusted EBIT = 240 + 90 = 330. You'd get the same from EBIT + rent − implied ROU depreciation = 240 + 180 − 90, since in steady state rent ≈ depreciation + interest. Adjusted NOPAT = 330 × 0.75 = 247.5, and adjusted IC = 1,200 + 1,500 = 2,700. Adjusted ROIC = 247.5 ÷ 2,700 = 9.2%. Against an 8–9% US hurdle the "15% business" is running at the line. An IFRS 16/Ind AS 116 twin would already report EBIT ≈ 330 with the liability on its sheet. Put raw 240 next to 330 across the border and you've walked into the section 4.7 trap with money attached.

P6 (independent; section 4.9). Goodwill: the double view and the impairment flattery. Meridian Labs ($m) has NOPAT of 240, IC ex-goodwill of 800 and goodwill of 1,200 from a decade of acquisitions. The US hurdle is 9%. (a) Compute both ROIC views and interpret. (b) Next year Meridian impairs 600 of goodwill, and NOPAT stays 240. What does with-goodwill ROIC become, and what should you say about it?

Solution. (a) Ex-goodwill, it's 240 ÷ 800 = 30.0%, an excellent franchise. With goodwill, it's 240 ÷ 2,000 = 12.0%, three points over the hurdle. Read it per section 4.9. The underlying business is superb, and management paid away most of the excess return in acquisition premia. It's a fine business that management keeps buying at a high price. (b) New with-goodwill ROIC = 240 ÷ 1,400 = 17.1%, "improved" by destroying $600m. What you say is that the gross-goodwill record is still 240 ÷ 2,000 = 12.0%. An impairment is a confession. Anyone who quotes 17.1% as capital-allocation progress has been played by the denominator.

P7 (independent; section 4.8, the promised proof). Steady-state R&D barely changes the verdict. Silverbirch Software ($m): reported NOPAT 200 on IC 1,000 (ROIC 20.0%). R&D has been flat at 40/year for many years. Use a 5-year life. (a) Build the research asset and this year's amortization, then compute adjusted NOPAT, IC and ROIC. (b) State the general contrast with a growing-R&D firm.

Solution. (a) Research asset = 40 × (1 + 4/5 + 3/5 + 2/5 + 1/5) = 40 × 3 = 120. Amortization = five prior years ÷ 5 = 200 ÷ 5 = 40. That equals current spend, because spend is flat. Adjusted NOPAT = 200 + (40 − 40) = 200, unchanged. Adjusted IC = 1,120, and adjusted ROIC = 200 ÷ 1,120 = 17.9%. All you get is a ~2-point drag from the denominator. Your numerator doesn't move, the 8–9-point spread verdict doesn't move, and two steady-state peers keep their ranking. So section 4.8 tells you to adjust when comparing across R&D intensities and to skip it when comparing like with like. (b) With growing R&D, current spend exceeds the amortization of smaller past spends, so NOPAT rises. But the research asset grows the denominator proportionally faster, and measured ROIC falls, as Microsoft's did from 32.8% to 28.9% in section 4.8. Headline ROIC hides the growth in R&D. R&D on its own barely registers.

P8 (independent; section 4.10). The compounder table. Tulsi Consumer (₹ crore; t already applied, so figures are NOPAT): NOPAT by year FY20–FY25: 400, 450, 500, 560, 630, 700. Year-end invested capital: 2,000, 2,300, 2,550, 2,800, 3,000, 3,200. Compute 5-year ROIIC, the reinvestment rate, the g = RR × ROIIC check against actual NOPAT CAGR, the lagged-window variant, and the verdict.

Solution. ΔNOPAT = 700 − 400 = 300 and ΔIC = 3,200 − 2,000 = 1,200, so ROIIC = 300 ÷ 1,200 = 25.0%. Cumulative NOPAT FY21–25 = 450 + 500 + 560 + 630 + 700 = 2,840, so RR = 1,200 ÷ 2,840 = 42.3%. Check it. g = 0.423 × 25.0% ≈ 10.6% against an actual CAGR of (700/400)^(1/5) − 1 = 11.8%. That's close. The residual is margin drift, and the identity is a consistency check and not a law of single years. Lagged, it's ΔNOPAT from FY21 to FY25 = 250 ÷ ΔIC from FY20 to FY24 = 1,000 = 25.0%, unchanged. So what's the verdict? Tulsi reinvests only ~42% of profits at ~25%, double the 12% hurdle. It grows ~11–12% and can distribute the other ~58%. That makes it a compounder and a cash returner. Only runway is left open, and that's for Phase 4.

P9 (timed, 25 minutes; blank-sheet build 1, India). Rajdhani Paints Ltd (synthetic; ₹ crore; FY26 consolidated). P&L: revenue 5,000 and total expenses 4,220, which include an exceptional VRS charge of 40. Other income is 70 (interest and gains on the investment book 60, scrap sales 10). Finance costs are 60, mostly lease interest under Ind AS 116. PBT is 790 and the cleaned ETR 25%. Balance sheet, assets: cash 150, liquid investments 450, receivables 550, inventory 700, other operating current assets 100, net PP&E 1,600, CWIP 200, ROU assets 250, operating intangibles 50, goodwill 100 (total 4,150). Liabilities & equity: trade payables 650, accrued expenses & provisions 200, short-term borrowings 100, current lease liabilities 30, long-term borrowings 370, long-term lease liabilities 220, net DTL 80, equity 2,450, minority interest 50 (total 4,150). Opening invested capital (computed the same way last year): 2,500. Deliver clean NOPAT, IC both ways with tie-out, ROIC on average capital (with goodwill), convention-ROCE, and a verdict vs the Indian hurdle.

Solution. Start with NOPAT. Operating profit before finance costs = PBT + finance costs = 790 + 60 = 850. Take out the non-operating slice of other income, keep the 10 of scrap sales, and add back the VRS one-off. Clean EBIT = 850 − 60 + 40 = 830. NOPAT = 830 × 0.75 = 622.5.

Operating approach: NWC = (550 + 700 + 100) − (650 + 200) = 1,350 − 850 = 500. Fixed & other = 1,600 + 200 + 250 + 50 = 2,100. That's 2,600 ex-goodwill and 2,700 with goodwill. Financing approach: debt (100 + 370) + leases (30 + 220) + equity 2,450 + MI 50 + DTL 80 = 3,300. Less non-operating (cash 150 + investments 450), it's 3,300 − 600 = 2,700, so it ties. Average IC = (2,500 + 2,700)/2 = 2,600, so ROIC = 622.5 ÷ 2,600 = 23.9%.

Convention-ROCE: current liabilities = 650 + 200 + 100 + 30 = 980, so CE = 4,150 − 980 = 3,170, and ROCE = (790 + 60) ÷ 3,170 = 26.8%. That's below the clean pre-tax ROIC of 830 ÷ 2,700 ≈ 30.7% (ending), mostly because the ₹600 cr treasury book bloats CE. Your verdict is ~24% post-tax against 12%, a +12-point spread, which is elite. Your next questions are persistence and reinvestment runway. Base rates say fade, so name the moat. Score yourself on the nine-move ritual of section 4.13. And the tie-out was non-negotiable.

P10 (timed, 25 minutes; blank-sheet build 2, US). Frontier Fitness Holdings Inc. (synthetic; $m; US GAAP; ASC 842). P&L: revenue 3,000 and reported operating income 330. That's stated after operating-lease rent of 200 and includes a one-time restructuring charge of 30. Interest income and expense sit below operating income, and t = 25%. Operating lease liabilities total 1,600 at ~5% implicit. Balance sheet, assets: cash 250, short-term investments 150, receivables 180, inventory 220, other operating current assets 50, net PP&E 900, ROU assets 1,550, goodwill 400, operating intangibles 100 (total 3,800). Liabilities & equity: trade payables 260, accrued expenses 140, deferred revenue (memberships) 200, short-term debt 40, current operating lease liabilities 90, long-term debt 560, long-term operating lease liabilities 1,510, net DTL 20, equity 980 (total 3,800). Deliver lease-adjusted clean NOPAT, IC both ways with tie-out, ROIC with and without goodwill, and the two classic wrong answers with their causes.

Solution. NOPAT. Clean reported EBIT = 330 + 30 (one-off restructuring) = 360. Then the ASC 842 numerator fix. Imputed lease interest = 5% × 1,600 = 80, so adjusted EBIT = 360 + 80 = 440, and NOPAT = 440 × 0.75 = 330.

Operating approach. NWC = (180 + 220 + 50) − (260 + 140 + 200) = 450 − 600 = −150. Members prepay, which is lovely to see (section 4.4's negative-working-capital pattern). Fixed & other = 900 + 1,550 + 100 = 2,550. Ex-goodwill = 2,550 − 150 = 2,400, and with goodwill it's 2,800. Financing approach. Debt (40 + 560) + operating lease liabilities (90 + 1,510) + equity 980 + DTL 20 = 3,200. Less cash 250 and investments 150, it's 3,200 − 400 = 2,800, so it ties.

ROIC. With goodwill = 330 ÷ 2,800 = 11.8%, and ex-goodwill = 330 ÷ 2,400 = 13.75%. Against an 8–9% hurdle that's a real but moderate spread. This is a solid operator and no franchise miracle.

The two wrong answers. (1) Forget the numerator fix while the ROU assets and lease liabilities sit in IC, and you get 360 × 0.75 = 270 ÷ 2,800 = 9.6%. The business is punished twice and the spread nearly vanishes, a matching-rule violation. (2) Strip leases from both sides the old way and you get ≈ 270 ÷ ~1,200 = ~22%, a fantasy fitness empire renting its capital. The two sides then disagree by the ROU–liability gap of 50, so the build won't even tie, and the reconciliation catches the mistake. In truth it's ~12%.

P11 (timed, 5 minutes; three judgment calls; sections 4.6, 4.11). One line each. (a) A screener shows an enterprise-software firm at 38% ROIC. Your default 10-year forecast, and the two kinds of evidence that would let you override it? (b) A steelmaker shows 6% ROIC at the bottom of the cycle. Write it off? (c) A tool prints "ROIC: 18%" for a bank. Use it?

Solution. (a) The default is fade toward the cost of capital, because competition works. Override it only with a named moat mechanism plus a reinvestment runway. Switching costs or network effects would count, with market-share stability as the numbers-test. Also, measured software ROIC is inflated by unbooked R&D capital (section 4.8), so adjust before you believe the level. (b) No. Cyclicals are judged on mid-cycle returns across a full cycle. 6% at the trough may be a 12% mid-cycle business, and the capital cycle (falling industry capex) may be setting up better returns. (c) No. Deposits are a bank's raw material, so separating operating from financing means nothing. Judge ROE vs cost of equity and ROA (>1% good, >1.5% excellent).

P12 (guided; sections 4.1, 4.11). Economic profit, in rupees. Vidarbha Cements Ltd (synthetic; ₹ crore) earns clean NOPAT of 360 on average invested capital of 2,000, against a 12% hurdle. (a) ROIC and economic profit today. (b) It reinvests a further ₹500 crore at a 15% incremental ROIC. Compute the firm's new total economic profit. (c) Capitalizing at the hurdle, how much value did that ₹500 crore create?

Solution. (a) ROIC = 360 ÷ 2,000 = 18%. Economic profit = NOPAT − hurdle × IC = 360 − 12% × 2,000 = 360 − 240 = +₹120 cr/yr, which you could also write as (18 − 12)% × 2,000. (b) The ₹500 cr adds 15% × 500 = ₹75 cr of NOPAT, and the capital charge takes 12% × 500 = ₹60 cr of it. Added economic profit = +₹15 cr/yr, so the new total = 120 + 15 = +₹135 cr/yr. (c) Capitalize the added economic profit at the hurdle: 15 ÷ 0.12 = +₹125 cr created. Check it another way. The ₹75 cr/yr stream is worth 75 ÷ 0.12 = ₹625 cr for ₹500 cr spent, which is +₹125 cr. A positive spread makes reinvestment a value engine. It's the mirror image of Worked Example 3's empire builder.

P13 (guided; section 4.5, method). Ending or average? Brightline Logistics Inc. (synthetic; $m) held invested capital of 1,500 at the start of the year and 2,100 at the end, all organic growth, and earned NOPAT of 240. The US hurdle is 9%. (a) ROIC on ending capital. (b) ROIC on average capital. (c) Which is the default, and why does the other mislead here? (d) Economic profit on average capital.

Solution. (a) 240 ÷ 2,100 = 11.4%. (b) Average IC = (1,500 + 2,100)/2 = 1,800, so 240 ÷ 1,800 = 13.3%. (c) Average is the default. The year's profit was earned on capital that climbed from 1,500 to 2,100 through the year. Year-end capital charges the business for money it didn't have for most of the period, a ~2-point understatement in a growth year. (d) EVA = 240 − 9% × 1,800 = 240 − 162 = +$78m. (One extension. Had the $600m increase been a November acquisition and not organic growth, even the average would overstate the deal's mature contribution. Time-weight the new capital or go pro-forma, as mistake #4 says.)

P14 (independent; section 4.6). The leverage identity, and what ROE alone hides. Nilgiri Steels (synthetic) earns a 14% ROIC; it borrows at 10% pre-tax (t = 25%) with a debt-to-equity ratio of 1.5. Its all-else-identical twin earns only a 9% ROIC on the same financing. Hurdle 12%. (a) Each firm's ROE via the identity ROE = ROIC + (ROIC − rd(1−t)) × D/E. (b) Which creates value? (c) What would ROE alone have told you?

Solution. After-tax cost of debt = 10% × 0.75 = 7.5%. (a) For the first firm, ROE = 14% + (14% − 7.5%) × 1.5 = 14% + 6.5% × 1.5 = 23.75%. For the twin, ROE = 9% + (9% − 7.5%) × 1.5 = 9% + 1.5% × 1.5 = 11.25%. (b) Only the first creates value (14% ROIC > 12% hurdle). Its twin destroys it (9% < 12%). (c) ROE flatters both. The first firm's 23.75% is 14% of operations plus 9.75 points of pure leverage. And the twin's 11.25% still exceeds its own 9% ROIC, because leverage manufactures ROE whenever ROIC beats the after-tax cost of debt, even for a value-destroying business. Isolate the machine with ROIC before you let ROE speak. This is Worked Example 2 as a drill.

P15 (independent; section 4.9). Goodwill: when the premium destroys value. Summit Instruments Inc. (synthetic; $m) has NOPAT of 120, invested capital ex-goodwill of 500 and goodwill of 1,200 from a decade of acquisitions. The US hurdle is 9%. (a) Both ROIC views and both economic profits. (b) The one-sentence verdict on business vs management.

Solution. (a) Ex-goodwill, ROIC = 120 ÷ 500 = 24.0% and EVA = 120 − 9% × 500 = +$75m. With goodwill, ROIC = 120 ÷ 1,700 = 7.1% and EVA = 120 − 9% × 1,700 = −$33m. (b) It's an excellent franchise (24%, +$75m of economic profit) that management assembled in a way that destroyed value. The $1,200m of premium dragged the full-cost record below the hurdle, vaporizing ~$33m/yr. Judged on what it actually paid, this serial acquirer earns less than its cost of capital. Compare Worked Example 5, where the premium was expensive but still added value. Here it crossed the line.

P16 (independent; section 4.10). The tollbooth: high returns, nowhere to grow. Nagpur Toll Roads Ltd (synthetic; ₹ crore; figures are NOPAT and year-end invested capital), with a 12% hurdle.

Compute the 5-year ROIIC, the reinvestment rate, the g = RR × ROIIC check against actual NOPAT CAGR, and the verdict.

Solution. ΔNOPAT = 365 − 300 = 65 and ΔIC = 2,060 − 1,800 = 260, so ROIIC = 65 ÷ 260 = 25.0%, double the hurdle. Cumulative NOPAT FY21–25 = 315 + 325 + 340 + 350 + 365 = 1,695, so RR = 260 ÷ 1,695 = 15.3%. Check it. g = 0.153 × 25.0% ≈ 3.8%, against an actual CAGR of (365/300)^(1/5) − 1 = 4.0%. Close enough. The returns are superb, with headline ROIC ~18% and incremental ~25%. But there's almost no reinvestment runway. It can put back only ~15% of profits, so it grows just ~4%. This is a tollbooth (section 4.10). Its value is in returning the other ~85% of profits, and there's no "growth" to celebrate. Compare Practice P8 (Tulsi Consumer). Its returns are of the same quality, but Tulsi has a real runway, and that's the difference between a cash cow and a compounder.

P17 (timed, 10 minutes; six rapid calculations; sections 4.1, 4.6, 4.9–4.11). One number each. (i) NOPAT ₹210 cr, average invested capital ₹1,500 cr, hurdle 12% → economic profit. (ii) Opening IC $800m, closing IC $1,200m, NOPAT $150m → ROIC on average capital. (iii) ROIC 16%, after-tax cost of debt 6%, D/E = 2 → ROE via the identity. (iv) NOPAT $90m, invested capital with goodwill $1,000m, hurdle 9% → economic profit, and what it says about the M&A. (v) ΔNOPAT ₹140 cr over ΔIC ₹560 cr (5-year window) → incremental ROIC. (vi) Reinvestment rate 80%, incremental ROIC 15% → sustainable NOPAT growth.

Solution. (i) 210 − 12% × 1,500 = +₹30 cr. (ii) Average IC = 1,000, so 150 ÷ 1,000 = 15%. (iii) 16% + (16% − 6%) × 2 = 36%. (iv) 90 − 9% × 1,000 = $0. The acquisitions earned the cost of capital and nothing more. That's treadmill M&A, a bigger company with the same owner value. (v) 140 ÷ 560 = 25%. (vi) g = RR × ROIIC = 0.80 × 15% = 12%. Scoring: 6/6 in time means you're fluent. If you got ≤4, redo sections 4.10–4.11 and Worked Examples 4–5 before the mastery check.

Applied mini-project

You'll take two returns machines and give each a fair verdict. The first is ITC Ltd. Use the latest consolidated annual report from the company's investor-relations page or BSE, plus the ten-year history on screener.in. The second is The Home Depot, Inc. Use the latest 10-K on SEC EDGAR (sec.gov full-text search), plus history from a free aggregator (stockanalysis.com or macrotrends.net). We paired them on purpose. ITC is an Indian giant with a famous cash-and-investments book, so convention-ROCE and clean ROIC come apart. Home Depot is a US retailer with heavy leases and negative shareholders' equity after two decades of buybacks. There ROE stops computing while ROIC keeps working. Allow 3–4 hours.

1. NOPAT, both companies. Clean EBIT first. Strip ITC's treasury/other income and any exceptional items, using the notes on both. For Home Depot, scan for one-offs inside operating expenses and add the imputed interest on operating leases. Under ASC 842 the liability and a disclosed discount rate are in the lease note. State and defend your tax rate for each.
2. Invested capital both ways, tie-out to the last crore/million. Decide and state homes for ITC's investment book and any hotel-asset nuances. Include both finance and operating leases for Home Depot. Then write one line on why the financing approach still ties despite negative equity (debt + leases − cash absorb it).
3. The ratio family. Compute ROIC with and without goodwill. Compute convention-ROCE for ITC, then reconcile your number to screener.in's printed ROCE in two lines. Compute ROE, and state what happens for Home Depot and why the number means nothing. Then ROA. Write one line each on what it answers here.
4. Incremental returns. The 5-year ROIIC table for both (ΔNOPAT ÷ ΔIC), reinvestment rate, the g = RR × ROIIC check, and the lagged variant. Flag any year where a one-off distorts the window.
5. Spread and persistence. Compare like-for-like against ~12% (India) / ~8–9% (US). Write the fade-or-defend call with named evidence (share stability, pricing power, reinvestment runway). Base rate first, override second.
6. Memo. Write one page per company, conclusion first: "Is this a good business, and is it getting better or worse at turning new capital into profit?" Every claim must carry a number, and every number must carry a source (filing page or note).

Scoring rubric (self-grade after writing; pass ≥ 13/16):

File both companies in your company files. Both ROIIC tables become permanent rows in your deep-dive workbook.

Case Lab: run case CL-2.3 in the app's Cases tab for this module, a guided returns-on-capital build-off that scores your NOPAT hygiene, the both-ways tie-out, the goodwill double view, average-vs-ending choice, and the ROIIC verdict against a hidden key.

Reading & resources

• McKinsey (Koller, Goedhart & Wessels), Valuation (8th ed., 2025): the chapters "Return on Invested Capital" and "Analyzing Performance" (chapter numbers vary by edition; both sit in the core-valuation part). The NOPLAT/invested-capital machinery simplified here, in full professional form. [Paid] [Intermediate→Advanced]
• Aswath Damodaran, "Return on Capital (ROC), Return on Invested Capital (ROIC) and Return on Equity (ROE): Measurement and Implications" (2007). Free PDF on his NYU site (pages.stern.nyu.edu). It's the best paper-length treatment of every measurement choice you made in this build. [Free] [Intermediate]
• Damodaran's R&D converter: the R&Dconv spreadsheet plus his paper "Research and Development Expenses: Implications for Profitability Measurement and Valuation" (both on his site, spreadsheets page). It gives suggested amortizable lives by industry, and it's section 4.8 in tool form. [Free] [Intermediate]
• Mauboussin & Callahan, "Calculating Return on Invested Capital" (Credit Suisse, June 2014). A free PDF, and the classic build-it-by-hand guide with the goodwill double view. [Free] [Intermediate]
• Mauboussin & Callahan, "Return on Invested Capital: How to Calculate ROIC and Handle Common Issues" (Morgan Stanley Counterpoint Global, October 2022): the updated version, post-ASC 842. Read the lease and intangibles sections against sections 4.7–4.8. [Free] [Intermediate]
• Mauboussin & Callahan, "The Base Rate Book" (Credit Suisse, 2016): the ROIC chapter, its transition matrices and its persistence quintiles, which is section 4.11's evidence in full exhibits. [Free] [Advanced]
• Mauboussin, "Measuring the Moat" (Credit Suisse, 2016 update): persistent spreads, CAP and mean reversion. You'll reread it in M4.05, so a first pass now pays off later. [Free] [Advanced]
• Mauboussin & Callahan, "Intangibles and Earnings" (Counterpoint Global, 2022): why measured ROIC systematically overstates asset-light champions. It's section 4.8's closing caveat at full length. [Free] [Advanced]
• Joel Greenblatt, The Little Book That Still Beats the Market, chapters 5–7: his return-on-capital definition (EBIT ÷ (NWC + net fixed assets)) is a stripped-down cousin of yours. Spot which of your nine moves he skips, and why it mostly works anyway. [Paid] [Beginner]
• William Thorndike, The Outsiders. Read now for pleasure, reread in M3.07: eight CEOs judged almost entirely on the spread and where they pointed it. [Paid] [Beginner]
• Zerodha Varsity, Fundamental Analysis module: the financial-ratios chapters (RoE/RoCE) for the Indian-convention vocabulary. [Free] [Beginner]
• screener.in. Read its ratio definitions (each ratio's formula is documented), then add a custom column for ROCE history on your watchlist. Your P4/P9 reconciliation skill is what makes the tool safe to use. [Free] [Beginner]
• Filings & data to pull (all free): Asian Paints consolidated annual report (company site/BSE) for section 4.12; Microsoft 10-K (EDGAR) for Worked Example 1; ITC annual report + Home Depot 10-K for the mini-project. The InterGlobe Aviation annual report has the lease note that dwarfs borrowings (section 4.7), and you'll want Avenue Supermarts annual reports + screener history for section 4.10's table. [Free] [All levels]
• Going deeper: G. Bennett Stewart III, The Quest for Value (HarperBusiness, 1991). The founding text of EVA / economic profit as a full management-and-valuation system. Read it as the long-form of Worked Example 4's NOPAT − hurdle × capital line, well before M3.07 formalizes it. [Paid] [Advanced]
• Going deeper: Aswath Damodaran, "Economic Value Added (EVA)" session notes and the EVA.xls spreadsheet (pages.stern.nyu.edu → Data/Spreadsheets): excess-return models built directly on ROIC, invested capital and the cost of capital. It's a free bridge from the economic-profit rupees you computed here to Phase 3's valuation. [Free] [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": "Clean NOPAT, invested capital by both approaches with a tie-out to the rupee, the four-ratio family, the three adjustments, and incremental ROIC against the hurdle",
  "use": "A ROIC build is a sorting problem — every balance-sheet line gets exactly one home, operating or financing — and sorting is the one part of this module a tool can genuinely accelerate. (1) Paste a balance sheet and ask for a two-column classification with a stated reason per line and an explicit flag on every judgment call: surplus versus operating cash, investments that are treasury versus strategic, deferred tax, capital work in progress. Then overrule it line by line, because the judgment calls are the module. (2) Ask it to list every adjustment the filing makes necessary that you have not made — leases where the standard differs, R&D, goodwill from a specific acquisition — and what each would do to the denominator's direction. (3) For the ROIIC table, ask it to extract the multi-year EBIT, tax-rate and balance-sheet series across ten years and flag every year with an acquisition, a demerger, a restatement or a standard change, since each of those breaks a ΔIC and the break is invisible in the arithmetic. (4) Ask it to state the strongest case that a high ROIC will fade, then go looking for the barrier that would refute it — §4.11 makes fade the default forecast and the burden of proof yours.",
  "tools": ["Chat assistants — Claude, ChatGPT for the operating-versus-financing classification pass and the fade argument", "Document/PDF AI — for the investments, leases, deferred-tax and goodwill notes that decide the judgment calls", "Coding assistants — for the ROIIC window panel and the noise simulation below"],
  "prompt": "Here is a balance sheet and the relevant notes: <PASTE>. Produce a table [Line item | Amount | Operating or Financing or Non-operating asset | One-line reason | Judgment call? Y/N]. Every line must appear exactly once and the two columns must be reconcilable. Do NOT compute NOPAT, invested capital or any ratio — I am doing that. Separately, list every line where a defensible analyst could classify it the other way, and what that alternative does to invested capital.",
  "verify": "The tie-out is the verification and it is not negotiable: the operating approach and the financing approach must reconcile to the rupee, exactly as §4.12 step 3 does, and no ratio is computed before they do. That single check catches almost every classification error a tool or a human can make, because a misplaced line lands in one column and not the other. Then verify the tax rate against the ETR reconciliation note rather than accepting a statutory rate, verify that other income and one-offs have been stripped from EBIT symmetrically, and verify every ΔIC year in the ROIIC table against the filings for an acquisition or restatement. A model asked to classify a balance sheet will classify all of it, confidently, including the lines that required you to decide (Guardrail 1-5).",
  "diy": "The gate is unaided: you build clean NOPAT and invested capital both ways from a real filing, reconcile them to the rupee, compute the four ratios, and explain every wedge between them on paper. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You built Asian Paints both ways and tied it out to the rupee. That build, with its judgment calls logged, is the core skill. What code adds is the ten-year, many-window view of incremental returns. It's how the compounder question really gets settled, and you can't do it by hand.

Today's workflow. Once NOPAT and invested capital are functions of a filing, and no longer a page of arithmetic, three things follow. First, the four ratios become a panel across ten years and a peer set. You learned to explain the wedge between convention-ROCE, clean pre-tax ROIC, post-tax ROIC and ROE in section 4.12 step 4, and now you can explain it for every company at once. Second, the three adjustments become switches. Capitalised R&D goes on or off, goodwill in or out, leases restated, and each choice produces its own ROIC series. That's the fair way to present a number that depends on choices. Third, incremental ROIC becomes a distribution over windows instead of a single figure. Nothing else you've learned here changes the conclusion more. On the DMart table you worked through earlier, the five-year ROIIC is 17.6% and the lagged version is 24.3%. Taken together, the one-year windows span −15.0% to +37.5%. Reporting one of those as the incremental return is a choice you should make deliberately, and the panel is what makes the choice visible.

Tools & sources (IN + US). Use pandas for the panel and the window sweep, and numpy for the simulation. For India, use the consolidated annual report for the balance-sheet sort and the investments, leases and deferred-tax notes that decide the judgment calls. Use screener.in for the ten-year EBIT and capital series, which you then reconcile against two or three filings by hand (BSE/NSE for the filings themselves). For the US, use EDGAR XBRL companyfacts (OperatingIncomeLoss, Assets, Liabilities, CashAndCashEquivalentsAtCarryingValue, ShortTermInvestments, Goodwill, OperatingLeaseLiability, ResearchAndDevelopmentExpense), the 10-K notes for the same judgment calls, Damodaran's ROIC- and cost-of-capital-by-sector tables at pages.stern.nyu.edu for the hurdle and the sector shape, and the Mauboussin/Counterpoint Global ROIC-persistence papers for the base rates behind section 4.11.

# DMart illustration, the module's own table, ₹ cr — R3: recomputed in-session
import numpy as np
nopat = np.array([1500, 1275, 1800, 2700, 3075, 3450], float)   # FY20..FY25
ic    = np.array([9500, 11000, 12800, 15200, 17800, 20600], float)
w = lambda i, j: (nopat[j] - nopat[i]) / (ic[j] - ic[i])         # ROIIC over one window
one = [w(i, i + 1) for i in range(5)]                            # every 1-year window
print(round(min(one) * 100, 1), round(max(one) * 100, 1))        # -15.0  37.5
print(round(w(0, 5) * 100, 1), round((nopat[5] - nopat[2]) / (ic[4] - ic[1]) * 100, 1))
#   17.6 (5-yr)   24.3 (lagged)  -> report the range, not a point
rr = (ic[5] - ic[0]) / nopat[1:].sum()                           # reinvestment rate 90.2%
print(round(rr * 100, 1), round(rr * w(0, 5) * 100, 1),
      round(((nopat[5] / nopat[0]) ** 0.2 - 1) * 100, 1))        # 90.2  15.9  vs 18.1 actual
print(round(100 * 0.08 / 0.12, 1))     # ₹66.7 of value per ₹100 reinvested at 8% vs a 12% hurdle

Verify. Nothing in the panel is usable until you've built one company in it by hand and tied it out both ways. Everything rests on a classification no data source makes for you. An aggregator's "capital employed" is somebody else's sort, usually total assets less current liabilities. Section 4.12 shows that convention giving 28.8% where the clean pre-tax figure is 31.8% and the post-tax figure 23.8%. Reconcile, then state which convention every published number uses. Check each ΔIC year for an acquisition, a demerger, a lease-standard adoption or a restatement. Any of those makes that year's incremental return arithmetic meaningless. And check the tax rate against the ETR note for every year. Once isn't enough.

Quantitative lens

Section 4.10 says never compute incremental ROIC on one year. That's correct, and you can turn the rule into a measurement. It's worth doing, because you'll actually obey a rule you've measured.

Take that same DMart table and compute ROIIC over every window it contains. The five one-year windows span −15.0% to +37.5%, a range of 52.5 percentage points. The four two-year windows span 9.1% to 33.9%, a range of 24.8 points. The three-year windows span 21.1% to 26.5%, a range of 5.4 points, and the four-year windows 3.7 points. The single five-year window gives 17.6%. The dispersion collapses by an order of magnitude between one year and three. And it does so on a company whose underlying incremental return never changed.

So this tells you about the estimator, and nothing about DMart.

Why does it happen? This mechanism generalises to every ratio of two differences you'll ever compute. ROIIC divides a change in NOPAT by a change in invested capital, and differencing amplifies noise. If each year's NOPAT carries a standard deviation σ, the difference carries σ√2. Meanwhile the denominator, ΔIC over a short window, is small. The signal shrinks with the window and the noise doesn't. Simulate it. Give each year's NOPAT a modest 10% lognormal error and compute ROIIC on the same table. Over one-year windows the interquartile range runs from −0.4% to +33.0%, and 25.4% of draws come out negative. That's on a business really earning about 18% on incremental capital. Over three-year windows, none do, and the interquartile range narrows to 19.0%–26.8%.

A one-year ROIIC isn't just a noisy version of the right answer. A quarter of the time it has the wrong sign.

This method has limits. Lengthening the window buys precision by assuming stability. If a company re-based its economics inside a five-year window, you're now measuring across two different businesses. So choose the window against the company's history, and not against a rule. The simulation's 10% noise is illustrative and independent across years. Real NOPAT errors are serially correlated, since a cyclical peak persists, and that makes the true dispersion of short-window estimates worse. And none of this touches the deeper problem section 4.11 raises. The arithmetic measures what incremental capital has earned. Every use of that number is a bet that it says something about what the next rupee will earn. Persistence base rates say you should place that bet cautiously, and with a named barrier.

Do it in code: sweep every window on a ten-year series and report the ROIIC range and the lagged variant alongside the headline. Then bootstrap the series to get a confidence band. Present the range in the memo, so the reader sees how much the window choice was doing. See G1/QM1.01 and G2/DA1.05.

Where this goes next: galaxy cross-links

• G2 · DA1.02 Pandas Data Wrangling: the balance-sheet sort and the ten-year, four-ratio panel as code, with the three adjustments as switches instead of re-typed tables.
• G2 · DA1.05 Statistics in Code: the window sweep, the bootstrap and the confidence band on an estimator built from differences. That's the fair way to publish a ROIIC.
• G1 · QM1.01 Probability, Distributions & Estimation: why differencing amplifies noise, and what sample size a ratio of two differences needs before it means anything.
• G3 · CN5.02 Market Attractiveness: the strategy side of section 4.11. The spread is an invitation to entry, and naming the barrier that protects it is a structured industry question more than a financial one.

Teach it back & journal

Feynman prompt. Explain to a smart 15-year-old why "how much profit does it make?" is the wrong first question about a business, and "how much money does it need to make that profit?" is the right one. Use two lemonade stands earning the same profit on very different setups, one tiny five-line ROIC build with invented numbers, and the phrase "every line gets exactly one home." End by explaining, in two sentences, how a company can grow its profit every single year while making its owners poorer. Keep it to one page, with no jargon you don't define.

Journal reflection. Take the company you most admired before this program for its profit growth. Pull five years of history (screener.in or the filings) and run the quick ROIIC table from section 4.10. Then write three sentences. Did the verdict survive the arithmetic? Which placement judgment (cash, deferred tax, goodwill) felt most arbitrary, and what convention will you commit to, so that every build you ever do is comparable with this one? Log the ROIIC number. It becomes a row in your deep-dive workbook.

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