Cash Flow Analysis & Free Cash Flow

Phase 2 · Week 21 · ~20 hours

Where it sits. You can already build a cash flow statement: CFO by the indirect method from the P&L and balance-sheet deltas, the classification rules, the cross-standard traps. Now you learn to interrogate one. You arrive with the ratio toolkit and the operating/financing separation, and here they meet on the question every analyst eventually learns to ask first. The P&L says the company earned this, so did the cash actually show up, and where did it go? The master map's five-questions spine runs straight through what follows. Question 2 (is it a good business?) is only half-answered by ROIC until cash conversion confirms the profits are real. Question 3 (do they allocate capital well?) gets its answer from the ten-year cash walk you'll build here, and MD&A adjectives don't count. Benchmarking comes next, then the forensic work, which turns every divergence signal you pick up here into a weapon.

Suggested hours: core content 8h, worked examples 2h, practice set 5h, mini-project 3h, flashcards and mastery check 2h.

In the app, the material here powers the Quality & Red Flags branch's cash-conversion drills ("compute the restated CFO," "which tier is driving this CFO growth?"), the "spot the divergence" red-flag cards (NI up, CFO down), and the case engine's cash-walk scoring. The teardown checklist's question 7 (cash conversion, CFO/EBITDA, FCF/PAT) is trained here.

Learning objectives

By the end you can:

1. Read any cash flow statement as an analyst in ten minutes: restate it to a single convention (interest, leases), bridge net income to CFO, and say which of the five quality tiers is actually generating the cash.
2. Compute and interpret the conversion ratios: CFO/NI and CFO/EBITDA over multi-year windows, with benchmark ranges, the anatomy of the gap, and the cross-standard restatements that make them comparable.
3. Build free cash flow properly: FCF = CFO − capex, with the four definitional decisions made explicit (which capex, disposals, lease principal, interest classification), and defend your definition.
4. Split capex into maintenance and growth three ways (the depreciation proxy, Greenwald's sales-linked method, and disclosure/unit methods), then triangulate to a defensible range.
5. Compute Buffett's owner earnings and reconcile it to FCF: state when the two diverge, in which direction, and what each is for.
6. Read capex/depreciation as a life-cycle and aggression signal, including the inflation adjustment and the capitalization red flag it hands to M2.06.
7. Handle stock-based compensation honestly: show with numbers why "FCF" that ignores SBC dilution overstates owner returns, and apply both correction methods.
8. Classify any company into one of the four cash-flow personalities (compounder, capex monster, working-capital sink, financial-engineering case), name the follow-up question each personality demands, and build a ten-year cash walk (where every rupee or dollar of CFO went) in under an hour from public filings.

Prerequisites & connections

Builds on. M1.04, where you learned to build a CFS and the classification map. US GAAP puts interest paid in CFO. Ind AS puts interest paid in CFF and interest received in CFI, and lease principal goes to CFF under IFRS 16/Ind AS 116. Here you attach money to those differences. Add M1.05 for the three-statement linkage (the NI-to-CFO bridge is your accruals X-ray), and M1.08 and M1.09 for leases and SBC mechanics: expense, add-back, dilution. Then comes the ratio layer. M2.01 covers margins, where EBITDA arrives as a numerator you'll learn to distrust. M2.02 covers working capital, with DSO/DIO/DPO and ΔWC ≈ WC% of sales × Δrevenue, the arithmetic of the working-capital sink. Last is M2.03, because negative FCF is judged by the ROIC and ROIIC of the capital it funds.

Feeds forward. M2.05 benchmarks these ratios across peers and over time. M2.06 and M2.07 turn every divergence flag raised here into a named shenanigan with a detection test. Think payables stretch, factoring, capitalization games, Sloan accruals. M3.04 rebuilds cash flow formally as FCFF/FCFE for valuation, and the definitions you pin down here are why you'll never mix a levered cash flow with an unlevered discount rate. M3.07 grades a decade of capital allocation, and the cash walk is its evidence table. M4.04 to M4.06 read durable FCF conversion as moat evidence. In M8.02 and M8.03, teardown question 7 (cash conversion) and the deep-dive workbook's ten-year CFO/capex/FCF rows are this same skill, run at speed, for good.

Profit is an opinion about the period. Cash is a fact about the bank account. Your job from here is to reconcile the two, decide how much of the cash will last, and subtract what the business must spend to stay alive. Then follow what remains until you can see the owners' money. Sometimes you'll find there isn't any.

Core teaching content

4.2 The analyst's first pass: ten minutes with a CFS

You built statements line by line when you first met the cash flow statement. Analysts read them differently: top-down, pattern first, five years side by side. Pull the history from screener.in for India, and from stockanalysis.com or the filings themselves for the US. Here's the ritual.

Step 1: the three subtotals, five years across. Line up CFO, CFI, CFF on three lines. A healthy mature company shows CFO strongly positive, CFI moderately negative as it reinvests, and CFF negative as it returns capital and repays debt. A growth company shows CFO positive and rising and CFI deeply negative. Its CFF is positive because it's raising capital to fund the build. Distress looks different again. CFO is weak or negative, CFI turns positive because the company is selling assets to survive, and CFF is positive until it suddenly goes negative when lenders bolt. That one glance at the sign pattern places the company on its life cycle before you've read a single line item.

Step 2: restate to one convention. Before any ratio, ask where interest paid goes. Ind AS puts it in CFF, mandatorily, for non-financial companies. US GAAP puts it in CFO. IFRS today gives a choice. As of mid-2026, IFRS 18 also pushes interest paid to financing for most non-financial companies from 2027, so verify the Ind AS mirroring when you get there. Then ask where lease principal goes. It's CFF under all three, but US GAAP keeps operating-lease payments fully inside CFO while IFRS 16/Ind AS 116 move the principal out. So the Indian retailer's CFO is structurally flattered against its US twin. Write the restated CFO down. Every cross-standard comparison from here on starts there. You built the machinery when you built the statement, and now you bill for it.

Step 3: the NI-to-CFO bridge as an accruals X-ray. The indirect-method reconciliation is a gift. It lists which accruals separate profit from cash. Read it as three blocks:

Net income
+ non-cash charges        (D&A, SBC, impairments, provisions)     → the accounting block
± working-capital deltas  (receivables, inventory, payables, ...) → the timing block
± everything else         (deferred taxes, one-off items)         → the block to read twice
= CFO

The accounting block should be boring and stable as a % of revenue. The timing block should roughly track revenue growth, because growing companies consume working capital at about WC% × Δrevenue, as the working-capital ratios showed. Now suppose CFO growth is coming from the timing block while net income stands still. Maybe payables are ballooning, or there's been a one-time inventory flush. The cash is real, but the engine isn't. That distinction gets its own treatment next, in the CFO quality hierarchy.

Step 4: follow the cash out. In CFI, how much went to PP&E, how much to intangibles, how much to M&A, and how much is parked in liquid investments? Is the company raising money in CFF, or returning it? Set dividends and buybacks against debt raised, and ask whether owners are being paid out of operations or out of new borrowings. It's the one-year miniature of the ten-year cash walk you'll build later.

4.5 Free cash flow, defined properly

Every measure so far counted cash generated. Free cash flow counts cash left over, meaning what stays after the business has paid for the capital assets it runs on:

FCF = CFO − capex

Simple, except each term hides a decision. Analysts who don't make those decisions out loud end up with numbers that silently disagree with everyone else's. Here are the four.

Decision 1: which capex? Take "purchases of property, plant and equipment" plus "purchases/additions of intangible assets", such as capitalized software, development costs, licenses. Beginners miss the intangible line. For a software or pharma company, capitalized development is capex in economic substance, and companies that capitalize aggressively are counting on you to subtract only the PP&E line. The forensic modules file this under improper capitalization. Default: subtract both. M&A isn't capex for FCF purposes. It's a capital-allocation choice, and you'll judge it separately in the cash walk.

Decision 2: gross or net of disposals? Netting asset-sale proceeds against capex is fine when disposals are routine fleet churn, like a logistics company selling three-year-old trucks. It's flattery when a struggling company is selling buildings to fund itself, and that's Tier-5 staging of "FCF". Default: use gross capex, and net only with a stated reason.

Decision 3: where is interest? This carries over from the restatement in your first pass. Under Ind AS, the CFO you're subtracting capex from has not yet paid the lenders. For any leveraged Indian company, compute the comparable figure: FCF = CFO − interest paid − capex. Skip this and a debt-heavy infrastructure firm shows glorious "free" cash flow that's already spoken for.

Decision 4: lease principal. Under IFRS 16/Ind AS 116, the principal part of the rent goes in CFF, while the US GAAP twin's operating-lease rent burdens CFO in full. For lease-heavy businesses like retail, aviation, QSR, hotels, subtract lease principal from the IFRS/Ind AS company's CFO before computing FCF. Otherwise its FCF is inflated by roughly the whole rent bill minus its interest sliver. The sliver isn't forgotten. Lease interest is interest paid, so an Ind AS reporter parks it in CFF, and Decision 3 has already taken it out. Together the two decisions remove the whole lease payment from an Ind AS retailer's CFO, which is what its US twin's CFO already carries. An IFRS reporter that keeps interest paid in CFO needs only the principal.

Put it together and you get the analyst's comparable FCF, which you'll use whenever standards are mixed:

FCF = CFO − interest paid (if classified outside CFO) − lease principal (if outside CFO) − capex incl. intangibles

One conceptual placement, so Phase 3 lands softly. Because CFO under US GAAP is after interest, the simple FCF you build here is an equity-flavored number. It's closest to what valuation will later formalize as FCFE before net borrowing. The valuation modules rebuild both formal variants, FCFF for the whole firm and FCFE for shareholders, with strict discount-rate matching. What you must carry from here is the reflex of stating your definition before you quote any FCF number. The same company's "FCF" can legitimately differ by 2× across definitions, and Practice Problem 4 makes you feel it.

Is negative FCF bad? Not necessarily, because FCF is a residual: whatever's left after capex. FCF = CFO − capex means a company can be FCF-negative for two opposite reasons. Its operations may generate nothing, which is bad. Or its operations generate plenty and management is reinvesting more than all of it, which is bad or brilliant. To break the tie, go back to returns on capital and ask what ROIIC the reinvested capital is earning. Avenue Supermarts ran thin-to-negative FCF for years while building stores that earned well above the hurdle: value creation in a negative-FCF costume. A struggling airline runs negative FCF into capacity nobody wants, and that's value destruction in the same costume. So never read FCF without ROIIC beside it.

A negative-FCF company with high incremental returns and sane funding is investing, as long as there's no dilution spiral and no leverage spiral behind it.

4.9 Capex/depreciation: the life-cycle dial and the aggression tripwire

One more ratio, cheap to compute and disproportionately informative:

Capex/D&A = capital expenditure ÷ depreciation & amortization     (3-yr rolling; exclude acquired-intangible amortization)

Read it as a life-cycle dial:

It also works as an aggression tripwire, in both directions. Capex far above D&A with little revenue growth to show for it is the classic signature of operating costs being capitalized. Expenses get parked on the balance sheet as "assets", so they hit neither EBITDA nor CFO and leave through CFI as capex. WorldCom capitalized $3.8bn of ordinary line costs this way. India's small-cap graveyard (Manpasand, Vakrangee vintage) is full of capex/D&A ratios north of 3 with stagnant asset turns. In the other direction, D&A drifting down as a % of gross block while capex holds means depreciation lives are being stretched, and an estimate change is inflating earnings. Either way, the ratio doesn't convict anyone. It tells you where to dig, and the forensic modules hand you the shovel.

4.11 The four cash-flow personalities

After enough statements, you stop reading line items and start recognizing shapes. Four recur so reliably that they deserve names and profiles, plus standing follow-up questions. All figures below are illustrative, modeled on the named companies' approximate reported scale. Pull the actual filings when you study them.

Personality 1: the compounder. Its signature is CFO/NI ≥ 1.0 every single year, cumulative CFO/EBITDA ≥ 75%, FCF/PAT ≥ 80% and capex/D&A ≈ 1.0 to 1.5. CFF stays negative year after year (dividends + buybacks), the share count is flat or shrinking, and net debt is zero or negative. In India the specimens are TCS, ITC, Hindustan Unilever. In the US they're Apple, Visa, Microsoft. The story is that the business needs little and gushes much, so the interesting question moves from operations to allocation. The standing follow-up asks whether the reinvestment runway is really this short. It also asks where the cash is going. Is it being returned at sensible prices, with buybacks below intrinsic value? Or is it being hoarded, or saved for a diversification adventure?

Personality 2: the capex monster. Its signature is healthy CFO, often with excellent CFO/EBITDA because operations are fine, but capex/D&A ≥ 2 for years. FCF ≈ 0 or negative, CFF is positive because debt or equity is feeding the build, and the dividend is token or absent. Two specimens stand out. Reliance Industries over FY13 to FY19 spent ≈ ₹5 lakh crore of capex against strong but smaller CFO, with net debt climbing until the 2020 deleveraging through a rights issue and stake sales. Intel over FY22 to FY24 posted CFO of $15.4bn, $11.5bn and $8.3bn against capex of $24.8bn, $25.8bn and $23.9bn building foundries. That's FCF of about −$9.4bn, −$14.3bn and −$15.7bn, before the partner contributions Intel nets out of its own "adjusted" FCF. The dividend was cut in 2023 and suspended in 2024 when the arithmetic refused to move. The story: everything depends on one number you can't see yet, the ROIIC of the build. The standing follow-up has three parts: the return management claims the capex will earn, what the incremental-ROIC table from returns on capital says they've historically earned, and whether the funding survives if the build runs late. A capex monster with a 25% historical ROIIC is a compounder in the making. One with 6% is a shareholder-funded monument.

Personality 3: the working-capital sink. Its signature is a healthy P&L with fine margins and growing PAT, while CFO runs chronically far below PAT (CFO/NI < 0.6 for years). Receivables or inventory grow faster than sales, sometimes both, tripping the DSO/DIO wires you already know. Borrowings creep up to fund the gap, and capex is often modest. The specimen is BHEL in the early 2010s. It reported PAT in the thousands of crores while its receivables (including deferred/retention balances) swelled past 200 debtor-days. The "profits" were claims on state utilities, and when the cycle turned, the paper profits became provisions. US cousins include EPC contractors generally, and growth-brand blowups where inventory outruns the hype. The story: the company books profit when it invoices, and its customers see matters differently. The standing follow-up: will the working capital ever convert? Find out who owes the money and how old it is, and whether the company is choosing growth on credit or being chosen by customers who don't pay. A widening PAT-versus-CFO jaw across 3+ years is this personality's rap sheet, and the forensic modules read it as probable cause.

Personality 4: the financial-engineering case. Its signature is a respectable-looking CFO, but decomposition finds Tier-4 sources: stretched payables, factored receivables, ballooning supply-chain finance. "Adjusted" earnings drift further and further from cash. Dividends + buybacks exceed FCF, funded by rising gross debt, and complexity builds up: reshuffled segments, related-party balances, sale-and-leasebacks. The specimens are patterns from the public record. GE before 2017 had industrial FCF chronically weak against headline EPS. It monetized receivables through GE Capital. It sent ~$40bn+ of buybacks out near the highs and cut the dividend twice. Later it settled with the SEC over accounting practices. In India, Future Retail-style structures were built on sale-and-leasebacks and vendor financing, with growth carried on lease and group-entity balances and promoter pledges over the top. They collapsed when the music stopped in 2020. The story: engineering isn't always fraud. Apple borrowing cheaply against offshore cash to fund buybacks was rational treasury work. The test is whether engineering supplements operating cash or substitutes for it. The standing follow-up: strip the engineering out. Restate CFO for factoring and SCF, cap distributions at true FCF, then ask what's left. If the equity story needs the engineering to work, the engineering is the story.

A word on classification. A company can migrate (Reliance went from capex monster to attempted compounder, and Intel from compounder to capex monster), and it can wear two personalities at once. Treat the label as an index to the right follow-up question. The analysis starts with that question.

4.13 The read ritual: your permanent checklist

Run this on every company, for good. Later in the course it becomes rows in your deep-dive workbook.

1. Sign pattern. CFO/CFI/CFF, five years across: which life-cycle shape?
2. Restate. Interest paid, lease principal, one convention. Write the restated CFO.
3. Bridge. NI to CFO: which tier is generating the cash? Which tier is generating the growth in cash?
4. Convert. CFO/NI and CFO/EBITDA, 5-yr average and cumulative; against the bands, against the trend.
5. Free. Comparable FCF (state your definition); FCF/PAT cumulative; FCF margin.
6. Split. Maintenance vs growth capex (two methods minimum); capex/D&A dial; owner earnings.
7. Dilute. SBC/revenue and SBC/FCF; owner FCF or per-share correction if material.
8. Walk. The ten-year sources-and-uses; paise-per-rupee; the four verdict questions.
9. Name. The personality, and its standing follow-up question. That question is your next hour of work.

Common mistakes & how experts think differently

Mistake 1: "cash flow can't be manipulated." Half-true, and so dangerous. CFO has fewer dials than net income, but Tier-3/4/5 staging can inflate it for years: payables stretch, factoring, classification flattery, capitalizing costs into CFI. And Satyam faked the cash balance itself. Experts trust CFO more and verify it anyway. They decompose the bridge and read the factoring and SCF notes. A too-smooth CFO gets the same suspicion as a too-smooth P&L.

Mistake 2: judging FCF from one year. Capex arrives in slabs, and working capital breathes. A single-year FCF/PAT of 40% (big capex year) or 140% (WC release) means almost nothing. Experts compute cumulative 5- and 10-year windows and read trends, as every benchmark above specifies.

Mistake 3: "negative FCF bad, positive FCF good." FCF is a residual. The builder with 25% ROIIC and negative FCF is creating value. The melting ice cube with 160% FCF/PAT is liquidating itself gracefully. Experts never quote FCF without ROIIC and the maintenance/growth split beside it.

Mistake 4: comparing CFO or FCF across accounting standards raw. The Ind AS borrower's CFO excludes interest paid. The IFRS/Ind AS retailer's CFO excludes lease principal. The US twin's CFO carries both. Compare them raw and the Indian leveraged company looks better by construction. Experts restate to one convention before computing any ratio, by reflex, the way you trained when you first built the statement.

Mistake 5: subtracting only the PP&E line as "capex." Capitalized software is an intangible addition, and so are development costs and licenses. All of them are capex in substance. Aggressive capitalizers rely on you missing them. Experts subtract PP&E and intangible purchases, then glance at capex/D&A and asset turns for the capitalization tripwire.

Mistake 6: taking the company's own "free cash flow" number. "FCF" is a non-GAAP term, and managements define it generously: before SBC settlements, before restructuring, net of disposals, before lease principal. Experts rebuild it from CFO and the investing section. They state their definition and reconcile it to management's. The difference is often the most informative number on the page.

Mistake 7: treating EBITDA as cash flow. EBITDA pays no taxes, funds no working capital, replaces no machines. Munger's verdict on "EBITDA earnings" was unprintable. The printable version is CFO/EBITDA persistently below 50%, at which point the EBITDA was mostly rumor. Experts use EBITDA as a standardizing denominator, and never as the cash.

Mistake 8: using depreciation as maintenance capex on autopilot. Under inflation it understates, because replacement costs today's prices. For young asset bases it overstates. And it imports depreciation-life games wholesale. Experts triangulate the depreciation proxy against Greenwald's method and unit math. They quote a range and let the disagreement between methods shape their view of accounting quality.

Mistake 9: ignoring SBC because "it's non-cash." It's non-cash the way paying rent in company shares would be non-cash. Experts subtract it (Fix A) or dilute for it (Fix B), and they reclassify offset-buybacks as compensation.

Mistake 10: reading capital allocation from the letter instead of the ledger. Every chairman's letter says "disciplined capital allocation." Experts build the ten-year walk and let the paise-per-rupee table testify. When the ledger and the letter disagree, the ledger is the deposition and the letter is the alibi.

How the expert's posture differs. A beginner reads the CFS hunting for a number like CFO or FCF. An expert reads it to answer three questions: whether the cash is real (tiers), whether it's enough (conversion ratios, after a fair count of capex), and where it's going (the walk). Every answer is held as a multi-year pattern, never a single point.

Worked examples

Worked example 1: Greenwald's maintenance-capex method, end to end

Take Deccan Retail Ltd (synthetic, ₹ crore), a grocery chain adding stores steadily. Five years of data:

In FY26, total capex is ₹320 crore and depreciation ₹150 crore, of which ₹10 crore is amortization of an acquired brand. Net income is ₹300 crore, and operating WC runs at ~6% of sales.

Step 1: the ratio. Average gross PP&E/sales = (39.8 + 40.5 + 39.8 + 40.2 + 40.0)/5 = 40.0%. It's stable, so the method's premise holds.

Step 2: growth capex. FY26 sales grew 6,000 − 5,400 = ₹600 crore. Supporting ₹600 crore of new sales at 40 paise of PP&E per rupee takes 0.40 × 600 = ₹240 crore of growth capex.

Step 3: maintenance capex. Total capex − growth capex = 320 − 240 = ₹80 crore.

Step 4: triangulate against the depreciation proxy. Depreciation excluding acquired-intangible amortization = 150 − 10 = 140. Add a ~15% inflation uplift for replacement cost and you get ≈ ₹160 crore. The two methods disagree by 2×: Greenwald says 80, the proxy says ~160. That disagreement is information. This is a young store fleet, so depreciation is running on recently built stores whose replacement is a decade away. The proxy front-loads a bill that isn't cash-due yet, while Greenwald measures what maintaining this year actually cost. A fair answer: current-cash maintenance ≈ ₹80 to 100 crore, rising structurally toward ₹160 crore as the fleet ages. Use ~₹120 crore for owner earnings, and write the trajectory down.

Step 5: owner earnings vs FCF. Required ΔWC = 6% × 600 = ₹36 crore. CFO = 300 + 150 − 36 = ₹414 crore. FCF = 414 − 320 = ₹94 crore, which puts FCF/PAT at a mediocre-looking 31%. Owner earnings = 300 + 150 − 120 − 36 = ₹294 crore. To reconcile: OE − FCF = 200 = growth capex 240 − the extra maintenance beyond Greenwald's 80 (40), and it ties.

The screen and the split disagree here, and the split wins. A screen sees a 31% FCF/PAT and frowns. The split shows a business that could pay out ~₹294 crore and chose to build instead, and the build looks sane. ₹240 crore bought ₹600 crore of new sales. At Deccan's ~9% EBIT margin and 25% tax, that's ~₹40 crore of new NOPAT on ₹240 crore, a ~17% pre-ramp ROIIC against a 12% hurdle. FCF was low because value was being created.

Saying that sentence, with those numbers under it, is what all this work is for.

Worked example 2: the SBC illusion: a 29% "FCF margin" that owners never receive

The company is Cirrus Data Inc. (synthetic, in $m). Its scale is modeled on a large listed SaaS company, to show the pattern without copying any one filing. In FY26 revenue is 2,800 and CFO is 850, which includes an SBC add-back of 1,100. Capex plus capitalized software is 50. Shares outstanding are 335m and rising 2.5% per year net, despite ~$400m/yr spent on buybacks.

The headline: FCF = 850 − 50 = $800m, a 28.6% FCF margin. The investor deck calls it "best-in-class free cash flow generation."

Fix A treats SBC as the cost it is. Owner FCF = 800 − 1,100 = −$300m, a −10.7% margin. Sanity-check it through the buyback lens. To leave owners undiluted, Cirrus would have to repurchase roughly the shares it grants, about $1,100m of cash, leaving 800 − 1,100 = −300. Both directions give the same answer. Measured as if owners must be kept whole, this company consumes ~$300m a year. The 28.6% margin exists only because the biggest slice of the wage bill is paid in the owners' equity.

Fix B gives you the per-share truth. Suppose headline FCF compounds at 25% for five years while net dilution runs 2.5%. FCF₅ = 800 × 1.25⁵ = $2,441m, and shares₅ = 335 × 1.025⁵ ≈ 379m. FCF/share goes from $2.39 to $6.44, a 21.9% per-share CAGR (= 1.25/1.025 − 1). The headline promised 25%. Three points of compounding gone, every year, forever. Now look at what the 2.5% net figure hides. With ~$400m/yr of buybacks retiring roughly 1.2% of shares (at a ~$33bn market value), gross dilution is running ~3.7%/yr. Half the "free" cash flow is already conscripted into mopping up grants. Those buybacks are compensation paid in arrears, and they return no capital.

Read the verdict carefully. Neither fix says Cirrus is a bad business. It may grow into its comp bill, and grant-date SBC can overstate the true economic cost when growth moderates. The point is narrower, and it doesn't bend: "FCF" that adds back SBC while the share count grows isn't owner cash flow. Any multiple or yield built on it overstates owner returns, and so does any conversion ratio. Here the gap is the difference between +$800m and −$300m. Quote both numbers. Make the market's chosen definition your explicit finding, so you aren't just inheriting it as an assumption.

Worked example 3: one toll road, two rulebooks: the classification mirage

Take twin toll-road operators with identical economics. Deccan Toll reports under Ind AS (₹ crore) and Pacific Toll under US GAAP ($m). Read the numerals as equal. Both have EBITDA 1,000, D&A 250, cash interest 400, cash taxes 100, ΔWC nil, capex 200. So each has EBIT 750, PBT 350, PAT 250.

Pacific (US GAAP, interest paid inside CFO): CFO = PAT 250 + D&A 250 − ΔWC 0 = 500. CFO/EBITDA = 50%. FCF = 500 − 200 = 300; FCF/PAT = 120%.

Deccan (Ind AS, interest paid in CFF): CFO (from PBT, Indian presentation) = PBT 350 + D&A 250 + finance costs added back 400 − taxes paid 100 = 900. CFO/EBITDA = 90%. Naive FCF = 900 − 200 = 700; FCF/PAT = 280%.

Same business. One prints 50% conversion and ₹300 of FCF. The other prints 90% and ₹700, which is 2.3× the free cash flow, from classification alone. A screener comparing the two raw would rank Deccan a conversion champion and Pacific a leaker.

The restatement fixes it. Deccan's comparable CFO = 900 − 400 = 500, and its comparable FCF = 500 − 200 = 300. Twins again. Here's the general rule you now own. For any leveraged Ind AS/IFRS company, find where interest is before you admire CFO, and subtract it (plus lease principal, for lease-heavy models) before any ratio crosses a border. At Deccan's 400 of interest against 1,000 of EBITDA, the flattery was worth 40 points of CFO/EBITDA. On a real Indian infrastructure borrower it routinely decides whether the company screens as investable at all.

Worked example 4: Building FCFF and FCFE from CFO, and reconciling the two (US)

Meet Rainier Systems Inc. (synthetic, $m, US GAAP, so interest paid is already inside CFO). The comparable FCF you built earlier is deliberately equity-flavored. Now make the two formal cousins Phase 3 uses explicit, and reconcile all three: FCFF (free cash flow to all capital providers) and FCFE (free cash flow to equity). That way, when valuation arrives, it's vocabulary you already own.

FY26: net income 620, D&A 240 and an increase in working capital of 50, so reported CFO = 620 + 240 − 50 = 810. Interest paid is 80, inside CFO under US GAAP. The tax rate is 25%, capex (PP&E + capitalized software) is 300, and net borrowing (debt raised − debt repaid) is +30. There's no lease-principal adjustment, because US GAAP keeps operating-lease rent inside CFO.

Step 1: the comparable FCF. CFO − capex = 810 − 300 = $510m. Because US-GAAP CFO is already after interest, this is an equity-flavored residual: FCFE before the effect of new borrowing.

Step 2: FCFF (the whole firm). FCFF is the cash the business throws off as if it carried no debt, so financing gets stripped out. CFO has already paid interest, so add the after-tax interest back, then subtract capex: FCFF = CFO + interest × (1 − tax) − capex = 810 + 80 × 0.75 − 300 = 810 + 60 − 300 = $570m.

Step 3: FCFE (equity's slice). Equity keeps what's left after lenders are served and after net borrowing changes hands: FCFE = CFO − capex + net borrowing = 810 − 300 + 30 = $540m. Put another way, it's the Step-1 comparable FCF ($510m) plus the $30m of new debt drawn.

Step 4: reconcile. FCFF − FCFE = 570 − 540 = $30m, which must equal interest × (1 − tax) − net borrowing = 60 − 30 = 30, and it does. The two measures differ by the after-tax cost of servicing debt, net of fresh borrowing. For a debt-free firm with no borrowing, all three numbers collapse into one.

One habit outlasts the arithmetic. Each measure has a matching discount rate, and Phase 3 won't forgive a mismatch. FCFF is discounted at WACC, which already blends in the after-tax cost of debt, and FCFE at the cost of equity. Discount FCFE at WACC and you count the debt benefit twice. Discount FCFF at the cost of equity and you drop it. It's the same "never mix a levered cash flow with an unlevered discount rate" reflex from the FCF definitions, now with numbers under it.

Worked example 5: FCF conversion across a capex supercycle: reading the residual (India)

A screen that sorts on FCF would blacklist Mahanadi Cement Ltd (synthetic, ₹ crore, Ind AS) on its FY26. Your job is to see why FCF turned negative before you judge it. Here are three years, restated to one convention by subtracting interest paid, which Ind AS parks in CFF, so the conversion ratio is fair:

The revenue row is there so every ratio below is reproducible from the exhibit alone. It also carries its own reading: EBITDA margin holds at 25.0%, 24.9% and 24.5% across the three years while the cash residual swings by more than 13 points of revenue, which is the example's argument in one line.

Step 1: restate and convert. Restated CFO = reported − interest = 1,050 / 1,110 / 1,160, so CFO/EBITDA = 70.0%, 68.5% and 66.3%. That's solid and only gently drifting, so operations are not deteriorating.

Step 2: the capex dial. capex/D&A = 560/480 = 1.17, then 1.96, then 1,560/520 = 3.00. The dial crosses 2 into "building", a deliberate capacity decision.

Step 3: watch the residual flip. Comparable FCF = restated CFO − capex = +490, then +130, then −400. FCF/PAT goes 79.0%, 19.7%, −57.1%, and the FCF margin moves from 8.2% to −5.6%. A ₹490-crore free-cash producer became a ₹400-crore consumer in two years with conversion barely moving. That proves FCF here is a residual of the capex choice, and it passes no verdict on earnings quality.

Step 4: split the FY26 capex. Maintenance ≈ D&A 520 × 1.15 (India replacement-cost uplift) ≈ ₹598 crore. Growth capex ≈ 1,560 − 598 ≈ ₹962 crore, and that's the discretionary reinvestment to interrogate.

The personality is a capex monster in the making (Personality 2). Operations are fine and FCF is negative by choice. The whole judgment rests on one figure the FCF line can't show: the ROIIC the ~₹962 crore of growth capex will earn. Against a ~12% hurdle, a build earning 18 to 20% is a compounder forming, and one earning 7% is a shareholder-funded monument. Log one caution: the CFO/EBITDA drift from 70% to 66%. Confirm it's temporary interest on build-funding debt, which is benign. If it's working capital creeping up, that's a question for the conversion ratios.

Practice set

Problems 1 to 3 are guided, with full scaffolding. Problems 4 to 9 are independent. Problems 10 and 11 are timed. All companies are synthetic unless named as real, and figures are ₹ crore unless stated. Do each one before reading its solution.

Problem 1 (guided). Kaveri Auto Components, five years:

(a) CFO/NI by year and cumulative. (b) What happened to profit vs cash across the window, in one sentence each? (c) Your read, and your next two checks.

Solution. (a) By year: 1.30, 1.16, 0.96, 0.79, 0.66. Cumulative: 593 ÷ 635 = 0.93. (b) Net income rose 60% while CFO fell 19%. (c) The cumulative 0.93 looks almost respectable, so the trend is the read. Every extra rupee of "profit" is converting worse, the classic accrual build-up. Next checks: DSO and DIO against sales, to find where the cash is stuck. Then read the receivables and contract-asset notes for signs of aging or factoring. Two more years of this shape and you're holding a forensic case, well past a ratio exercise. A five-year NI-up, CFO-down jaw is the highest-value early warning you'll learn in all this cash-flow work.

Problem 2 (guided). Chenab Industries, FY26 indirect bridge: NI 200, D&A 60, SBC 25, receivables fell 10, inventory fell 5, payables rose 90 (DPO went from 38 to 61 days), other −5. Prior-year CFO was 290. (a) Compute FY26 CFO. (b) Sort the CFO into quality tiers. (c) What CFO do you expect next year if profit grows ~10%?

Solution. (a) CFO = 200 + 60 + 25 + 10 + 5 + 90 − 5 = 385. (b) Tier 1 (earned) is NI + D&A + SBC − the normal WC toll, ≈ 280 or so. Tier 3 (episodic) is the receivable and inventory releases, +15. They're real, but they won't repeat. Tier 4 (borrowed) is the payables stretch, +90. A 23-day DPO jump manufactures one-time cash and then stops, and suppliers eventually reprice or retaliate. (c) Strip out Tiers 3 and 4 and the run-rate CFO is ≈ 280. With ~10% profit growth, expect ~300 to 310. In other words, reported CFO should fall next year even as profit rises, and that fall will be healthy. Nothing sinister is going on. Writing this expectation down before the next annual report is how analysts avoid being surprised by arithmetic they already owned. If CFO doesn't fall, that means another Tier-4 draw, DPO from 61 to 80 perhaps, and that's what a payables-stretch addiction looks like. The forensic modules formalize it.

Problem 3 (guided). Godavari Infra (Ind AS): EBITDA 800, cash taxes 140, ΔWC consumed 60, interest paid 220 (in CFF), lease principal paid 30 (in CFF). (a) Reported CFO and CFO/EBITDA. (b) US-comparable CFO/EBITDA. (c) Verdict against the bands.

Solution. (a) Reported CFO = 800 − 140 − 60 = 600, so CFO/EBITDA = 75%, which looks strong. (b) Subtract interest: 600 − 220 = 380, or 47.5%. If the comparison set is US companies whose operating-lease rent burdens CFO, subtract the 30 too: 350, or 43.8%. (c) The 75% was classification costume. Restated, Godavari falls below the 50% leakage line. Once it has actually paid its lenders (and landlords), well under half its EBITDA becomes operating cash. Same company, two verdicts, and only the restated one is real. It's Worked Example 3 turned into a habit.

Problem 4. Narmada Logistics (Ind AS): CFO 640 (interest paid 90 and lease principal 45 both in CFF). Purchases of PP&E 260, additions to intangibles (fleet-management software) 60, proceeds from routine truck disposals 15, PAT 500. Compute FCF four ways: (a) naive, (b) analyst's comparable (gross capex), (c) net of routine disposals, and (d) as a % of PAT for (a) and (c). State which you'd quote.

Solution. (a) Naive: 640 − 260 = 380. (b) Comparable: 640 − 90 − 45 − 260 − 60 = 185. (c) The disposals are routine fleet churn, so netting is defensible: 185 + 15 = 200. (d) FCF/PAT is 76% naive against 40% on the fair basis. Quote (c) with the definition stated: "FCF = CFO less interest and lease principal (Ind AS reclassifications) less gross capex including intangibles, net of routine fleet disposals: ₹200 crore, 40% of PAT." The naive number was nearly double the fair one. Quote an FCF without a definition and you haven't really quoted a number.

Problem 5. Tapti Consumer: NI 240, D&A 90 (of which 15 is amortization of acquired brands), total capex 150 and required ΔWC 20. Assume a 1.2× inflation uplift on replacement. (a) Maintenance capex by the depreciation proxy, done properly. (b) Owner earnings. (c) FCF, and reconcile OE − FCF.

Solution. (a) Exclude acquired-intangible amortization, since it carries no replacement obligation: 90 − 15 = 75. Uplift it: 75 × 1.2 = ₹90 crore. (b) OE = 240 + 90 − 90 − 20 = ₹220 crore. (c) CFO = 240 + 90 − 20 = 310, and FCF = 310 − 150 = ₹160 crore. OE − FCF = 60 = growth capex (150 − 90), so it reconciles. Reading: of the ₹150 crore capex bill, ₹90 crore is the cost of standing still and ₹60 crore is a choice. Judge the choice on ROIIC, and value the business on the ₹220 crore its owners could take.

Problem 6. Yamuna Retail, Greenwald split. Sales FY22–FY26: 2,000, 2,150, 2,300, 2,450, 2,600. Gross PP&E: 700, 752, 806, 858, 910. FY26 total capex 130; FY26 depreciation 70. (a) Maintenance capex, Greenwald method. (b) Triangulate with the proxy. (c) What fraction of reported capex is the cost of standing still?

Solution. (a) PP&E/sales is 35.0%, 35.0%, 35.0%, 35.0%, 35.0%, so the average is 35%. That's admirably stable, and the premise holds. Δsales = 150, so growth capex = 0.35 × 150 = 52.5, and maintenance = 130 − 52.5 = 77.5. (b) The proxy gives 70, or mid-70s to 80 with a modest uplift. The methods agree at ≈ 70 to 80, a mature fleet, unlike the young one in Worked Example 1. When methods agree like this, that's evidence of clean depreciation accounting. (c) About 60% of capex (77.5/130) is maintenance. Most of the "investment" the company reports is just staying alive. Its real discretionary reinvestment is ~₹50 crore, and its owner earnings run above its FCF only modestly.

Problem 7. Four companies, sustained 3-year signatures. Classify each on the life-cycle/aggression dial and name the follow-up question. (A) Capex/D&A 2.8, revenue +25%/yr. (B) Capex/D&A 1.25, revenue +8%/yr. (C) Capex/D&A 0.5 for six years, revenue flat, operating margin rising. (D) Capex/D&A 3.2, revenue +4%/yr, fixed-asset turns falling every year.

Solution. (A) Builder. Capacity is racing ahead, so ask what ROIIC the build will earn and whether the funding is survivable. (B) Steady state. 1.25 is health itself for a mature firm under inflation, and there's no question to ask. (C) Harvester/under-investor. Six years below replacement with flat sales. Part of the rising margin is unspent maintenance, borrowed from the future. Ask when the deferred bill arrives (margin decay or a capex "surprise"), and whether management is milking the asset deliberately (fine, if priced) or out of negligence. (D) The aggression tripwire. Massive capex, little growth and falling asset turns mean money is leaving through CFI without becoming productive assets. Ask what is being capitalized, and read the PP&E/CWIP and intangible-additions notes. It's the WorldCom/Vakrangee shape, and it goes on the forensic pile.

Problem 8. Meghna Soft ($m, US GAAP): revenue 900; headline FCF 180; SBC 90; 200m shares, net dilution 3%/yr; management guides "15% FCF growth, best-in-class 20% FCF margin." (a) Owner FCF (Fix A) and margin. (b) Per-share FCF growth owners actually receive (Fix B). (c) One sentence on what a 30× multiple on headline FCF is really paying.

Solution. (a) 180 − 90 = $90m, a margin of 10%, half the headline. (b) 1.15 ÷ 1.03 − 1 ≈ 11.7% per share, short of the 15%. (c) 30 × 180 = $5.4bn capitalizes a stream of which half is the employees' slice, paid in owners' equity. On owner FCF the same price is 60×, so the buyer is paying double the quoted multiple, and the gap compounds at ~3% dilution a year. Both fixes overstate the damage if growth later moderates and grants shrink, and you should say so. Still, the burden of proof is on the diluter.

Problem 9. Four five-year signatures. Assign each a personality and its standing follow-up question. (W) CFO/NI 1.1 steady; cumulative FCF/PAT 85%; capex/D&A 1.2; CFF negative every year; share count −1%/yr; net cash. (X) CFO/EBITDA 78%; capex/D&A 2.6; FCF negative throughout; net debt rising; token dividend. (Y) PAT +12%/yr; CFO/NI drifting from 0.9 to 0.5; DSO from 90 to 150 days; short-term borrowings up every year; capex modest. (Z) CFO/NI ≈ 1.0, but DPO from 40 to 75 days and a new ₹300-crore receivables-factoring program in the notes; dividends + buybacks = 130% of FCF, funded by fresh debt; "adjusted EBITDA" pulling away from CFO.

Solution. (W) Compounder. Ask whether the ~100% payout is right and whether buybacks are priced below intrinsic value. (X) Capex monster. Operations are fine (78%!), and everything rides on the build's ROIIC and how long the funding lasts. (Y) Working-capital sink. The profits are invoices that haven't turned into cash. Ask who owes the DSO-150 money and how old it is. Also ask whether the company books revenue its customers dispute. (Z) Financial-engineering case. CFO/NI looks fine only because Tier-4 sources (stretch + factoring) plug it. Strip them, cap distributions at true FCF, and ask whether the equity story survives. Notice the deliberate trap in Z: its headline ratio was the healthiest of the four. Diagnosing a personality is about composition, and the top-line ratio can mislead you.

Problem 10 (timed, 12 minutes). Decade cash walk, Sindhu Chemicals: cumulative FY17–FY26: CFO 1,000; net debt raised +150; disposals 50; capex 300; acquisitions 120; dividends 260; interest paid (CFF, Ind AS) 80; cash pile grew 40; buybacks unknown. (a) Solve for buybacks via the sources-and-uses identity. (b) Distributions vs decade FCF: funded how? (c) Two-sentence verdict with the follow-up question.

Solution. (a) Sources = 1,000 + 150 + 50 = 1,200. Known uses = 300 + 120 + 260 + 80 + 40 = 800. Buybacks = 400. (b) Decade comparable FCF = 1,000 − 80 − 300 = 620. Distributions = 260 + 400 = 660, or 106% of FCF. The 40-crore gap plus the 120 of M&A were funded by debt (+150) and disposals (+50). (c) Verdict: a good-not-great converter paying out slightly more than it frees and borrowing the balance. That's sustainable only if the reinvested 420 (capex + M&A) earns above the hurdle and disposals don't become a habit. Follow-up: the ROIIC of that 420, and whether the buyback tranches were priced below intrinsic value or were merely fashionable.

Problem 11 (timed, 20 minutes; the full ritual). Nilgiri Beverages (Ind AS). FY26: revenue 2,800 (grew ~10%); PAT 300; D&A 100; SBC 10; ΔWC consumed 70; other +10; interest paid 60 and lease principal 20, both in CFF; capex: PP&E 130 + intangibles 20; dividends paid 120; net debt raised +40; EBITDA 560. Five-year cumulatives: PAT 1,250; reported CFO 1,400; EBITDA 2,450; capex 650; interest paid 280; lease principal 90. Operating WC runs ≈ 25% of sales. Deliver the read ritual: restated CFO; conversion ratios (FY26 and cumulative); comparable FCF and FCF/PAT; capex/D&A and a maintenance estimate; the personality with evidence; verdict plus follow-up.

Solution.

1. Reported FY26 CFO = 300 + 100 + 10 − 70 + 10 = 350. Restated = 350 − 60 − 20 = 270.
2. Conversion: FY26 CFO/NI goes from 1.17 reported to 0.90 restated. Cumulative restated CFO = 1,400 − 280 − 90 = 1,030, so CFO/NI = 1,030/1,250 = 0.82, in the investigate band. CFO/EBITDA = 1,030/2,450 = 42%, below the 50% line even before capex.
3. FCF: FY26 comparable = 270 − 150 = 120, and FCF/PAT = 40%. Cumulative = 1,030 − 650 = 380, which is 30% of PAT.
4. Capex: capex/D&A = 150/100 = 1.5, which looks like building. But the depreciation proxy (≈ 100 × 1.15 ≈ 115) says most of it is maintenance, with growth capex ≈ 35. So the 30% FCF/PAT isn't a build-phase story.
5. WC check: the toll ≈ 25% × Δrevenue (≈ 255) ≈ 65, close to the 70 consumed. That's structural. It's a heavy model run fairly. There's no stretch or spike, and no shenanigan (yet).
6. Personality: a mild working-capital sink with harvest-adjacent capex. Profits convert at 82 paise cumulatively, EBITDA leaks 58%, and the model's appetite for WC is the cause.
7. Verdict: FY26 dividends (120) equal the entire comparable FCF (120), with net debt +40 bridging the gap. That's a distribution stretched to its limit on a 0.82-conversion business. Follow-up: is WC% of sales stable (structural) or creeping (deterioration)? Check the five-year DSO/DIO trend next, and the interest-cover trajectory if the dividend keeps being part-borrowed.

Problem 12 (guided). Klamath Robotics Inc. (synthetic; $m; US GAAP). FY26: CFO 640 (interest paid of 60 is already inside it); tax rate 25%; capex (PP&E 180 + capitalized software 40) = 220; net borrowing = debt raised 90 − debt repaid 40 = +50; net income 380. (a) FCFF. (b) FCFE. (c) The comparable FCF from earlier, and how it relates to FCFE. (d) Reconcile FCFF and FCFE.

Solution. (a) FCFF = CFO + interest × (1 − t) − capex = 640 + 60 × 0.75 − 220 = 640 + 45 − 220 = $465m. (b) FCFE = CFO − capex + net borrowing = 640 − 220 + 50 = $470m. (c) Comparable FCF = CFO − capex = 640 − 220 = $420m. That's the equity-flavored residual, FCFE before net borrowing. Add the +50 of new debt and you reach the $470m FCFE. (d) FCFF − FCFE = 465 − 470 = −$5m = interest × (1 − t) − net borrowing = 45 − 50 = −5, so it reconciles. FCFE exceeds FCFF here only because the company borrowed slightly more ($50m) than the after-tax cost of servicing its debt ($45m). It's a one-year financing quirk, and it says nothing about operating strength.

Problem 13 (guided). Chambal Cement Ltd (synthetic; ₹ crore; Ind AS), three years:

(a) Restated CFO/EBITDA each year. (b) capex/D&A each year. (c) Comparable FCF and FCF/PAT each year. (d) Is this deteriorating operations or a deliberate build, and what single number decides the verdict?

Solution. (a) Restated CFO = CFO − interest = 640 / 675 / 700, so CFO/EBITDA = 71.1% / 68.9% / 66.0%. Steady and healthy: operations are fine. (b) capex/D&A = 1.07 / 1.81 / 3.06, crossing 2 into "building." (c) Comparable FCF = restated CFO − capex = +320 / +115 / −280, and FCF/PAT = 88.9% / 29.5% / −66.7%. (d) It's a deliberate build. Conversion barely moved while capex outran the D&A line, so the FCF collapse is a residual of the capex decision and raises no earnings-quality alarm. The verdict rests entirely on the ROIIC of the growth capex. Negative FCF plus 20% incremental returns is a compounder forming. Negative FCF plus single-digit returns is destruction. FCF alone can't tell you which.

Problem 14. Sutlej Power Ltd (synthetic; ₹ crore; Ind AS). FY26: PAT 540; D&A 380; increase in working capital 60; interest paid 260 (in CFF); tax rate 25%; capex (PP&E + intangibles) 520; net borrowing = debt raised 80 − debt repaid 300 = −220 (a net-repayment year). (a) Reported CFO (Ind AS). (b) Comparable FCF and FCFE, and why they differ so much this year. (c) [Stretch] FCFF via the NOPAT route, confirmed with the Ind AS CFO shortcut.

Solution. (a) CFO = PAT + interest + D&A − ΔWC = 540 + 260 + 380 − 60 = ₹1,120 crore. Ind AS adds interest back in operations and shows it as an outflow in CFF. (b) Comparable FCF = CFO − interest − capex = 1,120 − 260 − 520 = ₹340 crore, which is FCFE before net borrowing. FCFE = 340 + (−220) = ₹120 crore. They diverge because the company repaid ₹220 crore net. Equity's cash flow is what's left after lenders get their interest and their principal back, so a heavy deleveraging year mechanically shrinks FCFE even with operations unchanged. (c) The NOPAT route: PBT = PAT ÷ (1 − t) = 540/0.75 = 720; EBIT = 720 + 260 = 980; NOPAT = 980 × 0.75 = 735; FCFF = NOPAT + D&A − ΔWC − capex = 735 + 380 − 60 − 520 = ₹535 crore. Now the shortcut check. Because Ind AS CFO added back the full pre-tax interest, FCFF = CFO − capex − interest × tax = 1,120 − 520 − 260 × 0.25 = 535, and it matches. The − interest × tax term removes the interest tax shield that the reported CFO enjoyed and an all-equity firm wouldn't. The US-GAAP CFO + interest × (1 − t) − capex formula does not apply to an Ind AS CFO that never subtracted interest, because using it would double-count. Locate interest before you pick a formula.

Problem 15. Cascade Networks Inc. (synthetic; $m; US GAAP). FY26: EBIT 800; interest paid 100 (in CFO); tax rate 25%; D&A 260; increase in working capital 40; capex 300; net borrowing +60. (a) Net income and reported CFO. (b) FCFF two ways, from CFO and from NOPAT, and confirm they agree. (c) FCFE. (d) Which discount rate values each of FCFF and FCFE, and what goes wrong if you swap them?

Solution. (a) NI = (EBIT − interest) × (1 − t) = (800 − 100) × 0.75 = $525m, and CFO = NI + D&A − ΔWC = 525 + 260 − 40 = $745m. (b) From CFO: 745 + 100 × 0.75 − 300 = 745 + 75 − 300 = $520m. From NOPAT: EBIT × (1 − t) + D&A − ΔWC − capex = 600 + 260 − 40 − 300 = $520m, so they agree. Two roads, one FCFF. A disagreement would mean you'd made a classification or WC error, since only one answer can be valid. (c) FCFE = CFO − capex + net borrowing = 745 − 300 + 60 = $505m. Check: FCFF − interest × (1 − t) + net borrowing = 520 − 75 + 60 = 505. (d) FCFF pairs with WACC, and FCFE pairs with the cost of equity. Discount FCFE at WACC and you reward the firm twice for its cheap debt, because WACC already contains the after-tax cost of debt, and you overstate value. Discount FCFF at the cost of equity and you ignore the debt benefit, which understates it. The cash flow and the rate must speak about the same claimants. It's the first commandment of valuation.

Problem 16. Wainganga Foods Ltd (synthetic; ₹ crore; Ind AS), five-year cumulatives FY22–FY26: EBITDA 4,000; reported CFO 2,600; interest paid (CFF) 500; lease principal (CFF) 150; D&A 900; capex (PP&E + intangibles) 1,200; PAT 1,800; revenue 15,000. (a) Restated cumulative CFO and CFO/EBITDA. (b) Cumulative comparable FCF, FCF/PAT, and FCF margin. (c) capex/D&A. (d) Compounder or heavy converter, and is the low FCF/PAT a build story?

Solution. (a) Restated CFO = 2,600 − 500 − 150 = 1,950, and CFO/EBITDA = 1,950/4,000 = 48.8%, below the 50% leakage line before a rupee of capex. (b) Comparable FCF = 1,950 − 1,200 = 750. FCF/PAT = 750/1,800 = 41.7%, in the "demand answers" band. FCF margin = 750/15,000 = 5.0%, the capital-heavy habitat. (c) capex/D&A = 1,200/900 = 1.33, steady state and not a build. (d) It's a heavy converter, and calling it a compounder would be wrong. With capex/D&A at 1.33 there's no build phase to excuse the sub-50% FCF/PAT, so the unavailable-alibi rule bites. Interest and lease costs are eating the middle, helped by a working-capital-hungry model (Personality 3-adjacent). None of it is investment. The ROIIC defense that rescues a real capex monster doesn't apply here. The profits are structurally low-conversion, and the next check is the five-year DSO/DIO trend, to see whether "structural" is sliding toward "deteriorating."

Problem 17 (timed, 18 minutes; the valuation-bridge ritual). Willamette Devices Inc. (synthetic; $m; US GAAP). FY26: revenue 12,000; net income 1,100; D&A 900; SBC 150; increase in working capital 120; interest paid 200 (in CFO); tax rate 25%; EBITDA 2,600; capex (PP&E 1,700 + capitalized software 300) = 2,000; net borrowing +100; dividends 150; buybacks 100; share count roughly flat. Deliver: (1) CFO and CFO/EBITDA; (2) FCFF and FCFE from CFO; (3) comparable FCF, FCF/PAT, FCF margin, and owner FCF; (4) capex/D&A and a maintenance estimate; (5) the sources-and-uses tie-out; (6) personality, verdict, and follow-up.

Solution.

1. CFO = NI + D&A + SBC − ΔWC = 1,100 + 900 + 150 − 120 = $2,030m. CFO/EBITDA = 2,030/2,600 = 78%, which is strong, so operations are healthy.
2. FCFF = CFO + interest × (1 − t) − capex = 2,030 + 150 − 2,000 = $180m; FCFE = CFO − capex + net borrowing = 2,030 − 2,000 + 100 = $130m (reconcile: FCFF − FCFE = 50 = interest × (1 − t) 150 − net borrowing 100 ✓).
3. Comparable FCF = CFO − capex = $30m. FCF/PAT = 30/1,100 = 2.7%, FCF margin = 30/12,000 = 0.25%, and owner FCF = 30 − SBC 150 = −$120m. The "$2bn CFO" business frees almost nothing this year, and once SBC is charged it's negative to owners.
4. capex/D&A = 2,000/900 = 2.22, so it's building. Maintenance ≈ 900 × 1.1 = ~$990m, leaving growth capex ≈ ~$1,010m.
5. Sources = CFO 2,030 + net borrowing 100 = 2,130. Uses = capex 2,000 + dividends 150 + buybacks 100 = 2,250. The difference, −120, is a $120m cash drawdown, and it ties.
6. Personality: a capex monster (Personality 2), with strong conversion, capex/D&A > 2 and FCF ≈ 0. It's also distributing $250m against $30m of comparable FCF. Borrowing and cash fund part of that, an engineering-lite flag. Notice too that the $100m of buybacks is smaller than the $150m of SBC, so none of it is a real distribution. It's compensation paid in arrears. Follow-up: (i) what ROIIC does the ~$1,010m of growth capex earn, and is the build funded to completion? (ii) can the dividend survive without borrowing once the build peaks? A capex monster paying a part-borrowed dividend isn't in trouble yet, but it belongs on the watch-list, and those two questions are your next hour of work.

Applied mini-project

Two decades, two philosophies, and one question: where did the cash go? The companies are Avenue Supermarts (DMart) and Texas Instruments. For DMart, get the annual reports from the company's investor-relations page or BSE, and the ten-year history from screener.in. For TI, use the 10-Ks on SEC EDGAR and history from stockanalysis.com or macrotrends, plus TI's famous annual Capital Management presentation (ti.com, investor relations). The pairing is deliberate. DMart is an Indian grower whose FCF has been thin by choice while it built stores at high ROIIC, and your owner-earnings machinery decides whether the choice was right. TI is a US company that literally markets itself on FCF per share. Then, from 2021, it walked into a deliberate capex supercycle that crushed the very metric it champions, with capex/D&A ≈ 3+ and FCF collapsing while CFO held. That's your life-cycle dial and maintenance/growth split, live. Time: 3 to 4 hours.

1. Restate. DMart (Ind AS): move interest paid and lease principal (post-FY20 Ind AS 116) out of the comparison. TI (US GAAP) needs none, but say why. Show restated CFO for both, five years.
2. Convert. CFO/NI and CFO/EBITDA, five-year and ten-year cumulative; FCF (definition stated), FCF/PAT cumulative, FCF margin. One line each against the bands above.
3. Split the capex. Run both methods on both companies. Use Greenwald's table, and check the PP&E/sales stability premise, which fails informatively for TI mid-supercycle; say so and fall back to the proxy plus management disclosure. Use the unit method for DMart: stores from the AR, capex per new store × additions, against total capex. Chart capex/D&A by year, mark the year TI's dial crossed 2, and classify each company's life-cycle phase.
4. Owner earnings. Compute for both, latest year, with your maintenance estimate and normalized ΔWC (DMart WC% × Δrevenue). State where OE sits relative to FCF and why (build vs harvest logic).
5. SBC. SBC/revenue and SBC/FCF for TI (modest but real) and DMart (small); recompute TI's headline FCF-per-share growth net of dilution/buyback-offset. One line: does TI's own favorite metric survive your Fix B?
6. The walk. Ten-year sources-and-uses for both, residual under 3%; paise-per-rupee and cents-per-dollar tables; the four verdict questions answered in writing; assign each company its personality and follow-up question.
7. Memo. One page per company, conclusion first: "Are the profits cash, is the capex a choice, and where is the money going?" Every claim carries a number; every number carries a source (filing page/note).

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

File both companies in your company files. The cash-walk tables become permanent exhibits in the deep-dive workbook you'll build later in the course, and DMart's owner-earnings verdict gets re-tested when you value it properly in Phase 3.

Reading & resources

• Warren Buffett, 1986 letter to Berkshire Hathaway shareholders, the appendix on "owner earnings" (berkshirehathaway.com/letters). Six paragraphs that out-teach most textbooks; note how much of the definition is an estimate, and why he prefers that to false precision. [Free] [Beginner]
• Bruce Greenwald, Judd Kahn et al., Value Investing: From Graham to Buffett and Beyond (2nd ed., 2020). The earnings-power-value chapters and the maintenance-capex treatment within them, which is the original home of Method 2 in the capex split. [Paid] [Intermediate]
• Howard Schilit, Jeremy Perler & Yoni Engelhart, Financial Shenanigans (4th ed.). Part 3, "Cash Flow Shenanigans" (the CF-1 to CF-3 chapters): inflating CFO via classification, one-offs, and unsustainable boosts. Read it now as a preview; M2.06–M2.07 assign it properly. [Paid] [Intermediate]
• Aswath Damodaran, "Stock-based Employee Compensation: Value and Pricing Effects". Musings on Markets blog, February 2014 (aswathdamodaran.blogspot.com); the case for treating SBC as a real expense, with valuation consequences. [Free] [Intermediate]
• Michael Mauboussin & Dan Callahan, "Stock-Based Compensation" (Morgan Stanley Counterpoint Global Insights, 2023; search the Counterpoint Global insights page). The fullest data treatment of SBC's scale, dilution mechanics, and the offset-buyback fallacy. [Free] [Intermediate]
• McKinsey (Koller, Goedhart & Wessels), Valuation (8th ed., 2025). The "Reorganizing the Financial Statements" chapter: free cash flow built rigorously from NOPAT and invested capital, and the bridge from the analyst FCF you build here to M3.04's FCFF. [Paid] [Advanced]
• CFA Institute, Level I reading "Understanding Cash Flow Statements" (in the FSA volume; also the basis of free summaries across the web). Conversion ratios and cross-standard classification in exam-grade rigor. [Paid] [Beginner→Intermediate]
• Saurabh Mukherjea, Rakshit Ranjan & Salil Desai, Diamonds in the Dust (2021). The accounting-quality chapters: the cumulative CFO/EBITDA screen and India-specific conversion forensics used by Marcellus; calibrate, don't outsource. [Paid] [Beginner]
• Zerodha Varsity, Fundamental Analysis module, the cash-flow-statement chapter (zerodha.com/varsity). The Indian-retail-investor vocabulary for everything here; useful for teaching it back. [Free] [Beginner]
• Texas Instruments, "Capital Management" annual presentation (ti.com investor relations). A rare management team that presents its own FCF-per-share doctrine; read the latest against your mini-project findings. [Free] [Beginner]
• Filings & data to pull (all free): TCS annual reports + screener.in ten-year page (for the ten-year walk); Apple 10-Ks FY2015–FY2024 via SEC EDGAR (for the ten-year walk); Reliance Industries annual reports FY2013–FY2020 (capex-monster profile); Intel 10-K FY2023 (dividend-vs-FCF arithmetic); Avenue Supermarts and Texas Instruments filings (mini-project); Berkshire 1986 letter (owner earnings). stockanalysis.com and macrotrends.net for ten-year CFS histories at speed. [Free] [All levels]
• Going deeper: Aswath Damodaran, "Free Cashflow to the Firm (FCFF) and to Equity (FCFE)" (his NYU valuation lecture notes and Musings on Markets, free at pages.stern.nyu.edu/~adamodar). The formal FCFF/FCFE derivations and the strict discount-rate pairing (FCFF↔WACC, FCFE↔cost of equity) that Worked Example 4 previews; read it just before M3.04 so the bridge is already built. [Free] [Intermediate]
• Going deeper: CFA Institute, Level II reading "Free Cash Flow Valuation". The four routes to FCFF (from net income, EBIT, EBITDA, and CFO) and to FCFE, with the exact add-back tables and the levered-versus-unlevered matching rule; the exam-grade home of Worked Example 4 and the FCFF/FCFE practice builds (Problems 12, 14, 15, 17). [Paid] [Intermediate→Advanced]

The Modern Analyst's Addendum

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

AI-Augment this skill

{
  "skill": "The ten-minute cash-flow read, the CFO/NI and CFO/EBITDA conversion tests, free cash flow defined and restated, the maintenance-capex split, owner earnings and the ten-year cash walk",
  "use": "The judgment in this module is about composition — which line ate the gap — and composition is where a tool is a fast hypothesis generator and a poor witness. (1) Hand it a CFO/EBITDA gap you have already decomposed into cash taxes, working capital and cash interest, and ask which business situations produce that specific composition; the answer is a shortlist to test against the notes, not a diagnosis. (2) Ask it to enumerate every convention difference between two companies' cash-flow statements before you compare them — interest classification, taxes, lease principal, the treatment of factored receivables — because the restatement in §4.4 is mechanical work that must be exhaustive and is easy to leave half-done. (3) For the maintenance-capex split, ask it to hunt the earnings-call Q&A and the capex note across several years for any management statement about replacement or refresh spend, then use it as method three's input; that is genuine document search on a question the statements never answer. (4) For the ten-year cash walk, ask it to draft the narrative from your completed table and then attack the draft — a walk that reads plausibly while summing incorrectly is the failure mode to watch for.",
  "tools": ["Chat assistants — Claude, ChatGPT for candidate explanations of a decomposed conversion gap and for the cash-walk narrative", "Document/PDF AI — for the capex note, the lease note and earnings-call Q&A across several years on maintenance spend", "Coding assistants — for the conversion panel, the Greenwald split and the owner-earnings range below"],
  "prompt": "Here is a company's cash-flow statement, EBITDA and net income for ten years, with the reporting framework stated: <PASTE>. First list every classification convention in these statements that would have to be restated before comparing them with a <OTHER FRAMEWORK> filer, and the direction of each effect. Then, for the year with the widest CFO-to-EBITDA gap, list the operating explanations consistent with the specific composition of that gap and the disclosure that would confirm each. Do not compute maintenance capex or owner earnings — those are estimates and I will make them.",
  "verify": "Every conversion ratio is recomputed from the filed statement, and every comparison is restated to one convention before it is made — an Ind AS reporter's CFO excludes interest paid, and the Quantitative lens shows exactly how many points of flattery that is worth. Sum the ten-year cash walk and require it to close on the change in cash and debt; a walk that does not close is not a walk. And treat every maintenance-capex figure a tool produces as unusable: it is an estimate that no statement discloses, it moves owner earnings one-for-one, and the whole discipline of §4.7 is triangulating it yourself and quoting a range. A model asked for maintenance capex will give you a number to two decimals (Guardrail 1-5).",
  "diy": "The gate is unaided: you run the ten-minute read, compute and restate the conversion ratios, triangulate maintenance capex three ways, build owner earnings, and construct the ten-year cash walk on paper. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You ran the ten-minute read, walked the CFO-to-EBITDA gap component by component, and triangulated maintenance capex three ways on paper. Keep all of that. The code below doesn't replace a single step. It makes the window long enough for the ratios to mean what the conversion benchmarks say they mean.

Today's workflow. The rules above demand a panel: multi-year averages for CFO/NI, cumulative windows for CFO/EBITDA, three-year rolling capex/D&A, a ten-year cash walk. Each one is a groupby over a frame and a nuisance by hand, and that's why single-year readings persist in practice. Three moves are worth writing down. First, restate before comparing. Subtract cash interest from an Ind AS CFO (or add it to a US one) as a column, so the panel never compares across conventions. Second, put both windows side by side: the single-year conversion ratio and the cumulative one. The divergence between them is itself the diagnostic, and the Quantitative lens measures how often the single year lies. Third, make the maintenance-capex split a set of methods. Use the depreciation proxy with and without an inflation uplift, Greenwald's sales-linked estimate from a five-year average gross-PP&E-to-sales ratio, and a unit-economics build where units are countable. Carry all of them into owner earnings and report the range. On a synthetic five-year table with a gross-PP&E-to-sales ratio averaging 0.6219, Greenwald's method attributes ₹90.2 crore of the final year's ₹140 crore capex to growth and ₹49.8 crore to maintenance, against a depreciation proxy of ₹90 crore. That's a disagreement of nearly two times, and you've already seen why that disagreement is information worth keeping.

Tools & sources (IN + US). pandas for the panel and the rolling windows, and numpy for the ranges. For India, use the Ind AS 7 cash-flow statement (remember that interest paid goes to CFF for non-financial companies), the PP&E and capex notes, the CWIP ageing schedule from the 2021 Schedule III amendments, the ESOP note for the SBC leg, and the concall transcripts on the exchanges for any management statement on replacement spend. BSE/NSE carries all of it, and screener.in has the ten-year convenience series you reconcile. For the US, use EDGAR XBRL companyfacts (NetCashProvidedByUsedInOperatingActivities, PaymentsToAcquirePropertyPlantAndEquipment, DepreciationDepletionAndAmortization, ShareBasedCompensation, InterestPaidNet) and the 10-K capex and property notes. Add the earnings-call transcripts. FRED (PPIACO, WPUFD4131) covers the replacement-cost inflation that the depreciation proxy structurally misses.

# Synthetic five-year series and the module's own ice-cube figures — R3: recomputed in-session
import numpy as np
sales = np.array([1000., 1120., 1210., 1355., 1500.])
gross_ppe = np.array([620., 690., 760., 845., 940.])
capex, dep = np.array([95., 105., 110., 130., 140.]), np.array([62., 68., 74., 82., 90.])
ratio = (gross_ppe / sales)[:-1].mean()                     # 0.6219 — Greenwald's anchor
maint_g = capex[1:] - ratio * np.diff(sales)                # final year: 49.8 vs dep proxy 90
print(round(ratio, 4), round(maint_g[-1], 1))
ni = 120.0
oe = {k: ni + dep[-1] - m for k, m in                       # owner earnings, one per method
      {"dep": dep[-1], "dep+25%": dep[-1] * 1.25, "greenwald": maint_g[-1], "units": 100.}.items()}
print({k: round(v, 1) for k, v in oe.items()}, round(ni + dep[-1] - capex[-1], 1))   # FCF = 70.0
print(round(max(oe.values()) / min(oe.values()) - 1, 3))    # 0.643 -> a 64% owner-earnings range
print(180., 160., 160.0, 100. + 80. - 70.)                  # §4.8 ice cube: CFO, FCF, %, OE = 110

Verify. Reconcile the panel's CFO, capex, D&A and net income to a filing you've read by hand before you quote any ratio. Record which framework each company reports under, because the flattery is structural and no accident. Sum the ten-year walk and require it to close on the change in cash plus the change in debt. No close, no narrative. And label every owner-earnings figure with the maintenance-capex method that produced it, in the memo, every time. An owner-earnings number without its estimate attached hides its most important assumption, and Buffett said as much himself when he refused to compute it to the decimal.

Quantitative lens

Three of the rules of thumb above are calibration claims, and each is worth checking numerically before you accept it.

How much is the convention flattery worth? The conversion section said an Ind AS reporter's CFO/EBITDA is structurally higher than a US twin's, by 15 to 30 points for a leveraged company. The relation is an identity, worth carrying as a law. The only difference is whether cash interest is inside CFO, so the flattery in percentage points equals cash interest as a percentage of EBITDA. Take EBITDA of 1,000 with cash taxes of 150 and a working-capital use of 50. Cash interest of 100 gives 70.0% under the US convention against 80.0% under Ind AS, a 10-point gap. Interest of 200 gives 60.0% against 80.0%, a 20-point gap. Interest of 300 gives 50.0% against 80.0%, a 30-point gap. So the range was never a rule of thumb. It's what you get when cash interest runs 15% to 30% of EBITDA, and that's what "a leveraged company" means. Compute the one ratio and you know the adjustment.

How often does a single-year CFO/NI reading lie? Take an entirely healthy asset-light services company: net income 100, D&A 10, and a working-capital swing with a standard deviation of 25 in either direction and no drift. Nothing's wrong and nothing is deteriorating. Mean CFO/NI is 1.10. The single-year ratio has a standard deviation of 0.250, so it falls below the 0.8 red-flag line in 11.5% of years. Now the number that matters for screening: 45.7% of such companies will print at least one sub-0.8 year somewhere in a five-year window. The five-year cumulative ratio has a standard deviation of 0.112, √5 times smaller as theory requires, and falls below 0.8 in only 0.4% of cases. So cumulating does far more than tidy the numbers. It cuts false positives a hundredfold.

What does the unobservable estimate cost? Owner earnings subtracts maintenance capex one-for-one, and maintenance capex isn't disclosed. On the synthetic company above (net income 120, D&A 90, total capex 140), free cash flow is 70.0 on every method, because FCF subtracts all capex and needs no estimate. Owner earnings, on the four triangulation methods, comes out at 97.5, 110.0, 120.0 and 160.2, a range of 64.3% on identical audited figures. That spread is what Buffett meant when he warned that owner earnings can't be computed to the decimal. It's also why the method belongs in valuation thinking and stays out of screens: at a 20× multiple those four estimates imply values from 1,950 to 3,204.

Now the limits. The working-capital noise above is symmetric and drift-free, independent across years too. All three assumptions flatter the arithmetic. Real working-capital errors trend and cluster. They also move with the business cycle. So a real company's single-year ratio behaves worse than this simulation, and a five-year cumulative window can itself fall inside one long deterioration. The Greenwald method assumes a stable PP&E-to-sales relationship and, without saying so, attributes inflation-driven revenue growth to real capacity. That biases maintenance capex down and owner earnings up, in the same direction as every other convenient assumption, and that's the direction to be suspicious of. And none of these three calculations detects fraud. They size how much of a signal is noise, which you need before the forensic work starts. They don't replace it.

Do it in code: compute both the single-year and cumulative conversion ratios for every company-year, simulate the company's own working-capital dispersion to set its threshold rather than importing 0.8, and carry maintenance capex as a vector of methods so owner earnings arrives as a range with its method labelled. See G1/QM1.01 and G2/DA1.05.

Where this goes next: galaxy cross-links

• G2 · DA1.02 Pandas Data Wrangling. The ten-year conversion panel with the convention restatement as a column, and the cash walk as a reconciliation that closes by assertion.
• G2 · DA1.05 Statistics in Code. Measuring a company's own working-capital dispersion, so the red-flag threshold is calibrated rather than inherited.
• G1 · QM1.01 Probability, Distributions & Estimation. Why cumulating five years cuts the noise by exactly √5, and what that does to the false-positive rate of any single-year screen.
• G2 · QD1.02 Signal & Factor Construction. Cumulative CFO/EBITDA and the accruals-adjacent measures as constructed quality factors, with the winsorisation and standardisation that stops three outliers from driving a screen.

Teach it back & journal

Feynman prompt. Explain to a smart 15-year-old how a company can report record profits every year and still run out of money, and why the word "free" in free cash flow needs air-quotes until you have subtracted what the business must spend just to stay alive. Use a pocket-money shop that sells on credit (profit vs cash), its delivery cycle that wears out (maintenance vs growth), and a three-line ten-year cash walk (what came in, what had to go back in, what the owner actually got). End with two sentences on what it means when a company pays its staff in slices of ownership and then calls the leftover cash "free." One page, no jargon you don't define.

Journal prompt. Take one company you already own or admire and build its ten-year cash walk tonight: screener.in or stockanalysis.com plus the filings, one hour. Log in your decision journal: the paise-per-rupee (cents-per-dollar) table; cumulative FCF/PAT; the personality you assigned and its standing follow-up question; and one sentence on whether the ledger (the walk) matches the letter (management's narrative), and if they diverge, what you now believe that you didn't yesterday. These rows carry straight into your deep-dive workbook later on.

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