Capstone Deep Dive: India

Phase 10 · Weeks 77–78 · ~40 hours

Where it sits. The second of the two independent capstone deep dives, and the twin of M10.01, which you completed last fortnight on a US company from the US approved list. That module proved you can run the full rapid-analysis system, certified in M8.05, unaided, at institutional grade, on a company you had never studied, and produce a thesis package a real investment committee would accept. Now you do it again on an Indian company, and it is not a mirror image. India adds a layer that the US capstone barely tested and that the Indian market makes non-negotiable: a full governance and promoter deep-read. In India, the question "is this management honest and aligned with me, a minority shareholder?" is not a soft coda to the analysis. For a large fraction of the market it is the analysis, because a great business run by an extractive promoter is a value trap wearing a compounder's clothes. The shared analytical spine of M10.01 carries over with the India layer bolted on, which makes the two capstones a matched pair, not a repeat. It is the last graded deliverable before the communication toolkit (M10.03) polishes your output and the final audit (M10.04) certifies core competence. Together, M10.01 and M10.02 are the two theses that must pass the institutional rubric to record the M11 milestone.

The one-sentence version. Run the complete deep dive on an unfamiliar Indian company to the same institutional bar as the US capstone (the ten-year history with incremental ROIC, the industry map, the four-lens moat proven in the numbers, the three-way valuation, and the capital-allocation read), and then do the thing that separates a competent Indian analyst from a naive one: read the promoter, the pledge, the related-party trail, the standalone-versus-consolidated gap, and the holding-company structure until you can defend, with evidence, whether the reported numbers can be trusted and whether the people who control this company are working for you or around you.

The company we carry. The worked walkthrough runs end-to-end on Marico Ltd (NSE: MARICO), a domestic promoter-family FMCG compounder (Parachute, Saffola), chosen deliberately: it is a good business run by a well-regarded promoter group, so the governance read resolves to "honest and aligned, with two things to monitor" rather than to a fraud. That is the harder and more common judgment, and the one the capstone must certify. Along the way it exercises every India-layer element: a promoter-holding and pledge check, a related-party read, a standalone-versus-consolidated reconciliation, and a mini sum-of-the-parts on its listed Bangladesh subsidiary. The governance section teaches the failure patterns you will not see at Marico by naming the real cases where they appeared. Every Marico figure here is illustrative, modeled on ~FY24/FY25 scale. Pull the actual annual report, results, and shareholding pattern before you rely on a number. Nothing here is investment advice; every company named is a teaching specimen.

Suggested hour budget (of the 40). Candidate selection + source-stack assembly 2h · Day 1 (history + incremental ROIC + notes en bloc) 8h · Day 2 (industry map, peers, model, reverse DCF) 8h · Day 3 (governance & promoter deep-read, scuttlebutt, calls, steelman) 8h · the SOTP/Ind AS/standalone-consolidated layer 4h · writing the thesis package (memo + governance dossier + exhibits) 6h · self-grading against the rubric + remediation 2h · flashcards, mastery check, teach-back, journal 2h. This is a two-week module; the calendar in the master map gives it weeks 71–72.

App link. This module drives the app's highest-tier flows: the independent case-engine chain on a real Indian company snapshot (history → ratios → moat → governance dossier → three-way valuation → thesis, rubric-graded); the governance-dossier builder (pledge %, RPT-as-percent-of-sales trend, LODR compliance checklist, standalone-vs-consolidated reconciler); the SOTP sandbox (value the parts, apply and justify a holdco discount); and the thesis grader that scores your package against the institutional rubric below and files it to your portfolio for the M10.03 review and the M10.04 audit.

Learning objectives

By the end you can:

1. Execute the full deep-dive system unaided, at institutional grade, on a never-studied Indian company: assemble the source stack from Indian primary sources, drive the three-day method from a blank workbook, and produce the complete thesis package to the M8.05 deep-dive bar.
2. Build the ten-year history with the India normalizations and compute incremental ROIC: apply the Ind AS 116 lease break (FY20), the FY20 corporate-tax-cut break, and exceptional/re-segmentation/acquisition adjustments so the series is comparable, then compute ΔNOPAT ÷ Δinvested capital and read the compounder test that headline ROCE hides.
3. Map the industry and prove or reject the moat in the numbers: run Porter plus Mauboussin's profit-pool and market-share-stability tests and all four moat lenses, confirming any moat claim in a persistent ROIC series rather than in adjectives.
4. Value the company three ways with Ind AS effects handled: intrinsic (FCFF/owner-earnings DCF), relative (multiples with India premium-multiple discipline), and expectations (reverse DCF), with internally consistent assumptions (g = RR × ROIC; terminal g ≤ nominal GDP) inside defensible bands.
5. Conduct the Governance & Promoter Deep-Read and deliver a defensible verdict: quantify promoter holding and pledge %; read related-party transactions from the notes, AOC-2, and the Board's Report against SEBI LODR Reg 23 materiality; check LODR compliance and Section 197(12) remuneration; reconcile standalone versus consolidated; and answer "is this management honest and aligned?" with evidence, recognizing the pledge, promoter-distributor-RPT, royalty-creep, and cash-upstreaming red-flag patterns.
6. Run the SOTP / holding-company-discount check where the target is a group company or holdco: value listed and unlisted stakes, apply and justify a holdco discount, and judge whether the structure creates or destroys minority value.
7. Write the institutional thesis package: a one-page thesis with a falsifiable variant perception, a supporting deep-dive memo, a governance dossier, clean exhibits, and a pre-mortem wired to pre-committed sell criteria.
8. Self-grade to the pass bar and diagnose your own misses: score the package against the capstone rubric, identify which criteria cost points, and remediate honestly.

Prerequisites & connections

Builds on. Everything, but most directly the Phase 8 signature skill you certified six weeks ago. M8.03: the ten-year history spreadsheet, the standard row template, the Ind AS 116 / FY20-tax normalizations, incremental ROIC, the industry map, the three-statement-lite model, and the reverse DCF are run here without scaffolding. M8.04: the Day-3 method (scuttlebutt via Fisher's fifteen points, the incentive read on the Indian Board's Report, the four earnings-call tells, the steelman, the one-page thesis, the pre-mortem) is the backbone of your governance and qualitative work; §4.4 of that module is where the Titan promoter/RPT read was first worked, and the capstone generalizes it. M8.05: the deep-dive rubric and its five hard gates (incremental ROIC, reverse DCF, steelman, one-page thesis, pre-mortem) are the floor this capstone builds on. M2.03 (NOPAT, invested capital, incremental ROIC), M2.08 (the forensic/quality-of-earnings toolkit and the Indian collapse cases: Satyam, IL&FS, DHFL, Yes Bank, Manpasand, Vakrangee), M3.07 (the five uses of capital and the proxy/board-report incentive framework), M3.10 (SOTP and holding-company discounts, banks and cyclicals valuation, IBC), and the Phase 5 sector playbooks (M5.01–5.10, the right lens, KPIs, and thresholds for whatever sector your candidate is in) are all assumed and used wholesale. M1.10 installed the anatomy you now read at speed: Ind AS and Schedule III, the auditor's report and CARO, BRSR, the shareholding pattern, and promoter-pledge disclosures. M6.05 gave you the written checklist you run before writing a word.

Feeds forward. M10.03 (the communication toolkit) takes this package as raw material: you will rewrite the thesis to professional standard and build the exec deck from it. M10.04 (the final audit) grades both capstone theses against the institutional rubric to record M11 — core competence, and re-opens the naive one-pager you sealed in week 3. Phase 9's process is where this thesis becomes a sized position with a journal entry and a calibration score. In the years after the program, this is the template every Indian deep dive you ever run will follow.

The competency it certifies. Primarily C10 — the signature skill, now demonstrated at institutional grade on the harder jurisdiction. But it is also the capstone home of the ethics & governance thread (assessed inside C2 — quality of earnings and C7 — management & capital allocation): the whole program has insisted that you end able to judge whether a management team is honest and shareholder-aligned, and in India that judgment carries more weight and needs more evidence than anywhere in the US market. And it carries C11 — communication: the package is graded partly on whether a decision-maker could act on it.

Core teaching content

4.1 What this capstone is, the institutional bar, and why India is the harder twin

The median large Indian listed company is controlled: a promoter family or group, not a dispersed base of institutions, decides what happens to it. The US capstone barely tested what that does to an analysis.

The two Phase-10 capstones are deliberately a matched pair with one asymmetry. Both demand the identical analytical spine, run unaided on a never-studied company and graded against an institutional rubric: a ten-year history with incremental ROIC, an industry map, a four-lens moat proven in the numbers, a three-way valuation, a capital-allocation read, a steelmanned bear case, a one-page thesis with a falsifiable variant perception, and a pre-mortem. If you passed M10.01, you already own that spine. Do not re-learn it; run it.

The asymmetry is the India layer, and it exists because the Indian market is structurally different from the US in four ways that change how you analyze it:

1. Ownership is concentrated in promoters. The median large Indian listed company is controlled: a promoter family or group typically holds anywhere from a third to two-thirds of the equity, versus the dispersed institutional ownership of the typical S&P 500 name. Control is not itself bad. A founder with 55% of the stock and a decades-long horizon can be the best possible steward (the benign case). But it means the central agency problem in India is not manager-versus-owner (the US textbook problem, solved with equity incentives); it is controlling-owner-versus-minority-owner, the risk that the promoter uses control to route value from the company (which you own a slice of) to entities the promoter owns entirely. That is a different problem, and it is read from different documents.
2. The disclosure surface is scattered and requires assembly. The US concentrates governance signal in one rich document, the DEF 14A proxy. India scatters the same truth across the Board's Report, the related-party (AOC-2) annexure, the corporate-governance report, the standalone-versus-consolidated statements, the shareholding pattern, and the pledge disclosures, each filed to BSE/NSE and MCA on its own schedule. You must gather and reconcile them.
3. Reported numbers require more distrust. India's forensic history (M2.08) is not ancient. Satyam (2009), the IL&FS and DHFL collapses (2018–19), Yes Bank, and a long tail of small- and mid-cap frauds are within a working analyst's memory. Auditor quality is more variable; promoter-controlled companies have more channels through which reported profit can be inflated or cash quietly extracted. "Over-trusting the reported number" is the single most expensive India mistake, and the capstone weights the forensic and governance read accordingly.
4. Structures are complicated on purpose. Holding companies, cross-holdings, listed subsidiaries, and group entities are pervasive, sometimes for legitimate historical reasons and sometimes to obscure. The standalone-versus-consolidated gap and the sum-of-the-parts / holdco-discount question are live for a large share of Indian names in a way they rarely are for a US single-entity operator.

So the institutional bar for M10.02 is the M8.05 deep-dive bar plus a governance and promoter dossier that could survive a hostile question from a portfolio manager who has been burned in India before. The order of work is unchanged (Day 1 machine, Day 2 industry and model, Day 3 qualitative and governance), but Day 3's governance block is expanded from a pass into a full deliverable.

A package that nails the DCF but hand-waves the related-party trail does not pass this capstone. In India, that package gets people's capital impaired.

4.2 The approved India candidate list

Pick one company you have never studied in the program: not in the starter bank you drilled from week 4, not the Titan/Asian Paints names worked to death in Phase 8, and not Marico (the worked walkthrough below). The list spans sectors on purpose, and it spans governance difficulty on purpose: some names are clean compounders where the governance read confirms a negative; a few are live cases where promoter, pledge, related-party, or group-structure questions are the center of the analysis, and where the whole India layer earns its weight. Choose by the rule the app enforces for certification: pick the one whose sector or governance texture you are weaker in, not the one you will find easy. Every note below is one line on what it tests; the figures and situations are as of the program's writing (mid-2026) and must be re-checked against current filings. Governance situations in particular evolve.

How to use the flags. If your honest self-assessment says governance is your weakest area (and for most learners leaving this program it is, because it is the least mechanical), deliberately pick a "Live" name. The clean compounders (Dabur, Infosys, Cholamandalam, Eicher, UltraTech) will teach you the disciplined confirm-the-negative read; the live cases (Nestlé's royalty, Kotak's cap, Sun's RPT history, Bajaj Finserv's holdco, Coal India's PSU dynamics, Vedanta's extraction, Adani's group questions, Zee's failure) will teach you what the red flags look like when they are real. A capstone on one of them, done honestly, is the strongest possible evidence for the M10.04 audit that your governance competency is genuine and not just a paragraph you know how to write. Whichever you pick, the deliverable is identical: the full package, with the governance dossier weighted as the rubric demands.

4.3 The deliverable spec: the shared analytical core (identical to M10.01)

The analytical spine is the M8.03–M8.05 deep dive run at the independent rung. It is unchanged from the US capstone; what follows is the checklist of what the package must contain, so you can drive it from a blank workbook. Each item is a heading in your deliverable.

A. The ten-year history spreadsheet, normalized. Build it from primary filings (annual reports FY16–FY25, quarterly results) using the M8.03 standard row template: segment revenue, the margin stack (gross → EBITDA → EBIT → PBT → PAT), ROIC/ROCE, invested capital (with and without goodwill), CFO, capex, FCF, net debt, share count, and dividends + buybacks. Apply the required India normalizations so the series is comparable end to end: the Ind AS 116 lease break at FY20 (EBITDA jumps when rent moves below the line, §4.5), the FY20 corporate-tax-cut break (the effective tax rate steps down when the company opts into the new ~25% regime, §4.5), exceptional items stripped, re-segmentation bridged, and acquisitions handled (ROIC with and without goodwill; split the incremental-ROIC window at the deal). For a bank, NBFC, or insurer, the operating/financing split fails. Use the M5.01–M5.03 templates instead (spreads, GNPA/credit cost, gearing, or VNB/EV).

B. Incremental ROIC: the compounder test. Compute ΔNOPAT ÷ Δinvested capital over multi-year windows and reconcile it to growth via g = RR × ROIC (RR = reinvestment rate). This is the M8.05 hard gate the headline ROCE hides: a company can show a gorgeous 40% headline ROCE on a legacy asset-light core while earning a mediocre return on the new capital it is deploying into acquisitions and new categories. The incremental number is the one that tells you whether reinvestment is compounding value or quietly destroying it.

C. The notes read en bloc. Revenue recognition, related-party, leases, contingent liabilities, segments, ESOP, and the auditor's report and CARO annexure: read together, findings surfaced. In India this read is also the first pass of the governance dossier (§4.4), so do not treat it as a formality.

D. The industry map. Porter's five forces (the real ones, not a recital) plus Mauboussin's supplier → firm → customer map, the profit pool, and the market-share-stability test (share moving more than ~2%/year signals a weak moat). Place your company's power in its chain.

E. The four-lens moat, proven in the numbers. Run all four lenses (Dorsey's five sources, Greenwald's barriers-to-entry, Helmer's 7 Powers, and Mauboussin's ROIC-persistence/CAP) and reject false moats (a hot product, high share, good execution, a charismatic promoter). Any moat claim must be confirmed in a persistent, high ROIC series across a cycle, or it is an adjective.

F. Peer benchmarking. Two or three peers on the same KPIs, normalized for accounting-policy and one-off differences (M2.05). India-specific: normalize for lease-accounting choices, different depreciation policies, and the standalone-vs-consolidated boundary before you compare anything.

G. The three-way valuation. Intrinsic (an FCFF or owner-earnings DCF with a currency-matched INR WACC, typically near the program's ~12% INR hurdle for a low-beta name), relative (multiples chosen by the sector lens, with India premium-multiple discipline), and expectations (the reverse DCF, the M8.05 hard gate, solving for the growth/margin the current price implies and judging it against base rates). Handle the Ind AS effects (§4.5) and, if the company is a group, the SOTP (§4.6).

H. The capital-allocation read. Where a decade of cash went across the five uses (reinvest, M&A, dividends, buybacks, debt paydown), and at roughly what incremental return. That is the M3.07 framework. In India this read blurs into governance: a promoter's capital-allocation choices are also incentive evidence (does the cash get reinvested at high returns, or does it drift toward promoter-adjacent entities and pet projects?).

Items A–H are the US capstone's deliverables too. Items in §4.4–§4.6 are the India layer that makes M10.02 distinct.

4.6 The India layer III: the SOTP / holding-company-discount check

Run this whenever your candidate is a holding company, a conglomerate, or an operating company with material listed or unlisted stakes (Bajaj Finserv, L&T, Sun via Taro, and the many pure holdcos: Bajaj Holdings, Tata Investment, Bombay Burmah, Maharashtra Scooters). A single blended multiple is meaningless for these; you must value the parts.

The method (M3.10, applied).

1. Identify the parts: each operating segment or subsidiary/associate, plus any pure investment stakes (listed and unlisted), plus net cash/debt at the parent.
2. Value each part on its own right lens: the sector multiple or a mini-DCF for operating businesses (a bank on P/B–ROE, an insurer on P/EV, an FMCG arm on EV/EBITDA), and market value for listed stakes (the cleanest input: Bajaj Finserv's stake in listed Bajaj Finance, Sun's stake in listed Taro, Marico's stake in listed Marico Bangladesh).
3. Sum to gross asset value, adjust for parent-level net debt and holdco costs, to get the SOTP/NAV.
4. Apply a holding-company discount, and justify it. Indian holdcos trade at 30–70% discounts to their SOTP/NAV, persistently. This is not irrational; the discount compensates for real frictions: minorities own the holdco, not the underlying cash flows, so they cannot direct capital allocation; dividends are taxed twice (once as they flow up to the holdco, again as the holdco pays out); the stakes are often illiquid or strategic and will never be sold; and capital-allocation uncertainty at the holdco level (will retained cash be reinvested well or trapped?). Your job is to estimate the appropriate discount from these drivers. A tightly-run holdco with a shareholder-friendly payout and a clear structure deserves a narrower discount than a sprawling one that hoards cash. Then ask the two edge questions: is the current market discount wider or narrower than justified (a wide-discount holdco can be a cheap way to own great underlying assets if a catalyst, a demerger or a buyback or a payout policy, might narrow it), and does the structure itself transfer value (a holdco that exists mainly to entrench promoter control at a discount to the assets is a governance issue, not just a valuation one).

The SOTP is where valuation and governance meet: a holdco discount is partly a price and partly a verdict on the structure. Say which part is which.

4.7 The thesis package: what you hand in

The deliverable is a package a real investment committee would accept, in four parts:

1. The one-page thesis (the M8.04 six-part structure): business · why it's a good business (moat + management/governance verdict) · the variant perception · the three things that must go right · valuation & expected return (a range) · sell criteria. Every spine question answerable from this page.
2. The supporting deep-dive memo (three to six pages): the ten-year history and incremental-ROIC exhibit, the industry map, the four-lens moat with its numbers-proof, the three-way valuation with the reverse-DCF read and the SOTP if applicable, the capital-allocation read, and the steelman + pre-mortem.
3. The one-page governance dossier (§4.4): the six checks, the red-flag scan, and the verdict, the deliverable that makes this the India capstone.
4. The exhibits: the history spreadsheet, the peer-comp table, the valuation output (base/bull/bear range), and, where relevant, the standalone-vs-consolidated reconciliation and the SOTP table.

Write it Minto-style (conclusion first; M10.03 will polish this), with every qualitative claim turned into a number (Damodaran's narrative-and-numbers discipline), and file it to your company folder and the app. This package, self-graded against the §8 rubric, is the graded output of the module and one of the two theses the M10.04 audit weighs.

Common mistakes & how experts think differently

Mistake 1: Over-trusting the reported number. The novice reads the Indian P&L the way they read a US one: as fact. The expert reads it as a claim made by a controlling shareholder, and asks, before trusting a single ratio, whether the forensic and governance checks support belief. In a market with India's fraud history (M2.08), the reported profit is a hypothesis until the cash-flow quality, the related-party trail, and the auditor's report corroborate it. This is not paranoia; it is the correct prior for a market where the person signing the numbers often controls the company. The reflex: quality of earnings and governance before valuation, always.

Mistake 2: Analyzing the wrong entity (standalone vs consolidated). Pulling ratios off the standalone statements when a quarter of the economics lives in subsidiaries, or off consolidated when the governance question is about cash trapped in a specific entity, produces confident nonsense. The expert always knows which entity earns the profit, holds the cash, and carries the debt, defaults to consolidated for the business analysis, and reconciles to standalone to read the structure and the governance.

Mistake 3: Treating a good reputation as a completed governance check. "It's a Tata / Birla / Bajaj company, governance is fine" is a prior, not evidence. The expert confirms the negative: pulls the shareholding pattern and reads the zero pledge, reads the RPT note, reads the auditor's report. The whole value of the check is that it is evidenced. Reputation tells you where to set your prior; the filing tells you the answer. (And the halo can be wrong: reputations were intact the day before several governance failures.)

Mistake 4: Missing the pledge, or reading it as merely a solvency issue. The novice skips the pledge line or, seeing a modest pledge, treats it as the promoter's private affair. The expert knows a rising promoter pledge is a triple signal: a leveraged promoter (forced-selling risk that can crater the price), a private incentive to prop the reported numbers, and usually evidence of a weaker venture elsewhere in the group being funded by this stock. The pledge check is three minutes and has saved more Indian capital than any DCF.

Mistake 5: Seeing an RPT and shrugging because "every company has them." True, and irrelevant. The signal is never existence; it is size, trend, and arm's-length-ness. The expert tracks the material RPTs as a percentage of sales over a decade and asks whether the number is small, stable, and fair — or a rising skim on a flat business. A brand fee flat at 0.5% of sales is fine; the same fee creeping to 2% is the thesis.

Mistake 6: Ignoring promoter incentives when reading capital allocation. The novice grades capital allocation on returns alone. The expert grades it on returns and on who benefits, because in a promoter-controlled company a value-destructive acquisition or a cash pile drifting toward group entities is bad allocation and, at the same time, an incentive revealed. "Show me the incentive and I'll show you the outcome" (M3.07) is doubly true where the allocator controls the board.

Mistake 7: Comparing multiples across the Ind AS 116 break, or trailing P/E on a cyclical/PSU. Two mechanical errors that produce wrong verdicts. Comparing a lease-heavy retailer's EV/EBITDA before and after FY20 (or against an owner-operator) without the lease adjustment overstates cheapness. And putting a trailing P/E on Coal India, Vedanta, or UltraTech at a commodity peak flashes "cheap" at exactly the wrong moment. That is the cyclical trap (M5.09): lowest P/E at the peak is a sell signal. The expert normalizes the accounting and the cycle before the multiple means anything.

Mistake 8: Valuing a holdco or conglomerate on a blended multiple, and ignoring the holdco discount. A single P/E on Bajaj Finserv or L&T is meaningless; the parts have different lenses and one may be a listed stake with an observable price. And the expert neither ignores the persistent 30–70% holdco discount nor treats it as free money: they estimate the justified discount from its drivers (double taxation, no control, illiquidity) and ask whether the market's discount is wider or narrower than that, and whether a catalyst might close it.

Mistake 9: Falling in love with a great business and under-weighting a governance veto. The hardest discipline. A company can have a wonderful moat, high ROIC, and a long runway, and still be un-ownable because a rising pledge or an escalating value-leaking RPT means you cannot trust that the returns will reach you. The expert holds two facts at once, "great business" and "pass", and knows that in India the second can override the first. Being right about the business is worthless if the governance loses the capital.

Mistake 10: Writing governance as a paragraph of adjectives instead of a dossier of evidence. "Management is high-quality and shareholder-friendly" is what the novice writes and the bull deck says. The expert writes a dossier: promoter holding 59%, stable; pledge 0%, confirmed; largest RPT 0.4% of sales, flat, arm's-length; pay in line with peers, below the 197 cap; auditor unmodified, no KAM alarm, no resignation; verdict aligned, monitor X and Y. The difference between the two is the difference between a feeling and a finding, and it is exactly what the capstone rubric scores.

Worked examples

Marico Ltd carries a single, fully-worked model execution here: the finished standard the capstone is graded against, run end-to-end. It is deliberately exemplary and long; read it as the answer key for your own package. All figures are illustrative, modeled on ~FY24/FY25 scale. Pull the actual filings before relying on any number. Marico is a teaching specimen, not a recommendation.

WE: Marico Ltd (NSE: MARICO): the full deep dive, carried through every step

The source stack (assembled first, ~90 min, before the clock). Annual reports FY16–FY25 (Marico investor-relations site and BSE/NSE); the last two years of quarterly results and the eight most recent earnings-call transcripts (screener.in hosts them); the shareholding pattern for the last twelve quarters (exchange filings / screener.in); the Board's Report, AOC-2 RPT annexure, and corporate-governance report from the latest AR; the Marico Bangladesh Ltd annual report and Dhaka Stock Exchange quote (for the listed-subsidiary carve-out); one competitor set (HUL, Dabur, Nestlé India, Britannia) and one bear note. Data pulled to a blank workbook; screeners for speed only, the filing for every load-bearing number.

The business in one line (Q1). Marico is a domestic-promoter FMCG company that makes money on branded coconut oil (Parachute), branded edible oils and foods (Saffola), and a spread of hair/skin/male-grooming brands. The real profit engine is Parachute, a low-growth but extraordinarily high-return, trusted cash cow whose profits fund the Saffola foods and digital-first (D2C) growth vectors. Roughly three-quarters of revenue is India; a quarter is international, of which Bangladesh (via a ~90%-owned, separately listed subsidiary) is by far the largest and most profitable.

Step 1: The ten-year history, normalized, and incremental ROIC

The spine of the whole dive. Illustrative consolidated series (₹ crore unless noted):

The required normalizations, flagged in the series:

• Ind AS 116 (FY20 break): Marico leases distribution and office space but is not lease-heavy like a retailer, so the EBITDA-margin step at FY20 is small (~0.3–0.5pp). Flag it, don't over-adjust. (Contrast: for a DMart or a Jubilant FoodWorks candidate this break would be large.)
• FY20 tax-cut break: Marico opted into the ~25% regime; the effective tax rate steps down at FY20/FY21, lifting PAT and headline ROCE for a fiscal reason. Use a consistent ~25% tax for the trend read.
• Acquisitions: FY21–FY25 saw a string of D2C bolt-ons (Beardo, Plix, Just Herbs, True Elements, and others). These inject goodwill and consideration that step up invested capital (the ~₹2,600 → ~₹4,600 cr jump is mostly this), without a proportionate near-term NOPAT. Compute ROIC with and without goodwill, and split the incremental-ROIC window at the deals.

Incremental ROIC: the compounder test (the hard gate). Read the two windows separately, because they tell opposite-feeling stories:

• FY16 → FY20 (the legacy core compounding): ΔNOPAT ≈ 1,000 − 700 = ₹300 cr; ΔInvested capital ≈ 2,600 − 2,000 = ₹600 cr. Incremental ROIC ≈ 300 / 600 ≈ 50%. The asset-light core (Parachute + Saffola) reinvested a modest amount of capital at a spectacular return, the signature of a genuine franchise.
• FY20 → FY25 (the diversification push): ΔNOPAT ≈ 1,500 − 1,000 = ₹500 cr; ΔInvested capital including acquisition goodwill ≈ 4,600 − 2,600 = ₹2,000 cr. Incremental ROIC ≈ 500 / 2,000 ≈ 25%.

Here is the insight the headline hides, and the one your reverse DCF must respect: headline ROCE is ~45% and rising, yet the incremental return on the last five years' capital is ~25%, half the core's, though still comfortably above the ~12% INR hurdle, so still value-accretive. The 45% headline is dominated by a legacy core that needs almost no capital. The marginal rupee, deployed into D2C acquisitions bought at high multiples and into lower-margin foods, earns meaningfully less. This is not a red flag (25% > 12% is good allocation), but it is the central analytical fact of Marico: the compounding math going forward depends on the return of the new capital, not the legacy 45%.

A valuation that extrapolates 45% ROCE onto Marico's future reinvestment is wrong.

Cash quality (a fast M2.04 pass): CFO/PAT runs consistently above ~1.0× and FCF conversion (FCF/PAT) is high (~90%): profits are real cash, not accruals. Working capital is tight (a branded FMCG with strong distribution). No divergence between profit and cash to worry about, one box the forensic read clears early.

Step 2: The industry map and the four-lens moat

Industry map (Porter + Mauboussin). Indian FMCG is a good-but-competitive profit pool: structurally attractive (huge, growing, staple demand; distribution as the barrier) but with intense rivalry among a handful of scaled players (HUL, Nestlé, Dabur, Britannia, ITC, Marico) and a rising tide of regional and D2C challengers eating niches. Five forces: rivalry high but rational among incumbents; new entrants low barrier to launch (D2C makes a brand easy to start) but very high barrier to scale (national distribution takes decades); buyer power low (fragmented consumers) but modern-trade and quick-commerce are raising channel power; supplier power the swing variable (copra for Parachute, edible oil for Saffola; input-cost volatility hits gross margin); substitutes moderate. Marico sits in the profit pool as a category leader in two niches (coconut oil, health-positioned edible oil) rather than a broad-line giant, a focused, high-return position.

Market-share stability test: Parachute's share of branded coconut oil is very high and very stable across a decade, the single strongest piece of moat evidence. Saffola's share of health-edible-oils is strong and stable. The D2C brands have no stable share yet — they are options, not moats.

The four lenses:

• Dorsey's five sources. The moat is intangibles/brand (Parachute = trusted purity in a category where adulteration is a real fear; Saffola = health) + cost/distribution advantage (one of India's deepest FMCG distribution networks, expensive and slow to replicate) + efficient scale in the coconut-oil niche (Parachute's dominance makes the niche unattractive to attack). Not network effects; modest switching costs (habit, not lock-in).
• Greenwald (barriers to entry). Collapse the forces to barriers: customer captivity (habit and trust on a low-ticket, frequently-repurchased product) + economies of scale in distribution + local/niche dominance. The share-stability test confirms the barrier in the core; it also honestly flags that the new categories have no barrier yet.
• Helmer's 7 Powers. Present in the core: Branding (Parachute/Saffola) and Scale Economies (distribution). Absent: network economies, counter-positioning (though the D2C brands attempt counter-positioning against the legacy model), cornered resource, process power. Each present power is "Benefit + Barrier"; each absent one is honestly marked absent.
• Mauboussin (ROIC persistence / CAP). The decisive numbers lens: a decade-plus of ~40%+ ROCE that has not mean-reverted is strong evidence of a durable moat with a long Competitive Advantage Period in the core. But the incremental-ROIC read (Step 1) sharpens it: the CAP is long and high for the legacy core; for the reinvestment, the durable-but-lower ~25% is the honest number.

Moat verdict: a real, durable, narrow-to-wide moat in the core (brand + distribution + niche dominance), proven in a persistent 40%+ ROCE; a set of unproven options (D2C, foods premiumization) around it. Not a false moat, because the ROIC series confirms it, but a moat whose width in the future depends on categories that do not yet have one.

Step 3: The capital-allocation read (M3.07)

Where a decade of cash went, and at what return:

• Reinvestment in the core: modest (the core is asset-light and barely needs capital), and at ~50% incremental returns, exactly right.
• Acquisitions (D2C bolt-ons): the growth strategy, buying digital-first brands (Beardo, Plix, Just Herbs, True Elements) to build a younger, premium, online-native portfolio. At ~25% incremental ROIC including goodwill, this is accretive but not spectacular, and it is the allocation choice to monitor: are these brands scaling into the distribution machine and earning their multiple, or is Marico paying up for growth that dilutes returns? The honest read today: reasonable, disciplined in size, but unproven, and the single most important capital-allocation question in the thesis.
• Dividends + buyback: high payout (~80–90%), the correct signal for a cash-cow core that cannot absorb its own cash flow at high returns; returning it beats hoarding it (the M3.07 discipline). Watch for tension: as the growth push needs funding, does the payout stay disciplined or does allocation get muddled?
• Balance sheet: near debt-free, net cash. Conservative, appropriate, and a governance positive (no leverage games).
• Share count: broadly flat, with modest ESOP dilution. The Fisher-13 "self-funded growth" positive, lightly qualified by the ESOP cost.

Capital-allocation grade: good, promoter-disciplined. It returns the core's cash, reinvests where returns justify it, and makes bounded (not bet-the-company) growth acquisitions. The one open question is the return on the acquisition/foods capital, which is the thesis's central uncertainty, not a governance flaw.

Step 4: The Governance & Promoter Deep-Read (the exemplary section)

The deliverable that makes M10.02 the India capstone is worked here to the finished standard. The output is the one-page dossier. Run the six checks, and notice that on Marico the value of the exercise is not catching a fraud (there is none) but earning the right to trust the numbers, and doing it with evidence rather than reputation.

Check 1: Promoter holding. The promoter is the Mariwala family, holding through family entities (historically ~59% of equity), a level that has been stable across the decade with no creeping dilution and no distress-driven selldown. Harsh Mariwala is non-executive Chairman; the company is run by professional management (a long-tenured, non-family MD & CEO). Reading: this is the benign concentrated-ownership case, a founding family with a large, long-held, aligned stake and per-share value as its dominant incentive, but which has professionalized management rather than installing family in the executive chair. The 59% means minorities cannot outvote the promoter, so protection rests on disclosure, the board, and LODR, all of which check out below. Dossier line: promoter (Mariwala family) ~59%, stable a decade, professionally managed; high aligned skin in the game.

Check 2: The pledge. Pull the shareholding pattern and read the encumbered line: pledge is nil. You confirm this from the filing; you do not infer it from Marico's reputation. Dossier line: promoter pledge 0%, confirmed from the shareholding pattern, not assumed. Why this matters even for a clearly-clean company: it is the three-minute check that, had it read "18% and rising," would have overridden everything above it, because a pledged promoter is leveraged to the stock, exposed to forced selling, and privately incentivized to prop the reported numbers. Reading zero here is not a formality; it is the removal of the single largest India-specific tail risk, and it belongs in the dossier as evidence.

Check 3: The related-party trail. Go to the RPT note, the AOC-2 annexure, and the corporate-governance report. What the read surfaces:

• No parent royalty or brand fee: Marico is a domestic promoter company, not an MNC subsidiary, so the single most extractable RPT (the royalty-to-parent that dominates the Nestlé/HUL read) is simply absent. That structural fact removes a whole category of leakage risk.
• Transactions with the listed Bangladesh subsidiary and other group entities: these are ordinary intra-group operating flows (the ~90%-owned Bangladesh sub, other international subs); consolidated away and, on inspection, at commercial terms. Track the material ones as a percentage of sales: small and stable (well under 1% of consolidated sales, no rising trend).
• Promoter-linked entities: the Mariwala family runs philanthropic/venture vehicles (e.g., the Marico Innovation Foundation, the Ascent Foundation); check for any material commercial RPT or CSR routing. The disclosures show these as immaterial and arm's-length, not a channel for value transfer.
• No promoter-distributor pattern, no inter-corporate loans/guarantees to promoter ventures, no bailout acquisitions of promoter assets: the four patterns that leak value are absent, and you say so having looked, not having assumed.
• LODR Reg 23 lens: no RPT approaches the materiality threshold (for a company this size, the first band: 10% of consolidated turnover) in a way that raises a flag; the routine approvals are in order. Dossier line: RPTs small (<1% of sales), stable, arm's-length; no royalty, no promoter-distributor, no group loans/guarantees, no bailout deals; no value-leakage pattern present.

Check 4: Remuneration and CARO. The Section 197(12) disclosures show professional-management pay at market levels (the CEO's package is equity-linked and peer-consistent, not racing profits toward the statutory cap); the promoter, being non-executive, does not draw an extractive salary. ESOP dilution is modest and disclosed, a real but small cost to flag, not a red flag. The auditor's report is unmodified, with no going-concern or qualification, no alarming key-audit-matter, and no mid-tenure auditor resignation; the CARO annexure is clean on related-party loans, defaults, and fraud. Dossier line: pay market-level and equity-aligned, below the 197 cap; auditor unmodified, no KAM alarm, no resignation; CARO clean; modest disclosed ESOP dilution.

Check 5: LODR compliance and board quality. The corporate-governance report shows an adequately independent board (with a non-executive chair, the composition satisfies Reg 17), a functioning audit committee (Reg 18), and proper material-subsidiary governance for the Bangladesh entity (Reg 24). Reading the substance, not the tick-boxes: the independent directors are of genuine standing, director/CFO churn has been normal, and material events are disclosed on time. Proxy advisers (IiAS, SES) have not flagged Marico's resolutions as governance concerns. Dossier line: LODR-compliant in substance; genuine independent board; no proxy-adviser red flags.

Check 6: The earnings-call candor cross-check (from Day 3, M8.04 §4.5). Across eight quarters spanning a copra/edible-oil input-cost spike (a genuine bad stretch for gross margin) and stronger periods, management named the mechanism specifically ("gross margin compressed on the copra spike; here is the pricing action and the expected recovery"), owned the soft patches in the first person, kept its multi-year guidance on the foods/premiumization trajectory broadly, and narrated the same core metrics (India volume growth, Saffola foods scale-up, gross margin, the D2C portfolio's progress) through good and bad quarters rather than rotating to whatever was flattering. Candor: strong.

Check 6: The verdict. Synthesizing: Governance verdict: ALIGNED AND HONEST. High, stable, aligned promoter stake; zero pledge (confirmed); no value-leaking RPT patterns; professional management at market pay; clean auditor; genuine independent board; strong management candor. You can trust the reported numbers and own the business on its merits. Two monitoring items (not flaws, but the things a good analyst watches): (a) capital allocation on the D2C acquisitions: does the ~25% incremental ROIC hold, or do future deals dilute returns?; (b) the payout-vs-reinvestment balance as the growth push scales: does the disciplined high payout persist, and is the Bangladesh sub's cash (a ~90%-owned, FX-exposed, separately-listed entity; Step 5) repatriated efficiently to minorities? These two become explicit break-variables in the thesis.

Notice what the dossier did. It converted "Marico is well-governed" (a reputation) into a finding backed by six evidenced checks, and it did so with the same six checks that, run on Vedanta or Zee, would have surfaced upstreaming, pledges, and RPT leakage. That symmetry is the point: the governance read is a single disciplined instrument that confirms a clean company and catches a dirty one, and the capstone certifies that you can run it either way.

Step 5: Ind AS, standalone vs consolidated, and the mini-SOTP

Standalone vs consolidated. Marico standalone is essentially the India business; consolidated adds the international operations: mainly Bangladesh (the ~90%-owned listed subsidiary), plus Vietnam, MENA, and South Africa. The gap is ~25% of revenue and a higher share of profit, because the Bangladesh business runs very high margins and ROCE. Three analytical consequences you must handle:

1. Do the business analysis on consolidated. The moat, ROIC, and growth read above are all consolidated, correctly. The Bangladesh profit is real and high-quality.
2. Discount the subsidiary profit slightly in your mind. Profit earned in a ~90%-owned foreign, separately-listed subsidiary is worth marginally less to a Marico minority than profit in the wholly-owned parent: there is a ~10% non-controlling interest taking its slice, FX exposure (BDT/INR), and repatriation friction (dividends must be upstreamed across a border and a minority). This is a mild quality qualifier, not a red flag, but a rigorous analyst notes it.
3. The consolidation is clean (subsidiaries line-by-line with the NCI shown; no equity-method opacity hiding the international economics), which is itself a governance positive.

The mini-SOTP (because part of Marico is a listed stake). Marico is not a holdco, but it contains a listed subsidiary whose market price you can read directly, a clean and small SOTP exercise:

• Marico's ~90% stake in Marico Bangladesh Ltd has an observable market value on the Dhaka Stock Exchange, illustratively on the order of ₹8,000–10,000 crore (pull the DSE quote and the BDT/INR rate; check).
• Carve that listed stake out of Marico's ~₹84,000 crore market cap (illustrative), and the remaining ~₹74,000–76,000 crore of value is attributed to the India-plus-rest business, which, against that business's share of EBITDA, trades at an even richer implied multiple than the blended number suggests.
• The SOTP read therefore reinforces the valuation conclusion (Step 6): once you price the visible Bangladesh stake at market, the market is paying an extremely full multiple for the India franchise. It also frames a governance/structure nuance you disclose rather than bury: a slice of your Marico value is a foreign-listed, minority-partnered entity.

This is the disciplined habit even where a full holdco-discount analysis is not needed: know which parts are separately valued in the market, carve them out, and see what multiple the market is really paying for the core.

Step 6: The three-way valuation

Cost of capital (INR). A low-beta staple: illustrative INR risk-free ~6.75% (10-yr G-sec; pull the live figure), an India equity-risk-premium of ~7.5% (mature-market ERP ~4.3% + India country-risk premium ~3.2%, Damodaran method; pull the January update), a levered beta ~0.7. Cost of equity ≈ 6.75% + 0.7 × 7.5% ≈ ~12%; near-zero debt, so WACC ≈ 12%, squarely the program's INR hurdle.

(1) Intrinsic: owner-earnings / FCFF DCF. Start from ~₹1,400 cr FCF (FY25), grow near-term at a fading rate (mid-teens tapering as the base scales and the foods/D2C engine matures), and, critically, hold reinvestment consistent with g = RR × ROIC at the honest ~25% incremental ROIC, not the 45% headline. Terminal growth 6% (below India's ~10–11% nominal GDP, appropriate for a mature FMCG). A two-stage model on these inputs yields an illustrative intrinsic value in a band roughly ₹520–620 per share (base case), with the width driven by the incremental-ROIC and margin assumptions. The point is the range, not a false-precision point.

(2) Relative: multiples with India premium discipline. At an illustrative ~₹650 price and ~130 cr shares (₹84,000 cr market cap): P/E ≈ 54× (on ~₹1,550 cr PAT), EV/EBITDA ≈ 40× (EV ~₹83,500 cr net of cash / ~₹2,100 cr EBITDA), FCF yield ≈ 1.7%. Benchmarked against the premium FMCG cohort (Nestlé India ~65×, HUL ~55×, Britannia ~50×, Dabur ~45×), Marico sits mid-cohort: richly valued, but not the most expensive, and India's structural premium-multiple regime (M3.06) means a top-quartile FMCG franchise does deserve a premium to the market. The relative read says "expensive, consistent with quality," not "cheap."

(3) Expectations: the reverse DCF (the hard gate). Invert the price. Fast Gordon inversion for a mature compounder: implied perpetual FCF growth ≈ required return − FCF yield ≈ 12% − 1.7% ≈ 10.3%. So the price implies Marico grows free cash flow at roughly 10%+ forever, about India's nominal GDP rate, at a 12% discount rate. Judge it: for a company whose core (Parachute) grows low-single-digit volume, ~10% perpetual FCF growth requires the foods, premiumization, and international engines to carry mid-teens growth for a long time and the incremental ROIC to stay well above the hurdle. Against base rates (only a small minority of firms compound at ~10%+ real-plus-inflation for decades), this is demanding but not implausible for a top-quartile FMCG franchise. It is precisely a bet that the diversification succeeds. The margin of safety at the illustrative ₹650 price is thin.

Reconciling the three: intrinsic (~₹520–620) sits below the ~₹650 price; relative says "full, quality-consistent"; the reverse DCF says "priced for successful diversification with little room for error." The three agree on what the method shows on these FY24/FY25-scale figures: a genuinely great, well-governed business trading close to (or slightly above) what a great business is worth. This is a quality-at-a-full-price situation, not a mispricing. The honest output is a watch-and-wait, with a defined price at which the margin of safety opens. That is the method's output on stated, illustrative figures, not a view on the stock today.

Step 7: The one-page thesis and the pre-mortem

The one-page thesis (the deliverable):

MARICO LTD (NSE: MARICO) — one-page thesis · [date] · price ~₹650 · data through FY25 AR + Q[x] concall · sources: annual reports FY16–FY25, screener.in, 8-quarter transcripts, shareholding pattern (12q), AOC-2/RPT annexures, Marico Bangladesh AR + DSE quote, HUL/Dabur/Nestlé/Britannia comps 1. The business. Marico is a domestic-promoter FMCG franchise; the profit engine is Parachute (branded coconut oil: a low-growth, ~40%+-ROCE, trust-based cash cow), whose cash funds Saffola (health-edible-oils → foods) and a portfolio of hair/skin/male-grooming and D2C brands. ~75% India, ~25% international (mostly a ~90%-owned, listed Bangladesh subsidiary). It makes money on brand-and-distribution margin, not on the commodity oil. 2. Why it's a good business. Durable moat (brand + one of India's deepest distribution networks + niche dominance), proven in a decade-plus of ~40%+ ROCE that has not mean-reverted. Near debt-free, high cash conversion. Governance: ALIGNED AND HONEST. Promoter (Mariwala) ~59% and stable, pledge 0% (confirmed), professional management at market pay, no royalty / no value-leaking RPT patterns, clean auditor, strong management candor. This is a business you can trust the numbers of. 3. The variant perception. The price implies ~10% FCF growth in perpetuity: the market is pricing the diversification to succeed. Two honest variant reads are possible, and the analyst must pick one. The bull variant: India's premium personal-care and foods runway plus the D2C portfolio give the reinvestment a longer, higher-return life than the market's fade allows. The skeptical variant (the one this thesis takes): the ~25% incremental ROIC (vs the 45% headline) and a low-growth core mean the price already fully credits a diversification that is unproven, leaving no margin of safety. The variant is about runway and the return on new capital, not about quality; everyone agrees it's a great business. 4. The three things that must go right (to justify the price). (i) The D2C/foods reinvestment keeps earning ~25%+ incremental ROIC as it scales (the central bet); (ii) the core (Parachute/Saffola) holds share and pricing through input-cost cycles without volume erosion; (iii) premiumization and international (Bangladesh + new markets) sustain double-digit growth long enough to validate a ~10% perpetual-growth price. 5. Valuation & expected return. Intrinsic ~₹520–620 (base, on honest ~25% incremental ROIC and 6% terminal); relative ~54× P/E (full but quality-consistent vs a 45–65× premium cohort); reverse DCF implies ~10% perpetual growth (demanding). On the illustrative figures the ~₹650 price sits above the base band, so the margin of safety is thin; the thesis records that reading with its date and names what would move the base case, not a price to act at. 6. Sell criteria (pre-committed). Exit / do not enter if: (a) incremental ROIC on the reinvestment falls below the ~12% hurdle for a sustained period (bad growth: runway exhausted or acquisitions overpaid); (b) the core loses branded share or pricing power (Parachute/Saffola volume + share erosion for consecutive years: moat eroding); (c) a governance break: a promoter pledge appears, an RPT or royalty-type leakage emerges, or auditor/board quality deteriorates (a Fisher-15 veto); (d) price rises so far above intrinsic value that expected return no longer compensates (a valuation trim; this is a compounder, so trim, don't necessarily exit, per M9.03).

The pre-mortem ("it is three years later; Marico was a disaster — what happened?"), inverted to a monitoring dashboard:

The right-hand column is the monitoring dashboard for the permanent Marico file, mapping one-to-one onto the sell criteria. Note the honest classification the pre-mortem forces: four of the five disaster paths are business/governance risks with clear triggers; the fifth is purely a valuation risk, which is why the verdict is "watch and wait for a better price," not "avoid a bad company."

Step 8: Scenario, sensitivity, and the probability-weighted value (the intrinsic range made explicit)

Step 6 quoted a base intrinsic band of ₹520–620 and inverted the price with a fast Gordon perpetuity. Making that same engine explicit and running it forward as a scenario-and-sensitivity answers the question a hostile PM asks after the point estimate: "which assumption is doing the work, and what do you actually expect?" On a compounder yielding only ~1.7% of FCF the answer is dominated by one number, the perpetual FCF-growth rate, and honesty means showing how violently the value swings on it.

Owner-earnings per share, FY25 (illustrative): FCF ~₹1,400 cr ÷ ~130 cr shares = ₹10.77/share. On the single-stage Gordon lens, value/share = FCF/share × (1 + g) ÷ (WACC − g), with WACC = 12% (Step 6's near-debt-free INR hurdle).

Sensitivity grid (₹/share):

Read three scenarios off the WACC = 12% column, each tied to a business story rather than an arbitrary ±:

• Bear: g = 7.0% → ₹230. The diversification bet fails: D2C never scales into the distribution machine, foods stays mid-margin, and Marico reverts to a Parachute-plus-Saffola cash cow growing only a little above the ~6% terminal rate.
• Base: g = 9.9% → ₹564. Modest, disciplined success of the new engines at a still-above-hurdle incremental ROIC, and the value lands squarely inside Step 6's illustrative ₹520–620 band, the cross-check that this transparent lens is calibrated to the fuller two-stage model.
• Bull: g = 10.8% → ₹994. Foods, premiumization, and international carry double-digit growth for a long time and the ~25% incremental ROIC holds: a long, high-return runway.

The market price of ~₹650 corresponds to g ≈ 10.2% (the ₹659 cell), between base and bull, precisely the "priced for successful diversification" reading of Step 6's reverse DCF.

Probability-weight it. Assign 30% bear / 45% base / 25% bull: 0.30 × 230 + 0.45 × 564 + 0.25 × 994 = ₹571/share. The expected value ₹571 sits ~₹79 below the ₹650 price on these illustrative weights. That is the arithmetic under Step 7's valuation reading, and the check it points to is the weights themselves: which evidence would move probability between bear and base. And the grid delivers the deeper lesson: a ±2 pp move in g swings value from ₹230 to ₹994, so on this stock a point-estimate forward DCF is false precision. The reverse DCF (what is the price implying?) is the more disciplined instrument.

fcf_ps = 1400/130                      # = 10.769  owner-earnings per share (illustrative)
r = 0.12
gordon = lambda g, rr=r: fcf_ps*(1+g)/(rr-g)
grid = {g: {rr: round(gordon(g, rr)) for rr in (0.115, 0.12, 0.125)}
        for g in (0.07, 0.08, 0.09, 0.099, 0.102, 0.108)}
bear, base, bull = round(gordon(0.07)), round(gordon(0.099)), round(gordon(0.108))
price_cell = round(gordon(0.102))
ev = 0.30*gordon(0.07) + 0.45*gordon(0.099) + 0.25*gordon(0.108)
print(round(fcf_ps, 3))                # 10.769
print(grid)                            # matches the sensitivity table cell-for-cell
print(bear, base, price_cell, bull)    # 230 564 659 994
print(round(ev), 650-round(ev))        # 571 ; 79 below the price

Step 9: The valuation cross-check and the bear/bull bridge (triangulate before you commit)

Never trust one method. This step does the two things a committee expects: it triangulates the three valuation lenses against each other from the module's own Step-6 inputs (the internal-consistency check that ties the whole dive together), and it bridges the bear-to-bull range into its driver, so the variant perception becomes a decomposition rather than an adjective.

(A) Triangulation: recompute each lens from one set of illustrative inputs and confirm they agree. Using market cap ~₹84,000 cr, PAT ~₹1,550 cr, EBITDA ~₹2,100 cr, net cash ~₹500 cr, FCF ~₹1,400 cr, WACC 12% (all illustrative; check against live filings):

• FCF yield = 1,400 ÷ 84,000 = 1.67%, matching Step 6's ~1.7%.
• P/E = 84,000 ÷ 1,550 = 54.2×, matching Step 6's ~54×.
• EV/EBITDA = (84,000 − 500) ÷ 2,100 = 83,500 ÷ 2,100 = 39.8×, matching Step 6's ~40×.
• Reverse-DCF implied perpetual growth = WACC − FCF yield = 12% − 1.67% = 10.3% (fast); the exact Gordon solve for a ₹650 price gives 10.2%, matching Step 6's ~10.3%.

All four independently recomputed figures reproduce Step 6's stated numbers, so the three lenses tell one coherent story: a top-quartile FMCG franchise at a full multiple that already embeds ~10% growth forever. The discipline for your own candidate: when the lenses disagree (a relative multiple whispering "cheap" while the reverse DCF shouts "priced for perfection"), do not average them; hunt the cause, usually a mis-set WACC, a peak-cycle earnings base (the cyclical trap), or an un-adjusted Ind AS 116 lease / FY20 tax break (Steps 1 and 5).

(B) The bear/bull bridge. On the Step-8 Gordon engine at WACC = 12%, walk the per-share value from bear to bull as a waterfall, where every step is a change in the perpetual-growth assumption alone:

The full bear-to-bull swing is +₹764 on a 3.8 pp move in g, the numeric face of Step 7's variant perception: the disagreement is entirely about runway and the return on the new capital, not about quality. The bridge also localizes today's decision to a single number: the market (₹650) is paying ~₹95/share more than your honest base (₹564), a small, specific bet on diversification you can accept or decline explicitly. And it is exactly that ₹95 (and much more) that a governance break would erase before any growth argument begins. A rising pledge, a value-leaking RPT, or the Bangladesh-cash-repatriation risk of Step 4 caps the multiple regardless of g, which is why the governance verdict is a veto over this valuation rather than an input to it.

mktcap, PAT, EBITDA, netcash, FCF, r, price = 84000, 1550, 2100, 500, 1400, 0.12, 650
fcf_ps = FCF/130
print(round(FCF/mktcap*100, 2))                       # 1.67  FCF yield %
print(round(mktcap/PAT, 1))                            # 54.2  P/E
print(round((mktcap-netcash)/EBITDA, 1))               # 39.8  EV/EBITDA
print(round((r - FCF/mktcap)*100, 2))                  # 10.33 implied g (fast)
g_exact = (price*r - fcf_ps)/(price + fcf_ps)
print(round(g_exact*100, 1))                           # 10.2  implied g (exact Gordon @650)
gordon = lambda g: fcf_ps*(1+g)/(r-g)
vals = [round(gordon(g)) for g in (0.07, 0.099, 0.102, 0.108)]
print(vals)                                            # [230, 564, 659, 994]
print([vals[i+1]-vals[i] for i in range(3)], vals[-1]-vals[0])  # [334, 95, 335] 764

What this worked example demonstrates for your own capstone. It carried a single Indian company through every required step, and it treated the governance read as a dossier of evidence rather than a paragraph of praise, confirming a clean company with the same instrument that would catch a dirty one. It surfaced the one analytical fact the headline hides (incremental vs headline ROIC), handled the Ind AS and standalone-vs-consolidated realities, carved out a listed subsidiary in a mini-SOTP, valued the business three ways to a range, and produced a falsifiable one-page thesis with a monitorable pre-mortem. That is the finished standard. Your candidate will differ, and a bank, a cyclical or a live-governance name will stress different parts, but the shape of the finished package is exactly this.

WE2: Sum-of-the-parts and the holding-company discount (India, ₹)

Synthetic. Bharat Holdings Ltd and every figure below are invented case data, not a real company. Reason against the numbers as given; every figure was recomputed in Python for this module (snippet at the end).

A holding company cannot be valued on a single P/E, because its earnings are a blend of stakes it does not fully own. You value the parts, sum them to a net asset value (NAV), and then ask the only question that matters for the wrapper: what discount is the market putting on the holdco, and is there a catalyst to narrow it?

Bharat Holdings Ltd owns three assets and holds net cash:

• 52% of listed Op-Co A (market cap ₹40,000 cr). Mark a listed stake to market: 0.52 × 40,000 = ₹20,800 cr.
• 40% of listed Op-Co B (market cap ₹15,000 cr): 0.40 × 15,000 = ₹6,000 cr.
• 100% of an unlisted cash-cow subsidiary. No market price, so value it on a multiple: at EV/EBITDA 8× on ₹500 cr EBITDA = ₹4,000 cr (8× is a mature, cash-generative business held a touch below its listed peers for illiquidity; state and defend the multiple).
• Net cash at the holdco: ₹1,200 cr (add it).

NAV = 20,800 + 6,000 + 4,000 + 1,200 = ₹32,000 cr. Bharat's own market cap is ₹19,200 cr, so the holding-company discount = 1 − 19,200 / 32,000 = 40.0%.

The market values Bharat 40% below the sum of what it owns. Holdcos trade at a discount for real reasons: tax leakage (unlocking value by selling a stake triggers capital-gains tax the look-through NAV ignores); capital-allocation drag (minorities don't control when or whether value is unlocked, and cash can be redeployed at the promoter's discretion); no look-through control premium; and duplicated holdco-level costs. What narrows it: a buyback, a demerger or separate listing of the unlisted arm, a higher dividend pass-through, or an explicit holdco capital-return policy.

The analyst's discipline: a 40% discount is both a possible opportunity and a possible trap. Write it in the thesis with a benchmark and a trigger: not "buy, 40% discount," but "40% discount vs a peer-holdco median of ~25–30%, justified narrowing only if catalyst X (e.g., the announced subsidiary listing) actually completes." No catalyst, no re-rating: the discount is just the price of the drag.

stakes = 0.52*40000 + 0.40*15000 + 8*500 + 1200
disc = 1 - 19200/stakes
print(round(stakes), round(disc*100, 1))   # 32000  40.0

WE3: Ind AS 116 and why the "same" company looks more levered (India, ₹)

Synthetic. Kirana Retail Ltd and every figure are invented case data. Every number was recomputed in Python for this module (snippet at the end); the depreciation and interest are illustrative under the stated assumptions.

A change in accounting, with zero change in economics, can move a leverage ratio by 3×. If you compare a post–Ind AS 116 Indian retailer to a pre-116 comparable, or to a foreign peer on a different basis, without adjusting, you will mis-rank their balance-sheet risk. This is the capstone's "Ind AS specifics that change the numbers" layer made concrete.

Kirana Retail Ltd runs a large leased store base.

Pre–Ind AS 116 (operating-lease view). Store rent of ₹500 cr/yr is an operating expense, so reported EBITDA is ₹800 cr (already after rent). Leases sit off the balance sheet, so reported net debt is ₹1,000 cr. Net Debt/EBITDA = 1,000 / 800 = 1.25×, and the store leases are invisible.

Under Ind AS 116 (capitalise the leases). Recognise a lease liability = PV of committed lease payments ≈ ₹4,000 cr (assume a ~10-year weighted lease life and a ~7% discount rate) and a matching right-of-use (ROU) asset ≈ ₹4,000 cr. The ₹500 cr rent leaves operating expense, so EBITDA rises to 800 + 500 = ₹1,300 cr, replaced below EBITDA by ROU depreciation ≈ 4,000 / 10 = ₹400 cr and lease interest ≈ 7% × 4,000 = ₹280 cr in year 1. Lease-adjusted net debt = 1,000 + 4,000 = ₹5,000 cr, so Net Debt/EBITDA = 5,000 / 1,300 = 3.85×.

The same economic business went from 1.25× to 3.85×, a 3× jump on presentation alone. Two second-order effects worth naming: EBIT rises by ₹100 cr (₹500 cr rent out, ₹400 cr depreciation in), but year-1 PBT falls by ₹180 cr because the ₹280 cr of front-loaded lease interest exceeds that ₹100 cr. That is the classic Ind AS 116 earnings front-loading, worse early and better late.

The fix is non-negotiable: put every company on one basis before you compare leverage or EV/EBITDA. Either lease-adjust everyone's net debt and multiples, or strip leases out for all of them, but never compare a lease-capitalised Indian retailer to an off-balance-sheet or pre-116 peer at face value. Getting this wrong is how an analyst "discovers" a leverage problem that is really just an accounting standard.

ebitda_post = 800 + 500
pre = 1000/800
post = (1000+4000)/ebitda_post
dep, interest = 4000/10, 0.07*4000
print(round(pre,2), round(post,2), dep, interest, (500-dep)-interest)  # 1.25 3.85 400.0 280.0 -180.0

Practice set

Eleven drills that build the India-specific components of the capstone, ramped guided → independent → timed. Each is followed immediately by its complete solution. Mini-companies with clean numbers are synthetic (labeled); patterns are drawn from real situations. Write your answer in full before reading the solution.

Problem 1 (guided): Read the pledge. Synthetic. "Girish Industries" has 100 crore total shares. The promoter group holds 62 crore shares; of those, 27.9 crore are pledged. (a) Promoter holding %? (b) Pledge as a % of the promoter holding, and as a % of total shares? (c) One-line interpretation and the action it triggers.

Solution. (a) 62 / 100 = 62% promoter holding. (b) Pledge as % of promoter = 27.9 / 62 = 45%; as % of total = 27.9 / 100 = 27.9%. (c) A 45%-pledged promoter is a loud alarm: the promoter is heavily leveraged against the stock, exposing minorities to forced selling if lenders invoke, and has a private incentive to prop the reported price. Action: the pledge is a governance flag that must be explained before any valuation is trusted. Trace why the promoter borrowed (usually a weaker group venture), size the invocation risk, and treat the reported numbers with elevated distrust. Absent a fully satisfactory explanation, this alone can move the verdict toward pass, un-ownable.

Problem 2 (guided): Classify the related-party trail. Synthetic. "Alpha Consumer Ltd" (consolidated sales ₹5,000 cr) discloses these RPTs for the year: sales of goods through a promoter-owned distributor ₹400 cr; a brand/royalty fee to the promoter ₹30 cr; an inter-corporate loan to a promoter-group entity ₹150 cr; office rent to a promoter entity ₹5 cr. Classify each and compute it as a % of sales. Which is the biggest concern and why?

Solution.

• Sales through the promoter distributor: ₹400 cr = 8.0% of sales. Watch closely. Not automatically bad, but a promoter-owned entity inserted in the sales chain can skim a margin the company could keep (the promoter-distributor pattern). Check: is the distributor's margin arm's-length and disclosed? Is the arrangement necessary or a rent-extraction?
• Brand/royalty fee: ₹30 cr = 0.6% of sales. Small; track the trend. Fine if stable; a concern if creeping up on flat performance (royalty creep).
• Inter-corporate loan to a promoter entity: ₹150 cr = 3.0% of sales. The biggest concern. This is cash leaving the balance sheet to fund a promoter venture, the classic pre-collapse pattern (IL&FS/DHFL-style). Even at 3% of sales it is a governance red flag because it is value leaving, not a commercial transaction; check terms, security, and whether it is recurring/growing.
• Rent: ₹5 cr = 0.1% of sales. Immaterial; note and move on.

Biggest concern: the inter-corporate loan. RPT sales at least involve the company's product, but a loan to a promoter entity is pure capital diversion with no operating rationale. The distributor arrangement is second (size + margin-skim risk). The discipline: size, trend, arm's-length-ness, not mere existence.

Problem 3 (guided): Reconcile standalone vs consolidated. Synthetic. "Beta Ltd" reports standalone PAT ₹800 cr and consolidated PAT ₹1,160 cr. The consolidated figure includes a 70%-owned subsidiary earning ₹300 cr PAT and an associate contributing ₹60 cr via the equity method; there is no other difference. (a) What is the non-controlling interest's share of the subsidiary's profit? (b) Roughly where does the profit "live," and which statements do you analyze for the business and why? (c) One governance question the gap raises.

Solution. (a) NCI share = 30% × ₹300 cr = ₹90 cr (so ₹210 cr of the subsidiary's profit is attributable to Beta's owners). (b) Standalone (₹800 cr) is essentially the parent's own operations; the extra ~₹360 cr of consolidated profit comes from the subsidiary (₹300 cr, of which ₹210 cr is yours after NCI) and the associate (₹60 cr, one line). Analyze consolidated for the business, which shows the whole economic entity, but reconcile to standalone to see where the profit and cash actually sit. (c) The gap means a meaningful chunk of "your" profit lives in a 70%-owned subsidiary (a 30% minority takes a slice, and dividends must be upstreamed) and in an associate you do not control (you get one profit line but cannot direct its cash). Governance question: can that subsidiary/associate cash actually reach Beta's minority shareholders, or is it trapped by the minority partner, by the lack of control, or by the promoter's structuring?

Problem 4 (independent): The Ind AS 116 EBITDA trap. Synthetic. A lease-heavy retailer reported FY19 EBITDA of ₹500 cr (pre-Ind AS 116; ₹120 cr of store rent sat in operating expenses) and FY20 EBITDA of ₹650 cr (post-Ind AS 116; rent now sits below EBITDA as depreciation + interest). Underlying operations were roughly flat. (a) How much of the ₹150 cr "EBITDA growth" is real? (b) What must you do to EV before using EV/EBITDA on FY20?

Solution. (a) Put FY19 on the new basis: FY19 EBITDA would be ~₹500 + ₹120 = ~₹620 cr comparable. So FY20's ₹650 cr vs ₹620 cr comparable is ~₹30 cr (~5%) of real growth; the other ~₹120 cr is pure accounting (rent reclassified below the line). Reporting "EBITDA grew 30%" would be nonsense. (b) Post-Ind AS 116, the balance sheet carries a lease liability; to compare like-for-like you must add the lease liability to EV (treat it as debt), because otherwise EV/EBITDA looks artificially low: EBITDA rose but the (lease) debt that produced it was ignored. Equivalently, use pre-lease "EBITDAR" consistently on both sides. Never compare a lease-heavy Indian company's post-FY20 EV/EBITDA to its pre-FY20 self, or to an owner-operator, without this adjustment.

Problem 5 (independent): Normalize the FY20 tax break. Synthetic. A company earned EBIT of ₹1,000 cr in both FY19 and FY20 on invested capital of ₹3,000 cr. It paid a 34% effective tax rate in FY19 and, opting into the new regime, ~25% in FY20. (a) Compute NOPAT and ROIC each year. (b) How much of the ROIC "improvement" is operational?

Solution. (a) FY19 NOPAT = 1,000 × (1 − 0.34) = ₹660 cr; ROIC = 660 / 3,000 = 22.0%. FY20 NOPAT = 1,000 × (1 − 0.25) = ₹750 cr; ROIC = 750 / 3,000 = 25.0%. (b) None of it. EBIT and invested capital were identical; the entire 3-percentage-point ROIC jump is the statutory tax cut, not operations. Normalize by applying a consistent tax rate (say 25%) to both years for the trend read, because otherwise you would wrongly credit management with a margin/return improvement that the finance minister delivered. FY20 is a double discontinuity in most Indian series (this tax break and Ind AS 116); handle both.

Problem 6 (independent): Incremental ROIC across an acquisition. Synthetic. Over FY20→FY25, a company's NOPAT rose from ₹600 cr to ₹950 cr and invested capital rose from ₹2,500 cr to ₹5,000 cr. Of the ₹2,500 cr capital increase, ₹1,500 cr was goodwill + consideration from a FY23 acquisition that contributed ₹200 cr of the ₹350 cr NOPAT increase. Compute (a) the naive incremental ROIC including goodwill; (b) the organic incremental ROIC excluding the acquisition; (c) the acquisition's own return. Interpret against a 12% hurdle.

Solution. (a) Naive: ΔNOPAT 350 / ΔIC 2,500 = 14.0%. (b) Organic: strip the deal. ΔNOPAT (350 − 200) = 150 / ΔIC (2,500 − 1,500) = 1,000 = 15.0%. (c) The acquisition alone: 200 / 1,500 = 13.3%. Interpretation: the organic core reinvested at ~15% (comfortably above the 12% hurdle, which is good) while the acquisition earned ~13.3% (only just above the hurdle, so value-accretive but thin, and sensitive to whether the deal's synergies materialize). The blended 14% hides both facts. The capstone lesson: always split the window at the deal, because a headline incremental ROIC that mixes organic reinvestment with an expensive acquisition can flatter a mediocre deal or penalize a great core.

Problem 7 (independent): Mini-SOTP with a holdco discount. Synthetic. Holding company "H Ltd" owns: 50% of listed "A Ltd" (A's total market cap ₹40,000 cr); 100% of unlisted operating business "B" (B's EBITDA ₹500 cr; peers trade at 12× EV/EBITDA; B has no debt); and ₹1,000 cr net cash at the holdco. H's own market cap is ₹16,200 cr. (a) Compute H's SOTP/NAV. (b) At what discount does H trade? (c) Name two reasons the discount exists and one thing that could narrow it.

Solution. (a) Listed stake: 50% × ₹40,000 cr = ₹20,000 cr (mark at market). Unlisted B: ₹500 cr × 12 = ₹6,000 cr EV = equity value (no debt). Holdco net cash: ₹1,000 cr. Gross SOTP/NAV = 20,000 + 6,000 + 1,000 = ₹27,000 cr. (b) H trades at 16,200 / 27,000 = 0.60 → a 40% discount to NAV. (c) The discount is real and compensates for: double taxation (dividends taxed as they flow up to H and again as H pays out), no control over the underlying cash (minorities own H, not A's or B's cash flows), and illiquidity/strategic-holding (the stakes will likely never be sold). It could narrow on a catalyst: a demerger, a buyback by H, a special dividend, or a shift to a shareholder-friendly payout policy. Judgment: a 40% discount is roughly typical-to-slightly-narrow for Indian holdcos (peers often 45–60%); it is a cheap way to own A and B only if you believe a catalyst or an improving payout will close part of the gap, and only if the structure is not being used mainly to entrench promoter control at a discount.

Problem 8 (independent): Reverse-DCF plausibility. Synthetic. An Indian compounder has an FCF yield of 2.2%; you require a 12.5% return. (a) Using the fast Gordon inversion, what perpetual FCF growth does the price imply? (b) Judge it against India's ~10.5% nominal GDP and base rates.

Solution. (a) Implied perpetual growth ≈ required return − FCF yield = 12.5% − 2.2% = ~10.3%. (b) That is essentially India's nominal GDP rate, forever. The market is pricing the company to grow free cash flow in line with the entire economy in perpetuity, at a demanding discount rate. Against base rates (only a small minority of firms sustain economy-level-or-higher growth for decades), this is demanding: it shifts the burden of proof onto you to name the specific, durable engine (a widening moat, a long formalization/premiumization runway, high-return reinvestment) that puts this company in the right tail, or to conclude the price already fully credits an outcome that most firms do not achieve, leaving no margin of safety.

Problem 9 (independent): LODR Reg 23 materiality. Synthetic. A company's consolidated annual turnover is ₹6,000 cr. It proposes an RPT, sales of goods to a promoter-group entity, expected to total ₹700 cr in the financial year. (a) Is the RPT "material" under the LODR Reg 23 threshold? (b) What approvals does it need, and who cannot vote?

Solution. (a) Turnover of ₹6,000 cr sits in the first band of the materiality gate, 10% of consolidated annual turnover for companies up to ₹20,000 cr, so the gate is 10% of ₹6,000 cr = ₹600 cr. The proposed RPT (₹700 cr) exceeds ₹600 cr, so it is material. (b) A material RPT requires prior approval of the audit committee and approval of shareholders by ordinary resolution, and all related parties are barred from voting on that resolution (so it turns on the minority/public vote). Analyst action: check that the vote actually happened, read the proxy-adviser (IiAS/SES) recommendation, and read the terms. The vote is a safeguard, but a promoter with allied institutional holders can still pass a bad RPT, so the substance still matters. (Check the current threshold; the bands were set by the LODR Fifth Amendment 2025, as of October 2026. The rule before it, the lower of ₹1,000 cr or 10% of turnover, gives the same ₹600 cr here.)

Problem 10 (timed, 15 min): The governance verdict. Synthetic. "Gamma Ltd" looks wonderful on the screen: 28% revenue growth, 22% ROCE, expanding margins. The governance file shows: promoter holding 55%, but promoter pledge rose from 5% to 40% over three years; a promoter-owned entity became the sole distributor last year (sales through it now 12% of revenue); the statutory auditor resigned mid-year citing "pre-occupation" and was replaced by a smaller firm; a large inter-corporate deposit to a promoter entity appeared in the RPT note. State your governance verdict and the action, and explain why the attractive financials do not rescue it.

Solution. Verdict: EXTRACTIVE, PASS. Un-ownable at any price. Three independent vetoes, any one of which is disqualifying, appear together: (1) a rising promoter pledge (5%→40%: leveraged promoter, forced-selling risk, and a private motive to inflate the reported numbers); (2) a promoter-distributor RPT capturing 12% of revenue (value routed to a promoter entity); (3) a mid-tenure auditor resignation replaced by a smaller firm (the single loudest pre-fraud tell), alongside an inter-corporate deposit diverting cash to the promoter. The attractive financials are not a mitigant; they are the reason to distrust. A promoter with a rising pledge and a private incentive to prop the price, whose auditor just walked, reporting 28% growth and 22% ROCE, is exactly the configuration in which reported numbers are least trustworthy (M2.08: Manpasand, Vakrangee, and the mid-cap fraud pattern). Being right about the "growth" is worthless when the governance scan says the numbers may not be real and the cash is leaving. Action: stop the analysis; do not value it; file it as a pass with the reasons. This is the discipline the capstone certifies: the willingness to decline a great-looking company on governance.

Problem 11 (timed, 15 min): The cyclical/PSU trap. Synthetic. A metals company trades at a trailing P/E of 6× and screens as "cheap." Its current EBITDA margin is 28% (a cyclical peak); mid-cycle margin is ~16%. It is simultaneously funding a large debt-financed capacity expansion, and net debt / EBITDA has risen from 1.2× to 2.8× over two years. (a) Roughly what is the P/E on mid-cycle earnings? (b) Give the verdict and the two red flags.

Solution. (a) Normalize earnings to mid-cycle: on the same revenue, earnings scale roughly with margin, so mid-cycle P/E ≈ 6 × (28 / 16) = ~10.5×. The "cheap" 6× is on peak earnings; on normalized earnings it is ~10.5×, not cheap for a cyclical. (b) Verdict: the cyclical trap. The low trailing P/E is a sell signal near the peak, not a buy signal. Red flag 1: valuing a cyclical on trailing (peak) P/E, because the lowest P/E in a commodity company typically occurs at the peak, when trailing earnings are highest and about to fall (M5.09). Red flag 2: debt-funded capacity expansion at the peak with net debt/EBITDA rising into a likely downturn, the classic value-destroying move that leaves the company over-levered when the cycle turns (and, if the EBITDA denominator falls, that 2.8× becomes far worse). Correct approach: value on mid-cycle EBITDA and EV/EBITDA vs replacement cost and cost-curve position, and weight the rising leverage heavily. (If this were a PSU cyclical like Coal India, add the government-promoter overlay: dividend upstreaming to the government's fisc, disinvestment overhang, and policy/ESG risk, the reasons for the structural "governance discount" multiple.)

Problem 12 (guided): Scenario table from the Gordon lens. Synthetic. "Deccan Staples" throws off ₹12 of FCF per share; you require 12%. Using the single-stage Gordon value FCF/share × (1 + g) ÷ (r − g), compute the per-share value in three scenarios (bear g = 6%, base g = 9%, bull g = 10.5%) and say what the spread of the three answers tells you.

Solution. Bear: 12 × 1.06 ÷ (0.12 − 0.06) = 12.72 ÷ 0.06 = ₹212. Base: 12 × 1.09 ÷ 0.03 = 13.08 ÷ 0.03 = ₹436. Bull: 12 × 1.105 ÷ 0.015 = 13.26 ÷ 0.015 = ₹884. A 4.5 pp range in perpetual growth produces a ~4× range in value (₹212 → ₹884), because g sits close to the 12% discount rate: the closer g is to r, the smaller the denominator (r − g) and the more violently value reacts. The lesson: for a low-yield compounder a single point estimate is meaningless without the scenario range, and you should lean on the reverse DCF (what g is implied by the price) rather than defend one hand-picked g.

Problem 13 (guided): Reverse-DCF cross-check against your honest base. Synthetic. "Konkan Consumer" trades at ₹800/share, throws off ₹14 of FCF per share, and you require 12%. (a) What perpetual FCF growth does the price imply (Gordon inversion)? (b) If your honest base case is 8% perpetual growth, what is intrinsic value? (c) The verdict?

Solution. (a) Solve 800 = 14 × (1 + g) ÷ (0.12 − g): 800(0.12 − g) = 14(1 + g) → 96 − 800g = 14 + 14g → 82 = 814g → g = 10.07% ≈ 10.1%. (Fast check: r − FCF yield = 12% − 14/800 = 12% − 1.75% = 10.25%, the same order.) (b) Intrinsic at g = 8% = 14 × 1.08 ÷ (0.12 − 0.08) = 15.12 ÷ 0.04 = ₹378. (c) The price embeds ~10% growth forever; your honest base supports ~₹378, less than half the ₹800 price, overvalued on your own assumptions unless you can defend that ~10%+ perpetual growth is genuinely deliverable (name the durable engine, or pass). The reverse DCF converts "it looks expensive" into a specific, testable growth claim, the discipline the capstone rewards.

Problem 14 (independent): Build the bear/bull bridge. Synthetic. "Sahyadri Foods" has ₹10 of FCF per share, required return 12%. Bear assumes g = 6%, base g = 9%, bull g = 10.5%. (a) Compute the three per-share values. (b) Build the waterfall: how much of the bear→bull swing comes from bear→base versus base→bull? (c) What does the shape tell you?

Solution. (a) Bear: 10 × 1.06 ÷ 0.06 = ₹177. Base: 10 × 1.09 ÷ 0.03 = ₹363. Bull: 10 × 1.105 ÷ 0.015 = ₹737. (b) Bear→base = 363 − 177 = +₹186; base→bull = 737 − 363 = +₹374; total bear→bull = +₹560. (c) The upper leg (base→bull) moves g only half as far (1.5 pp, from 9% to 10.5%) as the lower leg (3 pp, from 6% to 9%), yet adds roughly twice the value, the convexity of the Gordon formula as g approaches r. Practically: most of the value and the risk lives in the bull, so the bull's plausibility deserves the most scrutiny, and small over-optimism in g is expensive.

Problem 15 (independent): Probability-weight the scenarios and judge the price. Synthetic. You value "Nilgiri Beverages" at bear ₹250, base ₹520, bull ₹900 and assign probabilities 30% / 45% / 25%. The stock trades at ₹600. (a) Expected value per share? (b) Expected-return proxy versus the price? (c) Buy, watch, or pass?

Solution. (a) EV = 0.30 × 250 + 0.45 × 520 + 0.25 × 900 = 75 + 234 + 225 = ₹534. (b) EV ÷ price − 1 = 534 ÷ 600 − 1 = −11%, a negative expected return from ₹600. (c) Watch, don't buy: your own probability-weighted value sits ~₹66 below the price, so there is no margin of safety; you become interested nearer the base value (₹520) or below. Note the discipline: the base alone (₹520) is already under the price, and the bull is doing all the work holding the EV up; if you doubt the bull, the case is worse still. (The trap is a simple average of the three, ₹557, which ignores the probabilities.)

Problem 16 (independent): Triangulate three lenses from one set of figures. Synthetic. "Malabar Brands": market cap ₹60,000 cr, PAT ₹1,200 cr, EBITDA ₹1,800 cr, net cash ₹1,000 cr, FCF ₹1,000 cr; WACC 12%. Compute (a) P/E, (b) EV/EBITDA, (c) FCF yield, (d) the reverse-DCF implied perpetual growth. (e) Do the lenses tell one story?

Solution. (a) P/E = 60,000 ÷ 1,200 = 50.0×. (b) EV = 60,000 − 1,000 = 59,000; EV/EBITDA = 59,000 ÷ 1,800 = 32.8×. (c) FCF yield = 1,000 ÷ 60,000 = 1.67%. (d) Implied g = WACC − FCF yield = 12% − 1.67% = 10.33%. (e) Yes: a ~50× P/E, a ~33× EV/EBITDA, and a price needing ~10% FCF growth forever are three views of the same demanding story, a high-quality business priced for a long, high-growth runway. Had the P/E instead screened "cheap" while the reverse DCF implied ~10% growth, you would suspect a mis-set WACC, a peak-cycle earnings base, or an un-adjusted Ind AS 116 lease / FY20 tax break. Triangulation's job is to make one lens catch another's error.

Problem 17 (timed, 15 min, valuation × governance): The stretch integration. Synthetic. "Vindhya Consumer" throws off ₹9 of FCF per share; you require 12.5%. Your scenarios: bear g = 6% (35%), base g = 9% (40%), bull g = 11% (25%). The stock trades at ₹500. Part 1: (a) the three per-share values; (b) the probability-weighted value; (c) the buy/watch/pass call on valuation alone. Part 2: you then pull the shareholding pattern and find the promoter pledge has risen from 5% to 35% over three years, funding an unrelated group venture. State the final verdict and defend the sequencing.

Solution. Part 1: (a) Bear: 9 × 1.06 ÷ (0.125 − 0.06) = 9.54 ÷ 0.065 = ₹147; Base: 9 × 1.09 ÷ (0.125 − 0.09) = 9.81 ÷ 0.035 = ₹280; Bull: 9 × 1.11 ÷ (0.125 − 0.11) = 9.99 ÷ 0.015 = ₹666. (b) EV = 0.35 × 147 + 0.40 × 280 + 0.25 × 666 = 51.45 + 112 + 166.5 = ₹330. (c) On valuation alone: watch / pass, because the EV of ₹330 is ~₹170 below the ₹500 price (a −34% expected-return proxy), no margin of safety. Part 2: final verdict is EXTRACTIVE, PASS, un-ownable at any price. A promoter pledge rising 5% → 35% to fund an unrelated venture is a governance veto: forced-selling risk, a private incentive to prop the reported numbers, and a cash-diversion signal (Checks 2 and 3). The sequencing the capstone certifies: the governance veto is evaluated independently of, and overrides, the valuation. You do not "price it in" with a lower multiple or a bigger discount, because being right about the cash flows is worthless if the promoter is working around you. Even had the scenarios screened cheap (EV above price), the pledge alone would end the analysis.

Applied mini-project: THE INDIA CAPSTONE (a contributor to the M11 milestone)

This is not a mini-project; it is one of the two graded capstone theses that record the M11 milestone (core competence) and the module's entire purpose. Budget essentially all forty hours here. It is done, unaided, on one company from the §4.2 approved list that you have never studied. The integrity rule from M8.05 applies in full: never-studied means never-studied, no peeking at a bull/bear note first, and if a material development (an auditor change, a SEBI order, a restatement) has intervened, that company just became a better capstone target, not a spoiled one.

The deliverable. The full thesis package of §4.7: the one-page thesis, the supporting deep-dive memo, the one-page governance dossier, and the exhibits (history spreadsheet, peer comps, valuation range, and, where relevant, the standalone-vs-consolidated reconciliation and the SOTP). It must contain everything in §4.3 (the shared analytical core) and everything in §4.4–§4.6 (the India layer), to the standard of the Marico worked example. Then self-grade it against the rubric below to the pass bar.

The two-week time budget

At the immersive cadence, weeks 71–72. A guide, not a straitjacket: a bank or a live-governance name will shift the weight toward the governance and sector work.

Total ~50 module-hours; fold the daily loop (flashcards, the mastery check below) and the reading spine on top per the master-map weekly template.

The capstone grading rubric (criteria × 0/1/2)

Score every criterion 0, 1, or 2 against the anchor, as a hostile examiner. Maximum 38. The India-layer criteria (10–15) are what make this the harder, distinct capstone.

Pass bar: ≥ 31/38 (≥80%), no criterion scored 0, and the seven hard gates (2, 8, 10, 11, 15, 17, 18) each ≥ 1. The four US-capstone hard gates (incremental ROIC, reverse DCF, steelman/pre-mortem, one-page thesis) are joined here by the three that make it the India capstone: the pledge check, the related-party trail, and the governance verdict. A package that values beautifully but cannot evidence whether the promoter is working for you is exactly the package that impairs capital in this market. Both capstones (M10.01 and M10.02) must clear their bars for the M11 milestone.

The self-review protocol

Grade only after the package is completely written (working-agreement 3: write before you peek, even at your own rubric). Then:

1. Score the seven hard gates first. Any hard gate at 0 fails the capstone regardless of the total, so you know immediately whether you have a pass candidate or a specific hole to fill.
2. Score all nineteen rows as a hostile PM, one sentence of evidence-tied justification each. Effort is not a criterion; cold, evidenced capability is.
3. Run the India self-review checklist, the ten questions that catch the India-specific failure modes:

• Did I confirm the pledge from the filing, not assume it from reputation?
• Did I track the material RPTs as a % of sales over time, not merely note they exist?
• Did I ask "who benefits" on every capital-allocation choice?
• Did I analyze the right entity (consolidated vs standalone) and reconcile the gap?
• Did I handle the Ind AS 116 and FY20-tax breaks in the ten-year series?
• Did I carve out any listed stake before reading the multiple?
• For a cyclical/PSU, did I normalize to mid-cycle and refuse trailing P/E?
• Is my governance verdict a dossier of evidence, not a paragraph of adjectives?
• Is my variant perception falsifiable and my pre-mortem wired to triggers?
• Would this package survive a hostile PM who has been burned in India before?

1. Cross-check against the ten common mistakes (the "Common mistakes" section). Any that you committed is a targeted remediation.
2. Remediate honestly. Below bar is diagnosable: identify the weak criteria; if the failure is in specific sections (the governance dossier, the reverse DCF, the SOTP), rework those sections on the same company first (rework builds the skill; a new target only rebuilds hope); if the whole package is structurally thin, drop to the faded rung for one component and rebuild. Wait the mandatory two days, then re-sit any timed element on a fresh never-studied name. The deep-dive package itself is reworked, not restarted. Log the score and your top miss in the decision journal.

Passing this capstone is evidence for the C10, C2, C7, and C11 rows of the M10.04 final audit, the strongest single piece of evidence that you can do the whole thing, on the harder jurisdiction, unaided.

Reading & resources

The single most important resource is your candidate's own primary filings. Read them before any secondary summary. The program's own artifacts are the operational backbone: the artifacts/playbook-2-3-day-deep-dive.md workbook (the method you are executing), the checklist-red-flags-quality-of-earnings.md, the checklist-mental-models-and-moat.md, the sector-kpi-cheatsheets.md (the right lens for your candidate's sector), the valuation-workflow.md, and the phase-cross-referenced master-reading-list.md.

Primary Indian sources (all [Free] unless noted).

• BSE (bseindia.com) and NSE (nseindia.com) [Free][Intermediate]: company announcements, quarterly financial results, the shareholding pattern (promoter holding + pledge), the corporate-governance report, and RPT disclosures. Your first stop for the governance file.
• The company's Annual Report + Investor Presentation [Free][Intermediate]: the Board's Report, the AOC-2 related-party annexure, the auditor's report + CARO, the notes, and BRSR. The document the whole dossier is read from.
• MCA21 (mca.gov.in) [Free][Advanced]: statutory filings (AOC-4 financials, MGT-7 annual return, and the register of charges/pledges) for deeper promoter and group-structure work.
• screener.in [Free / Paid][Beginner–Intermediate]: ten-year financials on one screen, concall transcripts, shareholding + pledge history, and ratios, for speed, then verify against the filing. Trendlyne / Tijori [Paid][Intermediate] for deeper RPT and promoter data.
• SEBI (sebi.gov.in) [Free][Advanced]: the LODR Regulations 2015 text (Reg 17 board, 18 audit committee, 23 RPTs, 24 subsidiaries, 30 disclosure, 31 shareholding/pledge), the SAST Reg 31 pledge rules, and the orders/adjudications. Read a governance-case order to see the patterns in the regulator's own words.
• Companies Act 2013 [Free][Advanced]: Section 188 (RPTs), 197 (remuneration), and Schedule III, Division II (the Ind AS statement format and the newer promoter-loan/wilful-defaulter disclosures).
• Proxy-advisory research: IiAS, SES, InGovern [Free summaries / Paid][Advanced]: independent second opinions on RPT and remuneration resolutions; a governance firm flagging your candidate's resolution is a free red-team.

Cross-jurisdiction and valuation.

• Damodaran data library and India country-risk page [Free][Advanced]: the India ERP, country-risk premium, betas, and industry multiples, updated each January; pull the live numbers for your WACC.
• SEC EDGAR [Free][Intermediate]: for a candidate with a US-listed subsidiary or an MNC parent (e.g., Sun's Taro on the NYSE; a foreign parent's 20-F).

Books and lectures.

• Sanjay Bakshi — "Behavioural Finance & Business Valuation" (BFBV) lectures [Free][Advanced]: the definitive India-context teaching on promoter quality, moats, and governance judgment.
• Saurabh Mukherjea — The Unusual Billionaires and Coffee Can Investing [Paid][Intermediate]: India quality-and-governance screening in practice.
• Schilit, Perler & Engelhart — Financial Shenanigans (4e) [Paid][Advanced]: the forensic backbone for the "distrust the reported number" discipline; re-read the RPT and cash-flow chapters.
• Palepu & Healy — Business Analysis and Valuation [Paid][Advanced]: the deep-dive frame; the governance and accounting-analysis chapters.
• Damodaran — Narrative and Numbers [Paid][Advanced] (turn every governance judgment into a number) and Investment Valuation (SOTP, emerging-market risk premiums, holdco discounts).
• Zerodha Varsity — Fundamental Analysis (the annual-report and RPT modules) [Free][Beginner]: the India-specific quick reference.

Governance case reading (for method, evenhandedly). The public record on Satyam, IL&FS, DHFL, Yes Bank, Manpasand, Vakrangee (M2.08), and, reading both sides, the Adani–Hindenburg exchange and the Zee/Essel saga. Read short-seller reports and company rebuttals for method, not verdict; even a self-interested short often names the real fragility, and the company's rebuttal shows you how a promoter answers under fire.

• Going deeper — Aswath Damodaran, valuation spreadsheets + the "Pricing vs Value" / reverse-DCF blog posts (free spreadsheets at pages.stern.nyu.edu; posts at aswathdamodaran.blogspot.com) [Free] [Advanced]: rebuild Worked Steps 8–9 yourself. His two-stage / "ginzu" DCF spreadsheets run the scenario-and-sensitivity grid on live inputs, and the pricing-vs-value posts model the reverse-DCF cross-check that turns a full price into the growth it silently assumes. Pull the January India ERP/country-risk numbers while you are there and re-solve for your own candidate.
• Going deeper — Alfred Rappaport & Michael Mauboussin, Expectations Investing (revised edition) (Columbia Business School Publishing) [Paid] [Intermediate→Advanced]: the book-length method behind Step 9's expectations lens. Start from the price, read the market's implied expectations, then locate the operating "value triggers" (here, the D2C/foods incremental-ROIC and the runway) that would confirm or refute the bet. That is the disciplined upgrade to a point-estimate DCF on a full-priced compounder.

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 India capstone: the shared analytical spine plus the three India layers — the six-check Governance & Promoter Deep-Read, the Ind AS adjustments that change the numbers (116 leases, the FY20 tax break, 115, consolidation), and the SOTP / holding-company-discount check",
  "use": "As in M10.01, the capstone itself is unaided and this callout does not amend that. It is about the practice afterwards, and about the one place where the India layer creates a genuine retrieval problem rather than a judgement problem. (1) THE SCATTERED DISCLOSURE SURFACE is the real use. Section 4.2 says India spreads governance signal across the Board's Report, the AOC-2 annexure, the corporate-governance report, the shareholding pattern and the pledge disclosures, each on its own schedule; a retrieval system over ten years of those documents turns 'find every related-party transaction and its counterparty' from a week into an afternoon. Build the corpus, ask it questions with page citations, then read the pages. (2) THE RPT TREND TABLE: extraction of counterparty, nature, and amount per year is mechanical, and the analysis - is this small, stable and arm's-length, or large, growing and skimming? - is entirely yours. (3) SHAREHOLDING-PATTERN AND PLEDGE SERIES across twelve quarters: pure extraction into a table you then read. (4) THE ENGLISH-LANGUAGE DIFF: what changed in the corporate-governance report, the auditor's report, or the CARO annexure versus last year. A mid-tenure auditor resignation and a new emphasis-of-matter are the module's loudest tells and they are diffs.",
  "tools": ["Document/PDF AI and retrieval over your own corpus — for ten years of annual reports, AOC-2 annexures, CG reports and twelve quarters of shareholding patterns", "Coding assistants — for the RPT and pledge time series, the Gordon sensitivity grid and the SOTP carve-out", "Chat assistants — Claude, ChatGPT for the year-over-year diff of governance documents and for the hostile-PM read of a finished dossier"],
  "prompt": "Here are ten consecutive annual reports for <COMPANY> <ATTACH>. Build one table of every related-party transaction disclosed in the related-party note and the AOC-2 annexure, with columns: financial year, counterparty, relationship to the promoter group, nature of transaction, amount, and the page or note number it appears on. Do not classify, judge or summarise - I do that. Then a second table from the twelve attached shareholding-pattern filings: quarter, promoter and promoter-group holding %, shares pledged or encumbered as % of promoter holding, and % of total shares. If a figure is absent for a period, write ABSENT rather than interpolating.",
  "verify": "Every extracted number is checked against the cited page before it enters the dossier, and the check is not optional here: an RPT amount that is quietly wrong produces a governance verdict that is quietly wrong, and the governance verdict is a VETO in this module - it decides whether the company is investable at all, so an error there is not a valuation error, it is the whole position. Retrieval also fails silently on absence: a model asked for pledge disclosures in a year where the filing format changed will often return the nearest year's figure rather than saying nothing, which is why the prompt demands ABSENT and why you reconcile the row count against the number of filings. Never let it deliver the verdict: Section 4.4 Check 6 has three grades and one veto, and a fluent reader of a clean-looking document will grade generously. And re-derive the Ind AS adjustments yourself - the FY20 double discontinuity is exactly where an extracted series looks continuous and is not.",
  "diy": "The gate is unaided: you run the six governance checks off the primary documents, restate a ten-year series across the Ind AS 116 and FY20 tax breaks, compute incremental ROIC on a consistent tax basis, build an INR WACC from a G-sec yield and a country-risk-adjusted ERP, invert a price to its implied perpetual growth, carve out a listed subsidiary in a SOTP, and write a one-page thesis with a governance verdict that can veto it. No tool is in that room."
}

Modern Data Analysis

By hand first. You ran the six governance checks, restated a ten-year series across two FY20 discontinuities, split the incremental-ROIC window at the acquisitions, built an INR WACC from a G-sec yield and a country-risk-adjusted equity risk premium, ran a Gordon sensitivity grid to a probability-weighted ₹571, carved out a listed foreign subsidiary, and produced a one-page thesis whose governance verdict sits above the valuation as a veto. None of that is delegable and §4.1 means it.

Today's workflow. The reproducibility question M10.01 raises is sharper here, because the answer is worse, and the module's own design is what makes it worse.

• India's disclosure is a corpus, not an endpoint. §4.2's second asymmetry says the governance signal is scattered across documents on separate schedules. That is a retrieval problem before it is an analytical one, and it is the single place in Phase 10 where modern tooling changes what a two-week budget can cover.
• The FY20 double discontinuity is a reproducibility trap. §4.5 names both breaks (Ind AS 116 and the corporate-tax cut) and instructs you to normalize. A series that looks continuous across FY20 and has not been restated is wrong in a way no arithmetic check will catch.
• And the SOTP carve-out has live cross-border inputs. A Dhaka quote and a BDT/INR rate are not facts in a filing; they are prices with a date, which means the SOTP is the fastest-perishing exhibit in the package.

Tools & sources (IN + US). pandas for the RPT and shareholding time series, a local retrieval index over your own PDF corpus for the governance read, numpy for the Gordon grid, and a plain-text normalization log under version control. India: the consolidated annual report as the primary source for the panel, the related-party note, the AOC-2 annexure to the Board's Report, the corporate-governance report, the auditor's report and CARO 2020 annexure, and the Section 197(12) remuneration disclosures; the quarterly shareholding pattern filed to BSE and NSE under LODR Reg 31 (the pledge line lives here, and Checks 1–2 need twelve quarters of it); SAST pledge creation/release/invocation disclosures; NSE and BSE corporate-announcement archives for Reg 30 material events and Reg 33 results; MCA21 for filings the exchanges do not carry; proxy advisers IiAS, SES and InGovern for a free second opinion on material RPT resolutions under LODR Reg 23; screener.in for speed and cross-check only, never for a staked figure; RBI and CCIL/FBIL for the G-sec yield behind the INR WACC; MOSPI for the nominal-GDP series the terminal g must respect; Damodaran's country-risk-premium table for the India CRP. US (the comparison this capstone is defined against): EDGAR XBRL companyfacts at data.sec.gov, one free call returning every tagged fact a filer has ever reported with the accession number per fact, which is the thing India has no equivalent of and the reason the two capstones cannot promise the same reproducibility.

# Rebuild the walkthrough, then measure what the FY20 tax break does to it. R3, in-session.
fcf_ps, r, price = 1400/130, 0.12, 650                    # 10.769 owner-earnings per share
gordon = lambda g, rr=r: fcf_ps*(1+g)/(rr-g)
print([round(gordon(g)) for g in (0.070, 0.099, 0.102, 0.108)])       # [230, 564, 659, 994]
print(round(0.30*gordon(.070) + 0.45*gordon(.099) + 0.25*gordon(.108), 1))   # 571.3

# the module PRINTS a NOPAT row and then INSTRUCTS a consistent 25% tax for the trend read.
EBIT = {'FY16': 1000, 'FY20': 1300, 'FY25': 1950}
IC   = {'FY16': 2000, 'FY20': 2600, 'FY25': 4600}
NOP  = {'FY16':  700, 'FY20': 1000, 'FY25': 1500}         # as printed
print({y: round(1 - NOP[y]/EBIT[y], 3) for y in EBIT})    # {FY16: 0.3, FY20: 0.231, FY25: 0.231}
roiic = lambda n, a, b: (n[b]-n[a]) / (IC[b]-IC[a])
print(round(roiic(NOP, 'FY16', 'FY20'), 3), round(roiic(NOP, 'FY20', 'FY25'), 3))   # 0.500 0.250
norm = {y: EBIT[y]*0.75 for y in EBIT}                    # the module's own instruction applied
print(round(roiic(norm, 'FY16', 'FY20'), 3), round(roiic(norm, 'FY20', 'FY25'), 3)) # 0.375 0.244

Verify. The walkthrough reproduces cell for cell. Step 8: owner-earnings per share ₹10.7692; the sensitivity grid returns 256 / 230 / 210, 332 / 291 / 258, 470 / 391 / 335, 740 / 564 / 455, 913 / 659 / 516, 1,705 / 994 / 702 at WACC 11.5 / 12.0 / 12.5% for g = 7.0 / 8.0 / 9.0 / 9.9 / 10.2 / 10.8%, every printed cell; bear ₹230, base ₹564, the ₹650 price cell ₹659, bull ₹994; the probability-weighted value at 30/45/25 is ₹571.3, sitting ₹78.7 below the price. Step 9's triangulation: FCF yield 1.67%, P/E 54.2×, EV/EBITDA 39.8×, fast implied growth 10.33% and the exact Gordon solve 10.17%, matching the module's ~1.7 / ~54 / ~40 / ~10.3 / ~10.2. The bear-to-bull bridge on the rounded per-share values is +₹334 / +₹95 / +₹335 for +₹764. Step 1's incremental ROIC on the printed NOPAT row is 50.0% for FY16→FY20 and 25.0% for FY20→FY25.

Quantitative lens

The FY20 tax break is inside the module's own headline ROIIC, and normalizing it costs 12.5 points. §4.5 states the break plainly. India cut the corporate rate in September 2019, Marico opted into the ~25% regime, and "NOPAT, net margin, and ROIC jump for a reason that is fiscal, not operational". Then it instructs: use a consistent ~25% tax for the trend read. Apply that instruction to the module's own figures and watch what happens. The printed NOPAT row implies an effective tax rate of 30.0% in FY16, 23.1% in FY20 and 23.1% in FY25, so the row itself straddles the regimes. Window A on the printed row gives (1,000 − 700)/(2,600 − 2,000) = 50.0%. Recompute it from EBIT at a consistent 25%: NOPAT of ₹750cr and ₹975cr, so 225/600 = 37.5%, a 12.5-point fall. Window B is untouched: 25.0% printed against 24.4% normalized, because both its years sit inside one regime. The mechanism is exact and general: a cut in the effective rate from 35% to 25% raises NOPAT by 15.4% with no operating change whatever, and on Marico's FY20 base it lifts headline ROCE from 32.5% to 37.5%, five full points of purely fiscal improvement. So the walkthrough's most quotable contrast, a legacy core reinvesting at 50% against new capital at 25%, is really 37.5% against 24.4% once the module's own normalization is applied. The conclusion survives (the core still compounds at roughly half again the new capital's rate, and both clear the 12% INR hurdle), but the gap is 13 points, not 25, and any Indian ten-year series that crosses FY20 without this restatement carries the same distortion in the same flattering direction.

The reproducibility asymmetry, counted rather than asserted. §4.1 says India is the harder twin and gives four structural reasons. Reason 2, scattered disclosure, has a measurable cost, and here is one way to measure it. Count primary documents: separately-filed artefacts you must locate, download and read to obtain what each module's spec requires. M10.01 (US): twelve. One EDGAR companyfacts JSON, which returns every tagged fact for every year in a single free call; ten 10-Ks for the notes, segments and normalization detail; one DEF 14A, which carries the entire incentive read. M10.02 (India): twenty-four. Ten consolidated annual reports, because there is no cross-year fact store and each year's P&L, balance sheet, cash flow, related-party note, AOC-2 annexure, CG report, CARO annexure and Section 197(12) disclosure live inside its own document; twelve quarterly shareholding patterns, because §4.4's Checks 1 and 2 explicitly require the trend over eight to twelve quarters and each quarter is a separate filing; and two more for the mini-SOTP (the Bangladesh annual report and a DSE quote). That is a 2.0× document load for the same analytical spine. But the count understates it, because the decisive difference is addressability, not volume: in the US every mechanical cell has a (tag, period, accession) address at one free, documented, stable endpoint, so a figure is re-queryable forever; in India the reproducible unit is a document reference (filer, filing date, page or note number), so a figure is re-checkable but not re-queryable. That distinction is the whole of what each capstone can honestly promise. The US capstone can promise that its mechanical rows regenerate on demand. The India capstone can promise only that every figure is traceable to a dated document and a page. Both are legitimate standards; they are not the same standard, and a package that claims the first while delivering the second is claiming something it cannot do.

And the SOTP is the exhibit that perishes fastest. Step 5 carves Marico's ~90% Bangladesh stake out of an ~₹84,000cr market cap at an illustrative ₹8,000–10,000cr. Priced as a share of the whole, that band is 9.52% to 11.90% of market capitalisation, a 2.4-point spread on the residual value attributed to the India business, driven entirely by which day you took the quote. And the stake has two live inputs, not one: a Dhaka Stock Exchange price and a BDT/INR rate. On a ₹9,000cr midpoint, a ±10% currency move alone shifts the carve-out by ₹900cr, or 1.07% of market cap, before the underlying share price has moved at all. Neither input appears in any filing. The practical discipline is the one §4.7's package spec already implies and this makes concrete: stamp the SOTP exhibit with both input values and both dates, and re-price it before the thesis is re-used, because it is the one component that can be stale while every other page is still correct.

Honest limits, and the first is the largest. Every figure in the walkthrough is explicitly illustrative (the Marico series is stated as approximate, the Bangladesh stake as an order of magnitude), so reproducing it verifies the engine and the arithmetic and says nothing whatever about Marico Ltd, which is a teaching specimen and not a recommendation; pull the filings and rebuild. The 12.5-point ROIIC finding is a consequence of applying §4.5's own normalization instruction to §Step-1's own printed row; it is not a claim about what Marico's real incremental returns were, and on real filings the effective rate path, the goodwill treatment and the acquisition timing would all move it further. The document count is a count under a stated rule, and the rule is arguable: a reader who treats each quarterly results announcement as a primary document gets a much larger Indian number, and one who reads only the latest annual report gets a much smaller one. What is robust is the direction and the addressability point, not the 2.0×. The XBRL advantage on the US side is real but conditional: tags change across accounting standards, restatements exist at two accessions with two values, and a pull that does not carry the accession per cell is reproducible and wrong simultaneously. And the deepest limit is that reproducibility is a property of numbers only. The governance verdict, §4.4's veto and the thing that decides whether this company is investable at any price, rests on judgement about counterparties, intent and pattern that no source makes machine-checkable in either market. Its reproducibility is entirely the quality of the evidence trail you write down.

Do it in code: build the ten-year corpus first and index it, extract the RPT and pledge series into tables you then read yourself, restate the series across both FY20 breaks before computing any ratio, write the Gordon grid as a function so the implied-growth cell moves with the price, and stamp the SOTP with its two inputs and their dates. Then write the normalization log in prose a stranger could follow, because in this market that log is the whole of what reproducibility means. See G2/QD2.04, G3/AI0.04 and G2/DA1.03.

Where this goes next: galaxy cross-links

• G2 · QD2.04 Reproducible Research & Deployment: pinned inputs, dated outputs and the normalization log, and what "reproducible" can honestly mean when the source is a PDF rather than an endpoint.
• G3 · AI0.04 Retrieval & RAG: building an index over ten years of annual reports, AOC-2 annexures and shareholding patterns, including the chunking and citation discipline that makes every extracted figure traceable to a page, and the silent-absence failure that makes a missing disclosure look like a present one.
• G2 · DA1.03 Pandas Cleaning & Time-Series Bridge: structural breaks in a series, detecting the FY20 double discontinuity, restating across it, and why a break that is not flagged becomes a trend.
• G1 · QM1.01 Probability, Distributions & Estimation: what a probability-weighted value is and is not, and why three scenarios with assigned weights is a summary of your beliefs rather than a measurement of the world.
• G3 · CN1.02 Hypothesis-Driven Problem Solving: the governance deep-read as a set of falsifiable hypotheses with pre-named disconfirming evidence, which is what turns "the governance felt fine" into a dossier.

Teach it back & journal

Teach it back (Feynman). In one page, no jargon, explain to a smart 15-year-old: "Why can a company that is mostly owned by one family be either the safest kind of company to invest in or a trap — and what is a 'share pledge,' and why does it make grown-up investors nervous?" If you cannot make the controlling-owner-vs-minority idea and the pledge mechanism vivid and simple, you have not yet internalized the thing this whole capstone is built to teach. (Model move: the family owner who wants the family's shares worth more in twenty years is your best friend; the family owner who has secretly borrowed against those shares to fund a different failing business now needs your company's stock price to stay up today, and that changes everything about whether you can trust what they tell you.)

Journal reflection. Before you finish your capstone, write down which you feared more, that the reported numbers were wrong or that the price was wrong, and why. After the governance deep-read, write whether it changed your verdict (did a clean read let you trust numbers you would otherwise have distrusted? did a red flag override a beautiful valuation?), and name one governance pattern you can now no longer un-see in any Indian company. Then the calibration line for the decision journal: your confidence in the thesis, the one thing that would change your mind, and the date you will re-check the monitoring dashboard. This entry, written before the outcome is known and never edited after, is where your India governance judgment starts compounding into an edge.

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