The Full Filing + Capstone

Phase 1 · Accounting: The Language of Business · Week 17 · ~20 hours (≈6 h teaching content, ≈2 h practice set, ≈10–12 h capstone spread, ≈1 h mastery check)

Where it sits. You've reached the end of Phase 1 and the M2 milestone of the whole program. Behind you is the machinery: double-entry, the three statements, the linkage crown gate, and the topic modules on revenue, assets, liabilities, and equity and consolidation. Ahead is Phase 2, where the numbers you can now read become numbers you can judge. This week adds the last layer. That's everything in a filing that isn't a statement, which means policies, estimates, MD&A, the auditor's report, CARO, BRSR, the proxy and the shareholding pattern. Then you prove the whole phase by hand-spreading one full US 10-K and one full Indian annual report into a standard template, graded by a tie-out protocol.

You'll leave with two artifacts. The first is your spreading template, which becomes the historical block of every model you build in Phase 3. The second is the 45–60 minute statement-read checklist v1, and Phase 8 refines it into the final rapid-analysis artifact.

Learning objectives

By the end of this week you can:

1. Navigate both filings blind: name every section of a US 10-K (Item 1 through Item 15) and an Indian annual report (corporate overview → statutory reports → standalone and consolidated financials), say what each is for, and what an analyst checks in it.
2. Extract the policies that move numbers: identify the eight significant accounting policies worth reading closely, detect a year-on-year policy change, and classify a policy as aggressive or conservative relative to peers.
3. Read an auditor's report properly: classify the opinion (unmodified / qualified / adverse / disclaimer), distinguish an Emphasis of Matter from a modification, use KAM/CAM paragraphs as a map to estimation risk, and recognize going-concern language in both US and India formats.
4. Scan CARO 2020 like a forensic checklist: locate and interpret the high-signal clauses on bank stock statements, related-party loans, statutory dues, defaults and fund diversion, fraud, cash losses, and auditor resignation.
5. Read MD&A adversarially: test this year's promises against prior years' delivery, use India's mandated ratio-change explanations (LODR Schedule V), and detect KPI switching and blame asymmetry.
6. Dissect ownership and pay: pull comp structure, insider ownership, and related-party transactions from a DEF 14A; pull promoter holding, pledge percentage, and Reg 23 RPT materiality from Indian disclosures; compute the pledge margin-call math.
7. Police non-GAAP metrics: rebuild any adjusted-earnings reconciliation, classify every add-back, and compute the cumulative GAAP-vs-adjusted wedge.
8. Hand-spread a complete filing: spread five years of IS, BS, CFS, and key notes for a US and an Indian company into the standard template, and prove the spread with the self-check protocol: every balance sheet balances, cash rolls, equity rolls, EPS recomputes.

Prerequisites & connections

Builds on. Everything in Phase 1. M1.05's linkage work matters most, because the tie-outs here are the same checks run on real filings. M1.06–M1.09 supply the topic knowledge, so you can now understand the notes you're about to triage. M0.04 gave you the toolkit: EDGAR, BSE/NSE, screener.in, your spreadsheet. M0.05 was the guided first read, where you also met adjusting and non-adjusting subsequent events, and this capstone is that read done with judgment and finished with a spread.

Feeds into. Phase 2 computes every ratio off the spread you build here, and M2.01–M2.05 assume your template exists. M2.06–M2.08 (shenanigans) dig into the notes, CARO clauses and non-GAAP bridges you learn to read now. M3.08 (3-statement modeling) uses your spread as the model's historical block. M8.01 turns your checklist v1 into the final 45–60 minute artifact, and M8.06 drills subsequent events on live filings. M1.12 works the pension roll-forward and discount-rate sensitivity, and M1.13 the foreign-currency method comparison and ratio grid. The governance material here seeds C7 (management & capital allocation) and the ethics thread.

Competency map. This capstone certifies C1: read the language of business (hand-spread + explain any footnote), and plants the seeds of C2 (quality of earnings) and C7 (incentives).

Core teaching content

4.1 Why the pages without numbers decide whether you can trust the pages with numbers

The three statements are management's claims, expressed in a standardized grammar. Everything else in the filing exists to answer three questions about those claims:

• What assumptions manufactured these numbers? Accounting policies and critical estimates tell you.
• Does anyone independent vouch for them? That's the job of the auditor's report, KAM/CAM and CARO.
• What do the people in charge want, and are they aligned with you? Look at MD&A, the proxy/Board's Report, ownership and pledge disclosures, and the related-party notes.

Now map this to the five questions. The statements feed Q1 (what does it do) and Q2 (is it a good business). The narrative and governance layer feeds Q3 (good, honest, aligned people?) and Q4 (what can kill it?). It also decides whether the raw material for all five can be trusted at all. If you read only the statements, you're grading an exam where the student set the questions and wrote the answers, and the marking scheme too. You catch that in the rest of the filing.

It's a lot of paper. A US 10-K runs 100–300 pages and a modern Indian annual report 300–500, and BRSR alone can be 40+. Nobody reads every page with equal attention. The professional's triage goes like this: read ~30 pages deeply, skim ~50, and ignore the rest knowing what you're ignoring.

4.2 Anatomy of the US 10-K

The 10-K is the annual report filed with the SEC under the Securities Exchange Act. Know its skeleton cold. Page numbers change, but Item numbers don't.

The 10-K is the annual audited filing, and it comes with a family. Next is the 10-Q, which is quarterly and reviewed but not audited. The 8-K carries material events within 4 business days, and auditor changes, CFO exits and restatements land there first. The DEF 14A is the proxy, filed within 120 days of FY-end when Part III is incorporated by reference. The 20-F is the annual report of a foreign private issuer. Infosys files one for its NYSE listing, which makes it a useful bridge between Indian and US disclosure. Large accelerated filers file the 10-K within 60 days of FY-end. Accelerated filers get 75 days, and everyone else 90. The glossy "annual report to shareholders" is a marketing wrapper. Spread from the 10-K, because the 10-K is the legal document.

4.3 Anatomy of the Indian annual report

An Indian AR is assembled under the Companies Act 2013, SEBI LODR, and Ind AS. Its skeleton:

1. Corporate overview (the glossy). Chairman's letter, highlights, "value creation model." It's marketing. Read the chairman's letter once for tone and promises, since you'll test them later, and ignore the infographics.
2. Statutory reports.

• Board's Report (Directors' Report) with annexures: AOC-2 (related-party transactions not at arm's length / material RPTs), remuneration disclosures under s.197, which include the ratio of each director's pay to median employee pay, the CSR annual report (the 2%-of-profits spending rule under s.135), conservation of energy, and the secretarial audit report (MR-3).
• Management Discussion & Analysis, required by LODR Schedule V and covered in the MD&A section ahead.
• Corporate Governance Report: board composition and independence, committee memberships and attendance, auditor fees (audit vs non-audit), the shareholding pattern, and a practising company secretary's compliance certificate.
• BRSR: Business Responsibility and Sustainability Report, covered further below.

1. Financial statements, twice. Standalone (the parent legal entity alone) then consolidated (group). Each set carries its own independent auditor's report. The standalone report also carries the CARO 2020 annexure, covered further below, and the internal-financial-controls opinion under s.143(3)(i), which is India's SOX-404 analog.
2. AGM notice, sometimes bound in and sometimes separate. It lists the resolutions shareholders will vote on, including auditor appointments, director reappointments, RPT approvals, ESOP schemes.

Three rules govern how you use the standalone and the consolidated.

• Rule 1: Analyze the consolidated. It's the economic entity. All Phase 2 ratios and all valuation run on consolidated numbers.
• Rule 2: Check the standalone for cash and dividends. Dividends are declared out of the standalone entity's profits (Companies Act s.123). A parent can look great on consolidated numbers and still struggle to pay if its standalone entity is starved, with the profits trapped in subsidiaries. Only the standalone statements show the loans, guarantees and investments the parent has pushed into group entities. Read the standalone loans note, CARO clause (iii) and the RPT note together.
• Rule 3: Treat the standalone–consolidated gap as a diagnostic. If consolidated profit ≫ standalone, the value is in the subsidiaries. Ask which ones, and check NCI leakage with last week's consolidation work. If standalone ≫ consolidated, subsidiaries are losing money, so ask why the group keeps funding them.

India runs to its own timing. Audited annual results reach the exchanges within 60 days of FY-end under LODR Reg 33, so by ~May 30 for a March year-end. The full annual report arrives with the AGM cycle. The AGM falls within six months of FY-end, and the report goes out ≥21 clear days before it. In practice, from June to September you may be working from the results and the investor presentation while you wait for the full AR and its notes. Quarterly results in between are only limited-reviewed. Nobody audits them, so hold them more loosely.

4.5 Critical estimates and judgments: where the accounting is honest about being a guess

IFRS and Ind AS require a dedicated note on significant judgments and major sources of estimation uncertainty (Ind AS 1 paras 122/125). US filers put Critical Accounting Estimates inside MD&A. Either way, the company is telling you in plain sight that these numbers could be materially different within a year.

Read it as an inventory of soft spots. The usual cast: percentage-of-completion margins; goodwill and intangible impairment assumptions; ECL provisioning; useful lives; deferred-tax-asset recoverability (the "future taxable profits" bet you met with liabilities and financing); litigation provisions; pension assumptions; fair-value Level 3 inputs.

Two moves pay off most:

• Hunt for headroom. Impairment notes often disclose the cushion, as in "the recoverable amount exceeds carrying value by ₹310 crore". When required, they add a sensitivity too, such as "a 1% increase in the discount rate would reduce headroom to nil." When you see the phrase "a reasonably possible change in assumption would result in impairment", you're looking at a formal, auditable early warning. Log it.
• Cross-reference with KAM/CAM. The auditor's key-matters list, covered ahead, is nearly always a subset of this note. When both management and the auditor flag an estimate, your Phase 2 forensic work will start there.

4.9 BRSR: sustainability disclosure, analyst's dosage

The Business Responsibility and Sustainability Report is SEBI-mandated ESG disclosure for the top 1,000 listed companies by market cap, mandatory from FY2022-23. It's structured on the nine NGRBC principles. Section A is general (operations, employees, grievances), Section B covers management processes, and Section C covers principle-wise performance. BRSR Core is a subset of ~key performance indicators. It carries an assurance/assessment requirement being phased down from the top 150 companies toward the top 1,000. SEBI has revised the glide path and softened "assurance" to "assessment or assurance", so check the current SEBI circular before quoting specifics (as of mid-2026).

How much of it do you need? For an analyst, BRSR is awareness-level material. Here's what you can actually pull out of it:

• Employee data: headcount, turnover rates, contractual-vs-permanent mix (cross-check MD&A claims about "talent"), median pay and gender pay data.
• Resource intensity: energy and water per unit of output, for sectors where that's a cost (cement, metals, textiles). Its trend line tells you about future opex.
• Safety: LTIFR/fatalities for industrials, a leading indicator of how tightly operations are run.
• Grievances and penalties: Section A tables of complaints and regulatory fines.

Treat glossy ESG prose with the same skepticism as MD&A adjectives. And don't confuse BRSR, which is disclosure, with CSR. CSR is the s.135 Companies Act requirement to spend 2% of average net profits on listed causes. It's a real cash line, reported in the Board's Report and policed by CARO clause xx.

4.10 The proxy statement (DEF 14A): and its Indian equivalents

The DEF 14A is the document shareholders get before the annual meeting, and Part III of the 10-K usually just points to it. It's the single best US source on Q3 of the five questions, the one that asks who runs this and what they're paid to do. Here's its anatomy, followed by the three reads.

The document runs in a fixed order. Meeting proposals come first: director elections, auditor ratification, say-on-pay, shareholder proposals. Then the board section, with bios, independence, committees and meeting attendance. Then the CD&A (Compensation Discussion & Analysis). It carries the pay philosophy. Most important for you, it sets out the metrics and targets behind bonuses and performance shares. Compensation tables follow. Then beneficial ownership, covering all ≥5% holders plus directors' and officers' holdings. Then certain relationships and related transactions, where Item 404 of Reg S-K requires any related-person transaction above $120,000 to be described. Audit fees close it.

The Summary Compensation Table columns run left to right: salary, bonus, stock awards (grant-date fair value, linking back to last week's SBC mechanics), option awards, non-equity incentive plan compensation (the "bonus" that had a formula), change in pension value, all other compensation (planes, security, gross-ups), and total. There are two newer companions. One is the CEO pay ratio, CEO total comp ÷ median employee, required since 2018. The other is the Pay-versus-Performance table (Item 402(v), since the 2023 proxy season), which lines up "compensation actually paid" against TSR, net income and a company-chosen measure.

Read the proxy three ways.

1. What behavior is being purchased? List the incentive metrics and weights, then interrogate them. Value metrics (ROIC, FCF/share, relative TSR) buy different behavior from size metrics such as revenue and EBITDA. Size metrics reward empire-building and M&A at any price. Ask whether the basis is GAAP or "adjusted". Adjusted-EBITDA bonuses plus aggressive add-backs are management grading its own homework. Ask whether targets are per-share or absolute, since absolute targets are met by issuing shares. And check where thresholds are set against history. A "target" below last year's actual is a gift dressed up as a goal.
2. Skin in the game. Compare what executives own, from the ownership table, with what they're paid each year. A CEO owning 10× annual comp in stock thinks like an owner. One owning 0.3× with mega annual grants thinks like a renter. Check share pledging by US insiders too. Many proxies disclose anti-pledging policies, and the exceptions to them.
3. The related-party section. Family members on the payroll, entities owned by directors leasing property to the company, "consulting agreements" with former founders. Small dollars, large signal.

India has no proxy statement, so the same material is spread across five places. The Board's Report carries remuneration ratios under s.197, setting each director's pay against the median employee. It also shows the cap of 11% of net profits on managerial remuneration for public companies, beyond which shareholders must approve. Committees and independence, plus auditor fees, are in the Corporate Governance Report. The AGM notice carries the actual resolutions, and its explanatory statements are where RPT and ESOP approvals get argued. ESOP disclosures fall under the SEBI SBEB regulations, and the annual return/shareholding data completes the set. India's compensation numbers are usually tamer than the US. Here the governance risk concentrates in ownership structure instead, and the next two subsections deal with it.

4.12 The India stack: Ind AS, Schedule III, LODR, promoters and pledges

You've met the pieces. Now put the map together. Five layers govern what an Indian listed company shows you:

Layer 2: Schedule III (the presentation law). Schedule III of the Companies Act prescribes statement formats. Division I covers old GAAP companies, Division II covers Ind AS companies and is your default, and Division III covers NBFCs. Three format facts will hit you the moment you spread:

• Order is inverted vs US practice. The balance sheet runs non-current before current, and equity comes first on the liabilities side.
• The P&L is by nature, with no gross-profit or EBIT subtotal. Expenses appear as cost of materials consumed, purchases of stock-in-trade, changes in inventories, employee benefits expense, finance costs, depreciation and amortization, and other expenses. You must construct COGS, gross profit, and EBIT yourself. That's the single biggest mechanical difference from spreading a 10-K, and the capstone protocol ahead gives the mapping rules.
• "Other income" is part of total income. It holds interest, dividends, fair-value gains, and sometimes real operating items such as export incentives and scrap sales. Separating operating from non-operating other income is a judgment you make and log.

The 2021 Schedule III amendments (effective FY22) turned the notes into a small forensic toolkit. They made ageing schedules mandatory for trade receivables, trade payables, CWIP and intangibles-under-development. They also force disclosure of title deeds not held in the company's name, a promoter shareholding table with % change, transactions with struck-off companies, wilful-defaulter declarations, benami disclosures, undisclosed income and crypto holdings. Mirroring LODR, they require eleven ratios too, with explanations for any change >25%. The eleven are current, debt-equity, DSCR, ROE, inventory turnover, receivables turnover, payables turnover, net capital turnover, net profit, ROCE and return on investment. India now hands you, by law, the ageing and ratio drift that US analysts reverse-engineer.

Layer 3: SEBI LODR (the listing rulebook). The Listing Obligations and Disclosure Requirements Regulations, 2015, are the operating system of being listed. These are the regulations you'll actually touch:

• Reg 33: quarterly results within 45 days (limited review), annual audited within 60. Results must state whether the audit opinion is modified, with a Statement of Impact of Audit Qualifications when it is.
• Reg 23: related-party transactions. All RPTs need audit-committee approval, with omnibus approvals allowed for repetitive ones. A material RPT needs prior shareholder approval, and all related parties are barred from voting, which makes it effectively a majority-of-the-minority vote. The materiality threshold is 10% of annual consolidated turnover up to ₹20,000 crore of turnover, ₹2,000 crore plus 5% of the excess up to ₹40,000 crore, and ₹3,000 crore plus 2.5% of the excess beyond that, capped at ₹5,000 crore. That's per the SEBI LODR Fifth Amendment 2025, as of October 2026. Before it, the trigger was the lower of ₹1,000 crore or 10%. Brand-usage/royalty payments to related parties face a stricter 5%-of-consolidated-turnover materiality trigger (thresholds amended repeatedly since 2019). Listed entities must disclose their full RPT ledger to the exchanges half-yearly. The regulation exists because promoter groups historically drained listed companies through service fees, royalties, loans and asset sales. Reg 23 gives minority shareholders a veto, if they read the notice.
• Reg 30: material events within 24 hours/12 hours (rating changes, fraud, auditor resignations, key personnel exits…). It's the Indian 8-K.
• Reg 31: the shareholding pattern, quarterly. It shows promoter/promoter-group vs public holdings, the FII/DII breakdown, and the pledge/encumbrance table for promoter shares.
• Schedule V: contents of the annual report, the MD&A ratio rule described above, the CG report, and remuneration details.

Layer 4: promoter shareholding and pledging (the India-specific risk). A promoter is the controlling person/family/group. It's a formal, disclosed category under Indian securities law with no true US analog, though founder/dual-class control is the nearest cousin. Roughly three-quarters of Indian listed companies have a promoter at the wheel, and the median promoter stake is around half the company. With that much concentration, Indian governance analysis starts with the shareholding pattern. The board page comes later.

Pledging is the promoter borrowing personally against their shares. The disclosure trail starts with SEBI SAST Reg 31, which requires promoters to disclose pledge creation, invocation and release within days to the company and the exchanges. Encumbered shares then show up in the quarterly shareholding pattern. Since 2019, if combined encumbrance crosses 50% of the promoter's holding or 20% of total share capital, promoters must disclose detailed reasons (check thresholds against the current SAST text). Aggregators like screener.in show "pledged %" on every company page, but the primary trail is BSE/NSE.

How can the promoter's personal loan kill the company? Through the pledge spiral.

1. The stock falls, so collateral value falls. Lenders then demand a top-up in more shares or cash.
2. The promoter pledges more, and the free float of unpledged promoter shares shrinks.
3. A further fall and the lender invokes, seizing and selling pledged shares. Forced selling speeds up the fall. The promoter loses control, or bleeds the company to service personal debt through upstreamed dividends and related-party loans. That's CARO clause iii again, closing the loop.

The canonical case is Zee Entertainment / Essel Group. By late 2018 the promoters had pledged the large majority of their ~40%+ stake against group infrastructure debts. On 25 January 2019 the stock fell roughly a quarter in a day and lender standstills followed. In the end the promoters sold most of the company to repay pledge debt, and their stake ultimately fell toward ~4%. Those figures are approximate, from public reporting. Pull the shareholding-pattern history on BSE and watch it happen quarter by quarter, and Worked Example 4 does the math. The Street uses a few rules of thumb. Calibrate them and don't worship them. Pledging above ~20–25% of the promoter's holding means you investigate the why. Above ~50% it's a standing risk factor that belongs in your thesis. And pledging that rises while the stock falls means the spiral may already be turning.

Layer 5: Companies Act 2013 (the chassis). You've already met it piecemeal: books and audit (s.139–147), CARO under s.143(11), dividends from standalone profits (s.123), loans to directors (s.185), CSR (s.135), remuneration caps (s.197), the Board's Report (s.134). MCA21 is the registry where even unlisted group entities file. Phase 2 teaches you to pull a promoter's private-company filings from it when the related-party note smells.

4.14 The first-pass statement read: 45–60 minute checklist, v1

The capstone also seeds this artifact, and it answers a different question. The capstone project builds the full record. The checklist helps you form a structured first impression of the filing in under an hour. You'll run it weekly (company-of-the-week) from now on, and Phase 8 hardens it into the final rapid-analysis checklist. Here's version 1:

Before the clock starts, get the right documents in front of you. For the US that's the latest 10-K + DEF 14A. For India it's the latest AR (consolidated section) + latest quarterly shareholding pattern. Nothing else.

Minutes 0–5, identity and integrity scan:

• What does it sell, to whom, at what scale? One line, from Item 1's first paragraph or the corporate overview.
• FY-end, currency, units, consolidated?
• Auditor: name, then opinion type, then any going-concern language. Finish with the number and titles of KAMs/CAMs.
• India: does any CARO clause carry an exception? Note promoter holding % and pledge %.

Minutes 5–18, the statements, shape only:

• Revenue and EBIT, 5 years if in hand (else 2): direction and a one-line story.
• Gross and EBIT margin direction. Exceptional items in any year?
• PAT vs CFO, all years shown: do profits become cash? If CFO stays below PAT year after year, that's a Phase 2 alarm.
• Balance sheet: net debt (build it in your head as B18+B22−B1−B2), equity size, and the two biggest asset lines. Do they make sense for this business model? A "services" company where loans-given is a top-3 asset isn't a services company.
• Share count, 5 years: issuing or retiring?

Minutes 18–30, notes triage, nine stops:

1. Policies: anything changed vs last year (the diff trick)?
2. Segments: which segment is the profit engine? It may not be the revenue engine.
3. Related parties: total value of RP sales/purchases/loans vs size of company.
4. Contingent liabilities: total vs net worth. Above 15–20% of net worth, read the detail.
5. Debt: maturity wall inside 24 months? Rate mix? Covenant mentions?
6. The note behind the biggest KAM/CAM.
7. Employee benefits: funded status (plan assets − obligation). Add a material deficit to net debt before you quote leverage.
8. Foreign currency: which method the foreign operations use, translation or remeasurement, and how much of this year's reported growth came from currency and how much from volume or price.
9. Subsequent events: what happened between the year-end and the approval date, and whether the company classified each item as adjusting or non-adjusting.

Minutes 30–40, narrative and auditor, adversarially:

• MD&A: this year's three biggest claims, and whether last year's promises were kept. India: in the Schedule V ratio table, which ratios moved >25%, and does the explanation name a mechanism or wave at "market conditions"?
• Risk factors: the two specific ones (numbers, names, dates). Ignore boilerplate.
• Auditor's report in full (it's two pages): opinion wording, KAM detail, EOM paragraphs.
• India: the CARO 30-second scan of clauses (ii)(b), (iii), (vii), (ix), (xi), (xvii), (xviii), (xix), and (xxi) on the consolidated report.

Minutes 40–50, ownership, incentives, governance:

• US: the DEF 14A's comp metrics and their GAAP/adjusted basis. Then CEO ownership vs annual pay, and the related-party section. Add any 8-K auditor/CFO changes during the year.
• India: the shareholding pattern trend (promoter %, FII/DII, pledge history 4–8 quarters), Board's Report remuneration ratios, the AGM notice for any RPT or ESOP resolutions, and auditor tenure. Is a rotation year approaching?

Minutes 50–60, write the 10-line output, always, with no exceptions:

1. What it does and its scale (revenue, market position, one line).
2. Five-year shape: revenue growth, margin direction.
3. Cash reality: CFO vs PAT, capex weight, FCF sign.
4. Balance sheet state: net debt/equity, the one scary line if any.
5. The profit engine (segment) and its direction.
6. Accounting quality: policy flags, auditor flags, CARO flags.
7. Ownership & incentives: promoter/insider stake, pledge, what pay rewards.
8. Non-GAAP wedge, if the company leads with an adjusted number.
9. The two risks that actually matter.
10. Three questions I cannot yet answer, and which document would answer them.

One accelerator works on any filing PDF or HTML. Ctrl-F for "related party", "pledge", "contingent", "going concern", "restated", "resign", "exceptional", "one-time", "guarantee", "material weakness", "unbilled". Each hit earns 30 seconds of attention. Two stop rules apply. An adverse opinion or disclaimer stops the analysis, because the accounts aren't a usable input and that finding is the analysis. A pledge >50% of promoter holding moves governance to the top of your question list.

Common mistakes & how experts think differently

1. Reading the glossy and calling it the filing. The chairman's letter and "value creation model" are marketing. The statutory reports, notes, auditor's report, and CARO are the filing. Experts start from the back (statements, then notes, then auditor) and read the glossy last, if at all, as a record of claims to test.
2. Treating a clean opinion as a clean company. An unmodified opinion means the statements pass a materiality-bounded, sample-based examination. Satyam and Wirecard carried clean opinions into the abyss. Experts treat the audit as a floor and do their own forensic pass (Phase 2).
3. Confusing KAM/CAM and EOM with qualifications. A KAM doesn't modify the opinion. It discloses where the audit worked hardest. An EOM highlights a disclosed matter without modifying the opinion. Only "qualified/adverse/disclaimer" modify. Getting it wrong in either direction marks an amateur, whether that's panic at a KAM or a shrug at an actual qualification.
4. Spreading from press releases, decks, or screeners. Press-release tables are unaudited and curated. Decks are advocacy. Screeners embed silent mapping choices. Experts spread from Item 8 / the audited financial section, then reconcile the screener to their spread. It never goes the other way.
5. Netting and dumping. Shoving unclear items into "other" without a log, or netting gains against charges, quietly corrupts the derived panel (margins, days, ROCE) that Phase 2 will compute. Experts keep an auditable mapping log, so their spread is evidence.
6. Ignoring the standalone statements (India). Dividends are paid from standalone profits, and loans to group companies show up in the standalone notes and CARO. If you read only consolidated, you miss the cash trap and the diversion trail.
7. Computing Indian margins with "other income" inside. Treasury income inside EBIT flatters operating margin, sometimes by hundreds of basis points at cash-rich companies. Experts split other income at entry (mapping rule 5) and can quote both margins.
8. Taking the adjusted number on faith, or rejecting it on reflex. Both are lazy. Experts rebuild the bridge, classify each add-back (A/B/C/D), and then decide which earnings series describes the economics. Sometimes adjusted really is better, and rebuilding it is how you find out.
9. Reading promoter holding without the pledge table. "Promoters own 55%" reads as alignment. "55%, of which 70% pledged" reads as a margin call in waiting. Experts read past the headline stake to the encumbrance table and its four-quarter trend.
10. Believing MD&A is information. It's advocacy under legal constraints. Experts read it against prior years (the promises ledger) and against the numbers. The value is in the diffs and the omissions.
11. Doing tie-outs "approximately." A balance sheet off by 0.4% "close enough" hides a transposed digit that will surface as a wrong ratio in Phase 2 and a wrong model in Phase 3. Experts get Tier-1 checks to exactly zero and treat every residual as a message.

One difference runs underneath all of these. Amateurs read filings to extract numbers. Experts read filings to interrogate the people who wrote them. Every disclosure was drafted, reviewed and lawyered by humans with incentives, and those incentives leak through the narrative layer.

Worked examples

All four companies below are synthetic (numbers invented for clean arithmetic), modeled on real disclosure patterns.

Worked example 1: Running the self-check protocol (and catching two reconciling items)

FY25 spread extracts for Meridian Consumer Products Inc. ($m):

Income statement: Revenue 8,400 · COGS 5,040 → Gross profit 3,360 (40.0%). Then SG&A 1,850 · R&D 160 · Other opex 50 → EBIT 1,300, and CFS D&A 420 → EBITDA 1,720. Below EBIT sit Interest expense 190 · Interest income 40 · Exceptional (restructuring) −60 → EBT = 1,300 − 190 + 40 − 60 = 1,090. Tax 273 → Net income 817 · NCI 27 → NI to parent 790. Diluted shares 250m; printed diluted EPS $3.16.

Tier-1 checks.

• IS internal: 3,360 − 1,850 − 160 − 50 = 1,300 ✓, EBT recomputes ✓, 817 − 27 = 790 ✓.
• CFS internal: CFO 1,090, CFI −585, CFF −730, FX −15 → net change = 1,090 − 585 − 730 − 15 = −240. Opening cash 820 → closing 580 ✓ internal.
• Cash articulation: FAILS. The balance sheet shows cash & equivalents of 535, where you expected 580. That's a difference of 45, so go hunting. Ctrl-F "held for sale" takes you to Note 2, where a disposal group's assets include $45m of cash, presented within "assets held for sale." Write the reconciliation: C28 580 = B1 535 + 45 HFS cash. Now the check passes as reconciled.

Tier-2 checks.

• EPS: 790 / 250 = 3.16 ✓ (0.0% off printed).
• Equity roll: opening total equity 4,200. Movements: NI 817 · OCI −45 · dividends to shareholders −280 · dividends to NCI −12 · buybacks −350 · SBC +90 · option/ESPP issues +25. Expected closing = 4,200 + 817 − 45 − 280 − 12 − 350 + 90 + 25 = 4,445. Printed closing equity 4,457. Residual +12 = 0.27% of closing equity, within 1%, but the protocol wants a name. The statement of changes in equity shows "remeasurement of put option over NCI +12." Logged ✓.
• D&A triangulation: CFS D&A 420 vs PP&E-note depreciation 305 + intangible amortization 85 = 390. Gap 30. The CFS line label reads "Depreciation, amortisation and impairment", and Note 6 shows a $30m plant impairment inside the restructuring program. Logged ✓. That also splits I13's −60 exceptional. Is 30 of it cash severance in C4? No. The severance was cash and the impairment was non-cash, and the log records both.

The Tier-3 sanity bands come out clean. Implied interest: 190 / average gross debt 2,575 = 7.4%, inside the US band, matching the debt note's 6.9–7.8% coupons ✓. ETR: 273/1,090 = 25.0% vs ~21% federal + state ✓. So the spread passes with three named reconciling items, and that's what a passing spread looks like. Real filings almost never tie silently. They tie explainably.

Worked example 2: Policing a non-GAAP bridge

The figures for Cloudline Inc. (synthetic SaaS, $m):

Management's headline: "non-GAAP operating margin expanded 700bps to over 20%."

Step 1: verify the bridge arithmetic. FY25: −40 + 340 + 100 + 30 + 20 = 450 ✓. Always recompute. Errors happen, and they're never in the company's disfavor.

Step 2: classify each add-back. SBC goes in bucket C. It's a real, recurring cost of labor, at 340/2,200 = 15.5% of revenue. Amortization of acquired intangibles is C here too, because Cloudline acquires every year and the acquisition-cost line proves it. Excluding the amortization pretends the growth was free. Restructuring is B, since three consecutive years make it a run-rate cost in a costume. Acquisition costs are B for a serial acquirer.

Step 3: the wedge. Cumulative non-GAAP operating income = 215 + 330 + 450 = 995. Cumulative GAAP = −340. Wedge = 1,335 over three years. The adjusted series and the GAAP series describe different companies. SBC alone is 900 of it (67%).

Step 4: per-share reality. Share count: +8.3% over two years (~4%/yr). So SBC is an expense dodge and, more urgently, live dilution. Even if you accepted non-GAAP operating income, per share it grew ~16 points less than the headline suggests: 450/325 vs 215/300 = +93% total, against +109% unadjusted for shares. And on GAAP numbers the company still loses money.

Verdict (write it like this): "GAAP operations ≈ breakeven and improving. The entire reported 'profitability' is add-backs, two-thirds SBC. I will analyze Cloudline on GAAP operating income plus a memo line for FCF-after-SBC-dilution. The improving trajectory is real. The level is a fiction."

Worked example 3: A qualified opinion plus CARO, quantified

Suvarna Infratech Ltd is synthetic, reported in ₹ crore: PBT 520 · PAT 388 · net worth 2,850 · total loans given 559.

From the standalone auditor's report and CARO annexure:

Basis for Qualified Opinion: trade receivables include ₹412 crore due from Suvarna Devcon LLP, an entity in which a director is interested, outstanding for over three years, against which no loss allowance has been recognized. We were unable to obtain sufficient appropriate evidence regarding recoverability. CARO (iii)(f): loans aggregating ₹380 crore, repayable on demand, granted to promoter-group entities (68% of total loans granted). CARO (ix)(d): short-term funds of ₹610 crore have been utilised for long-term purposes. CARO (vii): undisputed statutory dues (GST, provident fund) of ₹18.4 crore outstanding for more than six months.

Step 1: restate as if the auditor is right. Provide in full against the ₹412 crore receivable and PBT falls to 520 − 412 = 108. The tax benefit is doubtful. Leaving aside the timing of a loss allowance's deductibility, recognizing a DTA requires probable future recovery, so take the conservative case of no current benefit. Restated PAT ≈ 388 − 412 = −24. With full tax benefit at 25.17%, it's ≈ 388 − 308 = 80. Either way, the reported ₹388 crore becomes ₹−24 to ₹80 crore. Once the disputed receivable is provisioned, the company earns roughly nothing. Net worth: 2,850 − 412 = 2,438 (−14.5%).

Step 2: aggregate the promoter-side exposure. Disputed receivable 412 + on-demand loans to promoter entities 380 = ₹792 crore. That's 27.8% of reported net worth (32.5% of restated) resting with the promoter's side of the fence.

Step 3: corroborate the stress. CARO (ix)(d) shows ₹610 crore of short-term money funding long-term assets. That's refinancing dependence, the IL&FS/DHFL failure mode. CARO (vii) shows PF and GST unpaid for six-plus months, which is what companies do when the till is empty. ₹18.4 crore is small, but the signal is about behavior. Materiality doesn't come into it.

Step 4: the write-up line and next actions. "Reported PAT 388. Economically ≈ 0 after the qualified receivable. ~28% of net worth is promoter-side exposure. Funding is maturity-mismatched. Statutory dues are aging. Next: pull Suvarna Devcon LLP's filings from MCA21, check whether the ₹380 crore of loans went through Reg 23 shareholder approval, and pull the pledge table." That's the core skill of the whole week. The auditor and CARO handed you a forensic case file, and all you had to do was read the annexure.

Worked example 4: The pledge spiral, computed

Kanchan Media Ltd is synthetic, modeled on the Zee/Essel pattern. Shares outstanding 96 crore · price ₹400 · market cap ₹38,400 crore. Promoter stake 42% = 40.32 crore shares, of which 90% pledged = 36.29 crore shares.

Quote a pledge on both bases, always, because they answer different questions. As a share of promoter holding it's 90%, which measures how much of the family's own position is spoken for. As a share of total capital it's 36.29 ÷ 96 = 37.8%. That's the number that matters to everyone else, because control changes hands on the total-capital base. India's takeover code sets its open-offer trigger at 25% of voting rights (SEBI SAST 2011, Regulation 3(1); perishable, verify against the current regulations). So this single pledged block is 1.5 times the size of the stake that would oblige a buyer to bid for the company. Whoever ends up holding it after an invocation holds a controlling position, and the promoter is left with 4.03 crore shares, 4.2% of the company. Quote the pledge only as a percentage of promoter holding and all of that disappears. 90% of a 5% stake is a private problem. 90% of a 42% stake is a change-of-control event waiting for a price fall.

• Collateral value: 36.29 crore × ₹400 = ₹14,515 crore.
• Lenders lend at 2.0× cover: promoter debt ≈ 14,515 / 2 = ₹7,258 crore. Top-up trigger: cover < 1.75×.
• Stock falls 25% to ₹300. Collateral = 36.29 × 300 = ₹10,886 crore → cover = 10,886 / 7,258 = 1.50×. Trigger breached.
• Top-up to restore 1.75×: need collateral of 7,258 × 1.75 = ₹12,701 crore → shortfall ₹1,815 crore = 6.05 crore more shares at ₹300.
• Promoter's unpledged shares: 40.32 − 36.29 = 4.03 crore. That isn't enough. Pledging all of them adds ₹1,209 crore and still leaves ₹606 crore short. The promoter must find cash or sell assets, or the lenders invoke.
• Invocation math: selling 36 crore shares = 37.5% of the entire company hitting a falling market. The price gaps down, which breaches other lenders' covers and forces more selling. That's the spiral.

Here's why company analysis cares about a personal loan. A promoter in this vise controls the company's cash. Watch for upstreaming: special dividends the company can't afford, brand-royalty payments (Reg 23's 5% trigger exists for this), loans to promoter entities (CARO iii). The real Zee episode ran this arc through 2019–20. There was a majority-pledged promoter stake, a one-day ~25% price fall, lender standstills, and finally the promoters selling down from ~40% to single digits to repay pledge debt. Those figures are approximate, from public reporting. Pull ZEEL's quarterly shareholding patterns on BSE and trace it yourself. It's the best free case study in pledge dynamics.

Worked example 5: Splitting "other income" on a by-nature Indian P&L (and what it does to the EBIT margin)

Saptagiri Consumer Ltd is synthetic, in ₹ crore. Its Schedule III (Division II) statement of profit and loss prints expenses by nature, with no gross-profit or EBIT line. You have to build both, then decide what belongs inside "operating." Here's the FY0 extract:

The other-income note splits the ₹520 crore into: interest on bank deposits ₹180 cr, dividend and fair-value gains on investments ₹200 cr, export incentives ₹90 cr, scrap sales ₹50 cr.

Step 1: construct COGS and gross profit (mapping rule 4). COGS = materials consumed + purchases of stock-in-trade + changes in inventories = 6,800 + 900 + (−300) = ₹7,400 cr. The negative changes line means inventory grew, so that cost was deferred to the balance sheet and wasn't incurred this year. Gross profit = 12,000 − 7,400 = ₹4,600 cr → 38.3% gross margin. Neither number appears anywhere on the face.

Step 2: the operating-cost block. Below gross profit come employee benefits 1,400, other expenses 1,720 and depreciation 540, all operating. Finance costs of 260 are non-operating and belong below EBIT. So the margin comes down to one question: how much of "other income" is operating?

Step 3: split "other income" (mapping rule 5). Interest (180) and dividend/FV gains (200) are non-operating and go to I11: ₹380 cr. Export incentives (90) and scrap sales (50) are earned by the operating activity itself, so they're operating and go to I6: ₹140 cr.

Step 4: EBIT three ways, and only one of them is right:

• Correct (split): EBIT = 4,600 − 1,400 − 1,720 − 540 + 140 = ₹1,080 cr → 9.0% EBIT margin.
• Lazy "leave all other income in operating" (the total-income habit): +520 instead of +140 → EBIT ₹1,460 cr → 12.2%, flattered by +317 bps of treasury income (interest + investment gains) that has nothing to do with selling groceries.
• Over-correcting "dump all other income below EBIT": +0 → EBIT ₹940 cr → 7.8%. Now you've thrown away ₹140 cr of real operating income and understated the margin by 117 bps.

Step 5: the tie that proves it. All three treatments recompute to the same PBT of ₹1,200 cr (1,080 − 260 + 380 = 1,460 − 260 + 0 = 940 − 260 + 520 = 1,200). Splitting never touches the bottom line. It only moves the operating/non-operating boundary, and with it the EBIT margin. So this number is a judgment you log, and you won't find it printed anywhere. Quote 9.0%, and be ready to show the 12.2% you refused to claim. That's common mistake #7, in numbers.

Worked example 6: Cross-footing a segment note to the face, and finding the profit engine (US)

Northwind Industrial Corp is synthetic, in $m. The face shows consolidated revenue $7,200m and operating income (EBIT) $1,040m. The segment note reports three segments with inter-segment sales, plus eliminations and unallocated corporate cost:

Step 1: cross-foot revenue (Tier-3 check 13). Σ total segment revenue 7,700 − inter-segment eliminations 500 = 7,200 = the face revenue. Put another way, the external-revenue column already sums to 7,200, because the eliminations are the inter-segment column. If this fails to tie, you dropped the elimination line, which is the commonest segment-note error.

Step 2: cross-foot EBIT. Σ segment EBIT 1,250 − inter-segment profit elimination 30 − corporate/unallocated 180 = 1,040 = the face operating income. Those two reconciling items, unallocated corporate cost and intra-group profit, are why segment profits never sum to the total. Naming them is the tie.

Step 3: revenue engine ≠ profit engine (notes-triage stop 2, covered above). Flow Control is the revenue leader ($3,500m, 45% of segment revenue), but Process Automation earns the higher margin, 19.6% vs 17.7%. Together the two carry 93% of segment EBIT (49.6% + 43.2%). Industrial Services is ~19% of segment revenue and only 7.2% of segment profit (a 6.2% margin). You study that segment to ask whether it earns its capital at all. Its top line isn't the point.

Step 4: the takeaway. The consolidated 14.4% EBIT margin hides two ~18–20% businesses subsidising a ~6% one, net of a fat corporate drag. When you spread this company, carry N1 (segment revenue) and N2 (segment EBIT) for five years and watch the mix move. Group-level margin drift is often just the profit engine gaining or losing share, with no segment's economics changing at all.

Practice set

Work each problem fully before you open its solution (Working Agreement 3). Problems 1–3 are guided, 4–10 independent, 11–12 timed.

P1 (guided). Classify the opinion. For each vignette, name the opinion type and the analyst response. (a) "In our opinion, the financial statements present fairly… We draw attention to Note 42 regarding the estimation uncertainty arising from pending litigation. Our opinion is not modified in respect of this matter." (b) "…except for the effects of the matter described in the Basis section, inventory of ₹240 crore for which we could not observe the physical count, the statements present fairly…" (c) "Because of the significance of the matters described, we were denied access to the accounting records of three material subsidiaries, we do not express an opinion…" (d) "In our opinion, because of the significance of the matters discussed, revenue has been recognized on contracts that do not meet the criteria of Ind AS 115, affecting substantially all reported periods, the statements do not present fairly…"

Solution. (a) Unmodified with an Emphasis of Matter. It's a clean opinion, and the EOM directs you to a disclosed note. Response: read Note 42 and quantify the litigation vs net worth. No restatement needed. (b) Qualified ("except for"), a material but not pervasive scope limitation. Response: restate mentally as if the ₹240 crore inventory were impaired, and ask why the auditor couldn't attend a count. (c) Disclaimer: no opinion possible. Denial of access is caused by management, which is worse than one caused by circumstance. Response: the accounts are unusable, and the denial is the finding. (d) Adverse: pervasive misstatement. Response: stop. Treat the accounts as uninvestable-grade until re-audited/restated.

P2 (guided). KAMs as a map. An Indian engineering company's auditor lists three KAMs: (1) "Revenue recognition on long-term construction contracts: estimation of costs to complete"; (2) "Recoverability of deferred tax assets on unabsorbed losses"; (3) "Valuation of investment in a joint venture following an adverse arbitration outcome." For each, which note do you open, and what specifically would you stress-test? And has the auditor qualified the opinion?

Solution. No. KAMs aren't qualifications, and the opinion here can be perfectly clean. (1) Open the revenue note + contract balances (contract assets/WIP). Stress the margin assumed on open contracts, since a small change in cost-to-complete swings percentage-of-completion profit. Check whether contract assets are growing faster than revenue, a billing-lag flag that Phase 2's earnings-games work names. (2) Open the tax note. What future taxable profits are assumed, and over what horizon? That's the "probable" test from your work on liabilities and financing. A company recognizing DTAs while losing money is booking optimism as an asset. (3) Open the investments/JV note + contingencies. Is the carrying value still supported post-award? Is there a guarantee to the JV's lenders? Check the RPT note + CARO iii.

P3 (guided). CARO clause mapping. Match each finding to its CARO 2020 clause and state the signal: (a) quarterly stock statements filed with banks differ from the books, (b) fresh loans granted to a group entity to repay its overdue loan, (c) GST of ₹9 crore unpaid for eight months, (d) short-term borrowings funded a new plant, (e) the previous auditor resigned and the new auditor comments on the predecessor's concerns, (f) the company has incurred cash losses in both the current and preceding year.

Solution. (a) (ii)(b): drawing power inflated with lenders. Books-vs-bank divergence is the DHFL-era smoking gun. (b) (iii)(e): evergreening. The loan book to related parties is rotting and being repainted. (c) (vii): acute cash stress. Companies stiff the government last-but-one, with employees last. (d) (ix)(d): ALM mismatch and refinancing risk (IL&FS failure mode). (e) (xviii): it forces the exit story into the record, so read it alongside the Reg 30 resignation disclosure. (f) (xvii): a blunt distress marker. Cross-check with going-concern language and clause (xix).

P4. The LODR Schedule V ratio table. An Indian company reports (FY24 to FY25): debtors turnover 6.1 to 4.2, inventory turnover 5.0 to 4.6, interest coverage 8.2 to 5.9, current ratio 1.9 to 1.6, debt-equity 0.35 to 0.52, operating margin 14.8% to 13.9%, net margin 9.6% to 6.8%. (a) Compute the % change in each and flag which require explanation (≥25%). (b) Management explains the debtors-turnover change as "strategic extension of credit to deepen dealer relationships in a competitive market." Grade the explanation.

Solution. (a) Debtors turnover −31.1% flag · inventory turnover −8.0% · interest coverage −28.0% flag · current ratio −15.8% · debt-equity +48.6% flag · operating margin −6.1% (relative change of the ratio) · net margin −29.2% flag. Four ratios need explanations. (b) Weak to alarming. It names a mechanism (longer credit) but frames a deterioration as strategy. "Deepening dealer relationships" with stretched credit is the classic channel-stuffing setup that Phase 2's earnings-games work names, with sales pulled forward by financing the channel. Cross-check the receivable days trend in your spread, contract assets, Q4 revenue skew and CARO (ii)(b). And notice the cluster: receivables slower + coverage down + leverage up + margins down = borrowing to fund receivables. The table just handed you Phase 2's opening hypothesis.

P5. Reg 23 materiality. Consolidated turnover ₹8,400 crore. Proposed with promoter-side entities this year: (a) purchases of ₹900 crore from a promoter company, (b) brand royalty of ₹450 crore to the parent, (c) a ₹120 crore loan to an associate. What approvals does each need?

Solution. Start with the material-RPT threshold. Turnover of ₹8,400 crore falls in the first band, so 10% × 8,400 = ₹840 crore. (a) 900 > 840, so it's material. It needs audit-committee approval + prior shareholder approval with related parties barred from voting. (b) The royalty/brand test is 5% × 8,400 = ₹420 crore. 450 > 420, so it's material under the royalty trigger despite being < 840, and it takes the same shareholder-approval route (check the current threshold text). (c) 120 crore isn't material by size, but audit-committee approval is still mandatory. Transactions with the same counterparty also aggregate across the financial year, so if the associate gets another ₹750 crore later, the aggregate crosses. And a loan to an associate shows up in CARO (iii) and the RPT note regardless. Approval doesn't make it invisible.

P6. Standalone vs consolidated. Parent standalone revenue ₹6,000 crore (including ₹1,500 crore of sales to subsidiaries). Subsidiaries' external revenue ₹4,500 crore. Standalone PAT ₹800 crore (including ₹150 crore dividends received from subsidiaries). Consolidated PAT ₹1,240 crore. (a) Compute consolidated revenue. (b) Explain why you cannot add standalone PATs across a group. (c) The company wants to pay ₹500 crore of dividends. Name the number that governs capacity. (d) Name where you look for money the parent has pushed into subsidiaries.

Solution. (a) 6,000 − 1,500 (intercompany eliminated) + 4,500 = ₹9,000 crore. (b) Double-counting. Intercompany sales/profits and intra-group dividends would count twice, and the ₹150 crore is the subsidiaries' profit showing up a second time in the parent's P&L. Consolidation eliminates both, as in last week's consolidation work. (c) Standalone profits/reserves (Companies Act s.123). Standalone PAT is really 650 ex-dividends-received plus the 150 actually upstreamed. The question is whether standalone distributable reserves and cash support ₹500 crore. Consolidated earnings trapped in subsidiaries can't pay parent dividends until they're upstreamed. (d) Standalone loans-and-advances note (B13 of your spread), the RPT note, CARO clause (iii), and guarantees in the contingent-liabilities note.

P7. The wedge. GAAP net income: 120, 95, 60 over FY23–25. Adjusted net income: 210, 225, 240. Add-backs across the three years total: SBC 230, restructuring 90, legal settlements 80. (a) Compute the cumulative wedge and express it vs cumulative GAAP earnings. (b) Classify the add-backs. (c) One-line verdict.

Solution. (a) Cumulative adjusted 675 − cumulative GAAP 275 = wedge 400 = 145% of GAAP earnings. Adjusted grows +7%/yr while GAAP falls −29%/yr. That divergence is the worst pattern. (b) SBC (230, 57% of wedge) is bucket C, a real cost. Restructuring 3 years running is B, recurring. Legal settlements take judgment. One large one-off that really is one-off could be A, but 80 spread over three years looks like B, a cost of how they do business. (c) "Economic earnings are shrinking. The adjusted series is 2.5× reality and widening. Value on GAAP-based earnings power, and treat the adjusted narrative as a management-credibility datum."

P8. Equity roll, solve for the missing item. Opening equity 2,400, NI 320, OCI +15, dividends −110, SBC +45, shares issued +10, closing equity 2,540. Buybacks aren't disclosed separately in your extract. Solve.

Solution. 2,400 + 320 + 15 − 110 + 45 + 10 − X = 2,540 → X = 2,680 − 2,540 = 140 of buybacks. Sanity: cross-check against CFF's "repurchase of shares" line, since treasury purchases are CFF cash-out. Settlement timing can create small differences, so log them if so. That's the power of the equity roll. Any one missing item is recoverable as the plug, and that's also why you're only allowed one plug. With two unknowns the roll proves nothing.

P9. Cash roll: find the 30. CFO 450, CFI −380, CFF −60, FX +5. Opening cash & cash equivalents 180. The balance sheet shows "cash and bank balances ₹225 crore." Does the roll tie?

Solution. Net change = 450 − 380 − 60 + 5 = +15 → closing cash & equivalents = 180 + 15 = 195 ≠ 225. Difference 30. Indian balance sheets commonly present "cash and bank balances" including other bank balances (unclaimed-dividend accounts, margin money/deposits pledged with banks). Those aren't cash equivalents. The CFS ties to the cash & cash equivalents note line (195), and the extra 30 is in "other bank balances." So the roll ties as reconciled, and your log names the 30. The US analog is restricted cash presented separately, with a required reconciliation table in the CFS notes. Rule: always tie the CFS to the note's cash-equivalents definition, whatever the biggest cash-looking line on the face says.

P10. Mapping drill. Assign each item to a template row (and note any log entry): (a) "Purchases of stock-in-trade", (b) "Changes in inventories of finished goods and WIP", (c) "Interest income on bank deposits", (d) "Export incentives received", (e) current maturities of long-term borrowings ₹210 crore, presented within "other financial liabilities", (f) "Unbilled revenue ₹95 crore", (g) right-of-use assets, (h) goodwill impairment of ₹75 crore presented within "other expenses."

Solution. (a) I2, a COGS build component in a by-nature P&L. (b) I2. The inventory-change line completes the COGS construction, and a negative number means inventories grew, with cost deferred to the BS. (c) I11, non-operating and out of EBIT. (d) I6, operating other income, because it's earned by the operating activity of exporting. Log the choice. (e) B18, always. Debt is debt whatever the presentation bucket, so log the reclassification. (f) B3 with a log entry. It's a contract asset, economically a receivable not yet invoiced, and DSO analysis must see it. (g) B9. (h) I13 exceptional, own row, with N9 detail (non-cash). Leave it inside I6 and your five-year "other expenses" trend and EBIT margin both lie.

P11 (timed: 8 minutes). Pledge math. Promoters hold 58% of a company's 40 crore shares, and 65% of the promoter holding is pledged. Price ₹300. Lenders hold collateral at 2.0× cover. (a) Promoter debt implied. (b) Price falls 35%: new cover? (c) Shares needed to restore 2.0×, vs free promoter shares. (d) What % of the company's total share capital is pledged after the top-up, and which disclosure threshold does that trip?

Solution. Promoter shares = 23.2 crore; pledged = 15.08 crore. (a) Collateral = 15.08 × 300 = ₹4,524 crore → debt = 4,524/2 = ₹2,262 crore. (b) Price ₹195 → collateral 15.08 × 195 = ₹2,941 crore → cover 1.30×. (c) Restore 2.0×: need 2,262 × 2 = ₹4,524 crore → shortfall ₹1,583 crore ÷ ₹195 = 8.12 crore shares, exactly the promoters' entire unpledged stake (23.2 − 15.08 = 8.12). They survive this call with zero shares to spare. (d) 23.2/40 = 58% of total capital pledged, far beyond the 20%-of-capital line that (with the 50%-of-holding line) triggers the SAST detailed-reasons disclosure. Verdict: one more downdraft means invocation. Treat control as contested and watch for cash upstreaming.

P12 (timed: 15 minutes). The 10-line output. Facts: Indian consumer-durables company. 5-yr revenue CAGR 14%. EBIT margin 9.8% to 7.2% over 5 years. Cumulative CFO/PAT 0.55. Receivable days 48 to 96. Promoter 51%, pledge 0% to 31% of holding over four quarters. Auditor unmodified, KAM on "measurement of dealer-scheme rebates and returns". CARO (ii)(b): quarterly bank statements differ from books, "not material" per auditor. MD&A attributes the −38% net-margin change to "challenging market conditions". Contingent liabilities 6% of net worth. Write the 10 lines.

Solution. Here's a model answer. Yours should match it in substance, though the wording can differ.

1. Consumer durables, ₹-mid-cap, dealer-distributed, and growing fast on paper (14% CAGR).
2. Growth is bought: EBIT margin down 260bps over the period.
3. Cash reality poor: only 55% of five-year PAT became CFO.
4. Balance sheet funding the gap: receivables doubled to 96 days, so half a year of one quarter's sales is with dealers.
5. Profit engine unclear ex-rebates: the KAM says even the auditor sweats the rebate/returns estimate.
6. Accounting quality: unmodified opinion, but KAM (rebates) + CARO (ii)(b) bank-statement divergence + receivables explosion cluster right where channel-stuffing happens.
7. Ownership: promoter 51% but pledge 0% to 31% in a year. Someone needs money, so watch for upstreaming.
8. No non-GAAP wedge disclosed, but check the investor deck for "adjusted" metrics.
9. Two real risks: dealer-channel saturation unwinding through returns/rebates, and a pledge spiral.
10. Next: aging of receivables (Schedule III table), rebate provision trend vs revenue, four more quarters of shareholding pattern, Q4 revenue skew. Verdict for now: the numbers aren't yet trustworthy enough to value.

P13 (guided). Construct COGS and split other income. Brindavan Foods Ltd (synthetic, ₹ crore) prints a by-nature P&L: revenue from operations 9,000; other income 300 (interest 120, dividend/FV gains 100, export incentives 60, scrap sales 20); cost of materials consumed 5,000; purchases of stock-in-trade 600; changes in inventories of FG/WIP/stock-in-trade +200 (inventories decreased); employee benefits 900; finance costs 150; depreciation 360; other expenses 1,200. (a) Construct COGS, gross profit, and gross margin. (b) Split other income and compute the EBIT margin correctly. (c) What EBIT margin would a lazy "total income" read report, and by how much is it flattered? (d) Prove the work by tying to PBT.

Solution. (a) COGS = 5,000 + 600 + 200 = ₹5,800 cr. A positive changes line means inventory was drawn down, which adds to cost. Gross profit = 9,000 − 5,800 = ₹3,200 cr → 35.6% GM. (b) Operating other income = export incentives 60 + scrap 20 = 80. Non-operating = interest 120 + dividends/FV 100 = 220. EBIT = 3,200 − 900 − 1,200 − 360 + 80 = ₹820 cr → 9.1% margin. (c) Leaving all ₹300 cr of other income in operating gives EBIT 1,040, or 11.6%, flattered by +244 bps of interest and investment income. (d) All roads tie: 820 − 150 finance + 220 non-op OI = ₹890 cr PBT = total income 9,300 − total expenses 8,410. The split moved the margin, never the bottom line.

P14 (guided). Find the profit engine. Cascade Diversified Inc (synthetic, $m) reports consolidated revenue $7,900m and operating income $1,150m. Segment note (total segment revenue / segment EBIT): Aerospace 4,000 / 560; Medical 2,400 / 600; Consumer 1,800 / 150; inter-segment eliminations −300 revenue and −20 EBIT; corporate/unallocated −140 EBIT. (a) Cross-foot revenue and EBIT to the face. (b) Which is the revenue engine, and which the profit engine? (c) What share of segment EBIT does the profit engine carry?

Solution. (a) Revenue: 4,000 + 2,400 + 1,800 = 8,200 total, and − 300 eliminations gives $7,900m ✓. EBIT: 560 + 600 + 150 = 1,310, then − 20 elimination − 140 corporate = $1,150m ✓. Those two reconciling items are why segment profits don't sum to the total. (b) The revenue engine is Aerospace ($4,000m, 1.67× Medical's revenue). The profit engine is Medical, with a 25.0% margin vs Aerospace's 14.0% and the larger absolute EBIT ($600m > $560m). Consumer is an 8.3%-margin also-ran. (c) Medical = 600 ÷ 1,310 = 45.8% of segment EBIT on 29% of segment revenue. Calling Aerospace "the business" because it books the most sales is the trap the profit-engine question exists to catch.

P15. D&A triangulation. Granite Materials Inc (synthetic, $m). The cash-flow statement's add-back line reads "Depreciation, amortization and impairment $640m." The PP&E note shows depreciation of $470m, and the intangibles note shows amortization of $95m. (a) Compute the reconciling gap. (b) What is it, and which template rows should the pieces occupy? (c) What breaks if you set I8 (D&A) = 640 straight off the cash-flow line?

Solution. (a) Gap = 640 − (470 + 95) = $75m. (b) The $75m is a goodwill/asset impairment bundled inside the cash-flow line label. Recurring D&A (I8) is $565m, and the $75m belongs in I13 (exceptional, non-cash) with N9 detail, never inside the D&A add-back. (c) You would overstate recurring D&A by $75m and bury a one-off impairment inside "depreciation." EBITDA (built as I7 + I8) and every D&A-based figure would then carry a phantom $75m of recurring charge, and your clean-vs-dirty EBIT trend would lie. The cash-flow line is a starting point, and the PP&E and intangibles notes are the decoder (mapping rule 1).

P16. Contingent liabilities vs net worth. Halcyon Engineering Ltd (synthetic, ₹ crore): equity attributable to owners (net worth) 4,200. Contingent-liabilities note: disputed tax demands 320; guarantees given on behalf of a 50:50 JV and associates 540; other claims not acknowledged as debt 180; bank guarantees in the ordinary course 120. (a) Total contingent liabilities and their ratio to net worth. (b) Does it clear the read-the-detail threshold? (c) Which single component is highest-signal, and which other notes must you open for it?

Solution. (a) Total = 320 + 540 + 180 + 120 = ₹1,160 cr = 27.6% of net worth. (b) Yes, comfortably above the 15–20%-of-net-worth trip-wire from the checklist above, so read the detail. Logging a total won't do. (c) The ₹540 cr of guarantees to the JV/associates (12.9% of net worth). A guarantee to a related entity can crystallise into real debt and is a classic diversion route. Open the RPT note, CARO clause (iii) and the standalone loans/guarantees note (Rule 2) together. The disputed tax (320 = 7.6%) is a read-the-forum-table item and lower signal. Bank guarantees in the ordinary course are usually benign.

P17. Effective-tax-rate sanity. Deccan Specialty Chemicals Ltd (synthetic, ₹ crore): profit before tax 1,600, total tax expense 288. India's headline corporate rate under s.115BAA is ~25.17%. (a) Compute the ETR. (b) Is it inside the sanity band, and where must the difference be visible? (c) Give one benign and one red-flag explanation.

Solution. (a) ETR = 288 ÷ 1,600 = 18.0%. (b) It's well below the ~25.17% statutory rate, a 7.2-point gap. Tax at statutory would be ₹402.7 cr, so ₹114.7 cr is "saved". That puts it out of band, and the difference must be reconciled line by line in the tax-reconciliation note you met with liabilities and financing. (c) Benign: SEZ/export tax holidays, weighted R&D deductions, or dividend/capital-gains income taxed at lower rates. They recur and they're disclosed, so they can last. Red flag: a one-off recognition of deferred-tax assets on carried-forward losses, a bet on "future taxable profits", propping up this year's PAT and unlikely to repeat. A large unexplained "others" line is another. A low ETR isn't good news until the note tells you why.

P18 (timed: 12 minutes, six parts). (i) By-nature P&L: cost of materials consumed ₹3,400 cr, purchases of stock-in-trade ₹500 cr, changes in inventories −₹120 cr (inventories increased). Construct COGS. (ii) Short-term borrowings ₹300 cr, current maturities of long-term debt ₹210 cr held inside "other financial liabilities", long-term borrowings ₹1,600 cr, cash ₹260 cr, current investments ₹140 cr. Compute gross debt and net debt. (iii) With EBITDA ₹950 cr, compute net debt / EBITDA. (iv) Cumulative adjusted net income $620m, cumulative GAAP $380m. Express the wedge as % of GAAP. (v) Consolidated turnover ₹9,500 cr. State the LODR Reg 23 material-RPT threshold. (vi) Interest expense ₹168 cr on average gross debt ₹2,100 cr. Compute the implied interest rate.

Solution. (i) 3,400 + 500 − 120 = ₹3,780 cr, since a negative changes line means cost deferred to the balance sheet. (ii) Gross = 300 + 210 + 1,600 = ₹2,110 cr, because current maturities are debt (mapping rule 7). Net = 2,110 − 260 − 140 = ₹1,710 cr. (iii) 1,710 ÷ 950 = 1.80×. (iv) (620 − 380) ÷ 380 = 63.2%, deep in "different company" territory. (v) First band, so 10% × 9,500 = ₹950 cr. (vi) 168 ÷ 2,100 = 8.0%, inside the Indian 6–12% band. Scoring: 6/6 in time = fluent. At ≤4, redo Worked examples 5–6 and P13–P14 before the mastery check.

Applied mini-project: the Phase 1 capstone: two full hand-spreads

Don't let the heading fool you. This is the M2 milestone of the entire program, the gate that certifies competency C1 (read the language of business) and the proof that thirteen weeks of accounting became a working skill. You'll hand-spread one full US 10-K and one full Indian annual report into the standard spreading template, pass the self-check protocol on both, and write the one-pager for each. The workbook you produce becomes the historical block of every model you build in Phase 3. The habits you build here are the ones Phase 2 assumes.

Budget: 10–12 hours (6–7 for your first company, 4–5 for the second, since the template does the remembering). Spread it over 4–6 sittings across the week. Tie-out work done tired produces errors that take longer to find than the rest did to make.

Part A: the US spread: Colgate-Palmolive (NYSE: CL)

Colgate is a deliberate choice. It's a ~$20bn-revenue consumer-staples company (illustrative, modeled on ~FY24/FY25 scale, so pull the actual filings), complex enough to teach and clean enough to finish. It carries everything this capstone taught. Geographic segments plus Hill's Pet Nutrition give you the segment-note work, and a non-controlling interest gives the NCI line a face you can visit, because its Indian subsidiary, Colgate-Palmolive (India), is separately listed. Recurring restructuring and productivity charges are bucket-B classification practice, a long buyback history feeds the Item 5 table and the equity roll, and the proxy carries a full CD&A.

Pull (SEC EDGAR: sec.gov → EDGAR → company search "Colgate-Palmolive" / ticker CL):

1. The three 10-Ks the sourcing rule above calls for: FY0 (latest filed), FY-2, FY-4. Save as CL_10K_FY20XX.pdf.
2. The latest DEF 14A (for the M-panel and the one-pager's incentives line).
3. Nothing else. No press releases, no decks.

Watch for (log each in the MappingLog): where D&A hides. It isn't on the P&L face, so you pull it from the CFS and the PP&E note (rows I8/C2). Then come the "Other (income) expense, net" line, split per mapping rule 5, restructuring charges (row I13, gross, with N9 detail), NCI in both the P&L split (I20) and equity (B33), and the pension note's funded status (B25 or B15). Treasury stock is the last one, a large negative equity component in B30. Colgate has repurchased for decades, so B30 is startlingly large against B29, and the equity roll is what makes sense of it.

Part B: the Indian spread: Asian Paints (BSE 500820 / NSE: ASIANPAINT)

Asian Paints is a deliberate choice too. It's a ~₹35,000-crore-revenue consumer company (illustrative, ~FY24/FY25 scale, so pull the actual filings) with a real group structure but no landmines, which makes it the right first Indian spread. It works every India-specific muscle. The by-nature Schedule III P&L means you construct COGS, gross profit, and EBIT yourself, per mapping rule 4. The large investment book generates other income you must split under mapping rule 5, a single judgment that moves the EBIT margin visibly. Consolidated-versus-standalone statements bring subsidiaries (Berger International, Nepal, Ethiopia, home-décor entities) and equity-accounted 50:50 JVs with PPG, which puts last week's one-line consolidation to live use. There is also an NCI line, a CARO annexure to scan, and a promoter group whose holding and (historically negligible) pledge you'll log in the M-panel from the shareholding pattern.

Pull:

1. Three annual reports, per the sourcing rule above: FY0, FY-2, FY-4, from asianpaints.com → Investors → Annual Reports, or bseindia.com → search "Asian Paints" → Corp Information → Annual Reports. With a March year-end, the newest AR may not have reached the site yet, because it arrives with the AGM cycle, typically June–September. In that case FY0 is the latest year with a full published AR, per the timing rhythm described earlier.
2. The latest quarterly shareholding pattern (BSE/NSE → company page → Shareholding Pattern): promoter %, pledge/encumbrance table, FII/DII split.
3. Spread the consolidated statements. Open the standalone only for the three Rule-2 checks (dividend capacity, loans/guarantees to group entities, CARO).

Watch for (log each): revenue from operations vs total income. I1 excludes other income, and confusing the two is the classic first-spread error on Indian filings. Next come the COGS construction and its five-year consistency, GST (revenue is net of it), "changes in inventories" sign conventions, and current maturities of long-term borrowings parked in "other financial liabilities", which mapping rule 7 sends to B18. Then the classification choices. Ind AS puts interest paid in CFF and interest received in CFI (mapping rule 9). Your CFS here starts from PBT, and net income isn't its starting line (Tier-1 check 5). Share of profit of JVs comes as one after-tax line, and dividend declared vs paid timing shows up in the retained-earnings roll (Tier-2 check 9).

You may pick different companies, within limits: non-financial (bank/NBFC/insurer formats wait for Phase 5), listed 7+ years, a consolidated group with subsidiaries and an NCI line, and an unmodified audit opinion, because your first spread should fight mechanics and leave forensics for later. Approved alternative pairs: Microsoft + Infosys (bonus: skim Infosys's 20-F beside its Indian AR for the Ind AS and IFRS bridge, free) or PepsiCo + Dabur India (bonus: Dabur's promoter-pledge history makes the M8 row non-trivial). Log your choice and why.

Part C: first flight of the checklist (1 hour, strictly timed)

Run the checklist v1 from earlier on a third company you've never spread, say ITC (bseindia.com) or Procter & Gamble (EDGAR), using the latest filing + proxy/shareholding pattern only. Set a 60-minute timer and produce the 10-line output. Note where you ran over. It's the first rep of the weekly habit Phase 8 will harden into the final artifact.

Deliverables (file all of it in your knowledge system)

1. Two workbooks in the standard template's exact row structure: 5 years × (IS + BS + CFS + notes panel + derived panel), M-panel filled, Source column on every row.
2. Two MappingLogs: every judgment call (COGS method, other-income split, reclassifications, restatements found) with row/filing/page/decision/reason recorded. A first spread with fewer than ~10 log entries per company almost certainly netted or dumped something.
3. Two completed self-check score sheets (the self-check protocol): every Tier-1/2/3 check, per year, with pass/fail/reconciled and the named reconciling items.
4. Two one-pagers in the step-12 format from the process above: the 10-line summary + five disclosure habits + three open questions.
5. The Part C 10-line output, with your actual elapsed time.
6. The aggregator diff (30 min): open screener.in (India) / stockanalysis.com (US) for your two companies and explain every material difference between their standardized numbers and your spread as their mapping choice or yours.

Grading: the self-check protocol is the answer key

No external answer sheet exists for a real filing. The self-check protocol is the grading key. Tier-1 checks catch transcription errors deterministically, since a balance sheet that balances in all ten company-years is no accident. Tier-2/3 checks catch mapping errors and force the reconciling items into daylight. Grade yourself against it, then score the rubric:

Pass = ≥16/20 AND the Tier-1 row at 2. No exceptions on Tier 1, because "the spread ties" is the phase's definition of done. If you fail, the protocol has already localized every error, since that's what it was designed to do. Fix, re-run the checks, re-score. Passing this project plus the mastery check below closes Phase 1.

Triage the three classic failures like this. (1) Balance check off by a weird amount: hunt mixed units first (₹ lakh typed as crore, or a "million" table in a crore workbook), then a subtotal typed as a component. (2) Cash articulation fails on the Indian company: look for other bank balances / unclaimed-dividend accounts outside cash equivalents (P9's pattern). (3) Equity roll residual large in one year: you missed buybacks or an NCI transaction. Open the SOCE and read it line by line. You'll find the answer printed there.

Case Lab: run case XL-1.2 (The Filing That Almost Tied) in the app's Cases tab for this capstone: a synthetic 10-K whose hand-spread fails three tie-outs until you locate the held-for-sale cash, the current maturities of long-term debt buried in "other current liabilities," and the impairment bundled inside the depreciation line.

Reading & resources

• SEC Investor Bulletin: "How to Read a 10-K/10-Q" (investor.gov → search the title). [Free] [Beginner]: the regulator's own two-page map of the document; read once before Part A.
• SEC EDGAR (sec.gov/edgar). [Free] [Beginner]: company search and full-text search, your permanent US primary source. Learn the full-text search operators once; they pay forever.
• Zerodha Varsity, Module 3: Fundamental Analysis, the annual-report and financial-statement chapters. [Free] [Beginner]: the friendliest India-first walkthrough of an AR's anatomy; pairs exactly with the Indian-AR anatomy section above.
• Thornton O'Glove, Quality of Earnings. [Paid] [Intermediate]: the 1987 classic on reading reports adversarially (differential disclosure, the auditor's report, inventories/receivables). Dated examples, timeless method; chapters 1–5 now, the rest in Phase 2.
• Schilit, Perler & Engelhart, Financial Shenanigans (4th ed.). [Paid] [Intermediate]: read chapters 1–2 now as a preview; the full taxonomy is your Phase 2 earnings-games spine.
• EY, US GAAP vs IFRS: The Basics. [Free] [Intermediate]: keep beside you when a Colgate policy and an Asian Paints policy for the same item read differently.
• Deloitte IAS Plus (iasplus.com). [Free] [Intermediate]: one-page standard summaries; use it to decode any policy-note reference (Ind AS numbering mirrors IFRS).
• ICAI, Guidance Note on CARO 2020 (icai.org → publications). [Free] [Advanced]: the clause-by-clause authority behind the auditor-interrogation section above; skim the clauses on loans, defaults, and fraud to see what auditors must actually test.
• SEBI LODR Regulations, updated master text (sebi.gov.in → Legal → Regulations). [Free] [Advanced]: Reg 23, 30, 31, 33 and Schedule V in the original; verify the RPT and pledge thresholds here whenever you quote them (they get amended).
• Schedule III, Companies Act 2013 (Division II) + the 2021 amendment notification (mca.gov.in), with ICAI's Guidance Note on Division II. [Free] [Advanced]: the presentation law behind every Indian spread, including the ageing schedules and the eleven-ratio disclosure.
• NFRA orders (nfra.gov.in → Orders). [Free] [Advanced]: India's audit regulator publishes its enforcement orders in full, named engagements, what the auditor missed, which clause failed. Free forensic case files; read one per month from Phase 2 onward.
• PCAOB AS 3101 (pcaobus.org) and the Center for Audit Quality's CAM resources (thecaq.org). [Free] [Advanced]: what a CAM legally is and how audit firms write them.
• Damodaran's accounting-for-analysts sessions (pages.stern.nyu.edu/~adamodar → teaching: free recorded classes). [Free] [Beginner→Intermediate]: the analyst's-eye recap of why each statement exists; good revision before the mastery check.
• Specimen filings to keep (all [Free]): Colgate-Palmolive 10-K + DEF 14A (EDGAR) and Asian Paints AR + shareholding pattern (BSE): your capstone set; Infosys AR + 20-F: the Ind AS↔IFRS bridge pair [Intermediate]; Vodafone Idea AR: a live "Material Uncertainty Related to Going Concern" section in the wild [Intermediate]; ZEEL quarterly shareholding patterns 2018–2020 (BSE): the pledge spiral in primary documents [Intermediate].
• screener.in and stockanalysis.com. [Free] [Beginner]: after the capstone, for the aggregator-diff exercise; from Phase 2 they become legitimate speed tools, because you can now audit them.
• Going deeper: Fridson & Alvarez, Financial Statement Analysis: A Practitioner's Guide (Wiley). [Paid] [Intermediate]: the chapters on the auditor's report, the notes, and analytical adjustments treat this capstone's narrative layer with a rigour Piper only sketches; the natural bridge into Phase 2's quality-of-earnings work.
• Going deeper: SEC Division of Corporation Finance, Financial Reporting Manual (sec.gov → Divisions → Corporation Finance). [Free] [Advanced]: the staff's own reference on what US filers must disclose and how; when a 10-K or DEF 14A disclosure question has no obvious answer, this is where the answer actually lives. A reference work, not a cover-to-cover read.

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": "Triaging a full filing — policies, critical estimates, MD&A, the auditor's report, CARO, the proxy and the shareholding pattern — and hand-spreading five years of statements onto one template",
  "use": "A 10-K runs 100 to 300 pages and a modern Indian annual report 300 to 500, so §4.1's triage — read thirty pages deeply, skim fifty, ignore the rest knowingly — is the module's real skill. An assistant changes what 'skim fifty' can mean, and nothing else. (1) Ask it to locate rather than to summarise: give me the page or note reference for the eight policies of §4.4, the critical-estimates list, every KAM or CAM paragraph, and each CARO clause that is not a clean 'yes'. A locator you verify is useful; a summary you trust has replaced the filing with a paraphrase. (2) Ask for a year-over-year comparison of the *policy* text and the risk-factor text between two filings and have it list what changed in wording, then read those paragraphs yourself — the sentence a company quietly deleted is one of the highest-signal objects in a filing. (3) Paste a non-GAAP reconciliation and ask it to assign each add-back to §4.11's four buckets with a reason, then overrule it, because the A-versus-B call depends on how many years the 'one-timer' has appeared and only you are holding five years. (4) Ask it what a filing does NOT contain that a peer's does — a missing ageing schedule, an absent segment disclosure, a ratio explanation that answers a different question.",
  "tools": ["Document/PDF AI — NotebookLM or Claude file upload for locating policies, KAM/CAM and CARO clauses across a long filing", "Chat assistants — Claude, ChatGPT for year-over-year policy and risk-factor text comparison and for a first-pass add-back classification", "Coding assistants — for the XBRL pull, the tie-out assertions and the pledge and wedge arithmetic below"],
  "prompt": "From the attached filing, produce a locator table only: [Item | What it is | Page or note reference | Verbatim quote of at most 25 words]. Cover the eight policies in my list, every critical estimate, the audit opinion type, every KAM or CAM heading, every CARO clause with an adverse or qualified answer, and the promoter shareholding and pledge figures. Do not summarise, interpret, or compute anything. If an item is absent from this filing, write ABSENT — do not infer it from a prior year or from a peer.",
  "verify": "Nothing from a tool enters the spread. The spread is built from the filing's own pages, and the tool's output is a table of contents you check page by page — a locator that points at the wrong note costs you thirty seconds, whereas a summarised figure that entered the template costs you the whole analysis. Then run §4.13's self-check protocol exactly as written: every balance sheet balances, cash rolls, equity rolls, EPS recomputes, segments cross-foot to the face. Those assertions are the only thing that certifies a spread, and they are indifferent to how the numbers arrived. Note also that an assistant reading a filing does not know what is absent from it unless you ask, and absence is where this module's best findings live (Guardrail 1-5).",
  "diy": "The gate is unaided, and this is the phase capstone: you hand-spread a full 10-K and a full Indian annual report, build COGS and EBIT from a by-nature Schedule III P&L, rebuild a non-GAAP bridge, and pass the tie-out protocol on paper. Nothing in this callout is admissible in that room, and the whole point of Working Agreement 2 is that the aggregator comes after the hand-spread, never instead of it."
}

Modern Data Analysis

By hand first. You spread two full filings by hand, built the mapping log, and passed the self-check protocol. That sequence isn't negotiable, because it's the only way you learn what an aggregator's "standardised" row has silently done to a by-nature Indian P&L. The code below is what you're allowed to do afterwards, for the third company and the thirtieth.

Today's workflow. A modern spread is a two-source job. For a US filer, the XBRL facts behind the statements are machine-readable at the company level, so five years of the income statement, balance sheet and cash-flow statement arrive as a tidy frame from one request. Every Phase 3 model's historical block is then a function, and you skip an afternoon of typing. For an Indian filer far less is tagged, so the reality is a hybrid. You use the exchange filing and the annual report PDF for the statements and notes, an aggregator for the convenience series, and your mapping log to reconcile the two. That's why you built the log by hand first. The part worth automating hardest is the tie-out, even more than the extraction. The self-check protocol is five assertions, and written as code they run on every company, every year, forever, instead of once when you remember. Assets equal liabilities plus equity. Opening cash plus CFO plus CFI plus CFF equals closing cash. Opening equity plus profit plus OCI minus dividends plus issuance equals closing equity. Net profit divided by the diluted weighted count reproduces reported diluted EPS. Segment revenue and segment profit cross-foot to the consolidated face.

Tools & sources (IN + US). pandas for the spread and the panel, requests for the filing pulls, sec-edgar-downloader for bulk retrieval, pdfplumber or camelot for Indian note tables, difflib for the year-over-year policy-text comparison. India: the BSE and NSE filing portals for annual reports, quarterly LODR results and the Reg 31 shareholding-pattern and pledge tables, MCA21 for the unlisted group entities the related-party note points at, and screener.in or stockanalysis-style aggregators for the convenience series you then reconcile. US: EDGAR full-text search, the companyfacts and submissions JSON endpoints, the SEC's quarterly Financial Statement Data Sets for bulk work, and the DEF 14A for the pay and ownership layer covered earlier.

# Synthetic, this module's own Worked Example figures — R3: recomputed in-session
S_out, P0, promoter, pledged = 96.0, 400.0, 0.42, 0.90      # crore shares, ₹
S_t = S_out * promoter; S_p = S_t * pledged                  # 40.32 and 36.29 cr
debt = S_p * P0 / 2.0                                        # 2.0x cover -> ₹7,258 cr
trigger = 1.75 * debt / S_p                                  # ₹350.00 -> a 12.5% fall
floor   = 1.75 * debt / S_t                                  # ₹315.00 if EVERYTHING is pledged
print(round(trigger, 2), round(floor, 2), round(1 - floor / trigger, 4))   # 350.0 315.0 0.1
gaap, adj, sbc = [-180, -120, -40], [215, 330, 450], [260, 300, 340]
wedge = sum(adj) - sum(gaap)                                 # 1,335 on cum GAAP of -340
print(wedge, round(sum(sbc) / wedge * 100, 1))               # 1335  67.4  (% of the wedge)
def tie_out(a, l, e, tol=0.5): assert abs(a - (l + e)) <= tol, (a, l, e)   # one of five

Verify. An aggregator's series is a hypothesis about your filing. Don't treat it as a copy. Reconcile at least revenue, EBIT, net profit, total debt and closing cash for the most recent year against the filing you spread by hand, and log every difference with its cause: standardisation of other income, a different treatment of exceptional items, consolidated versus standalone, a restated prior year. Where XBRL is the source, check that the tag you used means what you think. The same concept appears under several tags across filers and years, and a silent tag substitution is the modern equivalent of transcribing from the wrong column. Then let the five tie-out assertions decide. A spread that fails one of them isn't yet a spread at all.

Quantitative lens

The pledge spiral from the India-stack section looks like a narrative, but it's really a threshold system. That makes it the right thing to put into code, because thresholds are where intuition fails and arithmetic decides. Take Worked Example 4's own numbers: 40.32 crore promoter shares, 36.29 crore of them pledged, ₹7,258 crore of promoter debt at a 2.0× cover with a 1.75× maintenance trigger. The first top-up call doesn't wait for the 25% fall the example walks through. It arrives at a share price of ₹350.00, a fall of only 12.5%, and by ₹300 the cover is already 1.50× and two calls deep. Now ask the question the example leaves unsaid. How much cushion does the promoter's remaining unpledged stake actually buy? Pledging all 4.03 crore free shares moves the trigger price from ₹350.00 to ₹315.00, and 315/350 is exactly 0.90, which is 36.29/40.32.

That ratio is no coincidence. It's a general law worth carrying: when a fraction x of the promoter's holding is already pledged, pledging the entire rest lowers the trigger price by exactly (1 − x). A promoter 50% pledged can absorb another 50% fall before running out of collateral. One 90% pledged can absorb 10%. One 95% pledged can absorb 5% and is, for practical purposes, already out of ammunition. So the Street's rule of thumb screens on pledged percentage of the promoter's holding, and pledged percentage of total capital can't do the same job. That first number is the remaining cushion, expressed directly, and no other disclosure gives it to you.

The same threshold thinking exposes a limit in the other India-specific gift this capstone hands you. Schedule III requires an explanation for any of eleven ratios that moves more than 25%. That sounds like a filter on unusual events, and it's mostly a filter on ordinary noise. A ratio is a quotient of two series. If numerator and denominator each wobble by about 10% a year, roughly 7.8% of ratios breach a 25% band by chance alone, and at 15% wobble roughly 23.5% do. So both halves have limits. The pledge law assumes a single lender, one maintenance cover, no cash top-up and no other collateral, while a real promoter has several lenders with different covers and can post other assets. Read it as the shape of the risk. It won't predict a date. And a mandated ratio explanation is a prompt to look, never evidence that something happened. The signal is in which ratio moved, whether the explanation is specific enough to check, and whether it's the same explanation as last year.

Do it in code: build the trigger-price curve as a function of pledged fraction and maintenance cover for the promoter you're studying, mark today's price on it, and read off the distance to the first call. Then pull the quarterly shareholding-pattern history and plot pledged percentage against price. The spiral shows up as a rising line against a falling one long before it shows up in the results. See G1/QM1.01 and G2/DA1.04.

Where this goes next: galaxy cross-links

• G2 · CS2.05 APIs, HTTP & JSON: how the EDGAR companyfacts and exchange endpoints above actually work, so a five-year spread is a request you understand instead of a script you copied.
• G2 · DA1.02 Pandas Data Wrangling: the standard spreading template as a tidy frame, and the mapping log kept as data.
• G2 · DA1.04 Exploratory Data Analysis & Visualization: the pledge-versus-price plot and the non-GAAP wedge over five years, where a spiral and a widening wedge are both shapes before they are numbers.
• G1 · QM1.01 Probability, Distributions & Estimation: why a 25% band on a ratio of two noisy series flags what it flags, and how to set a threshold that catches events and lets ordinary variance pass.
• G3 · CN4.01 Reading an Industry: the filing as one witness among many. This capstone reads the company's own account of itself, and the industry module reads the industry around it.

Teach it back & journal

Feynman prompt. Write a one-page explainer for a smart 15-year-old titled "The exam where the student wrote the marking scheme." It must answer, in plain words with one running example (invent a bicycle company): (1) why the pages of an annual report without numbers decide whether you can trust the pages with numbers, (2) what the auditor actually promises, and the two things people wrongly assume the auditor promises, (3) why you copied every number into your own grid by hand instead of trusting a website that already did it. Use the words opinion, estimate, and tie out, and no other jargon. Here's the test. If the 15-year-old can't explain back to you why "the balance sheet balanced" is evidence of anything, rewrite the last third.

Journal reflection. This is the Phase 1 closing entry. Give it 30 minutes, in four parts. (1) Then vs now: reopen your notes from Phase 0's first guided filing read. List three things you now see on the same pages that were invisible fourteen weeks ago, and one thing you then believed about accounting that was simply wrong. (2) The judgment log: from your two capstone spreads, name the three hardest mapping calls you made, and write each as a standing rule for your future self ("other income at cash-rich Indian companies: split at entry, quote both margins"). (3) Disclosure habits: one sentence each on what your US company's and your Indian company's filings revealed about the people who wrote them: tone, candor, what they led with, what they buried. (4) The commitment: name the company-of-the-week you'll run checklist v1 on next week, and the fixed weekly slot it gets. Phase 2 assumes this habit exists.

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