The Document Lab: DRHP, Quarterly Results, 8-K, 13F and a Rating Rationale

Phase 8, the Rapid-Analysis System · Week 66 · ~16 focused hours · Mastery gate ≥ 85%

A draft red herring prospectus for a mid-size Indian manufacturer runs to about 640 pages. On a first pass you will read thirty-four of them, which is 5.3125% of the document, and if you pick the right thirty-four you will know more about that company than most of the people who read all 640 back to front.

The ratio is the point. Every document a company is compelled to file has a small number of pages where the compulsion actually bites, and a very large number of pages that exist because a lawyer wanted the risk covered. An analyst who cannot tell the two apart reads everything, learns little and runs out of week. An analyst who can tell them apart reads eight documents a quarter, in about forty-five minutes each pass, and comes out with a list of things that changed.

You already read two of the eight. The annual report and the 10-K were the first guided walks you ever took, and the full-filing capstone taught you to hand-spread one and to police what management calls adjusted. The other six get treated, almost everywhere in the profession, as reference material rather than as documents with a grammar. They are the quarterly result, the earnings release with its deck and its transcript, the event filing, the ownership disclosures and the rating rationale, plus the prospectus a company writes once, when it is asking strangers for money for the first time.

Each of those six has a fixed skeleton. The skeleton is fixed because a regulator wrote it down. The same three pages are therefore load-bearing in every filing of that type, forever, and the skill of finding them transfers from company to company and from country to country. That is a much better trade than it looks: nine document types, learned once, cover most of what a working analyst reads for the rest of a career.

The six also share a property that makes them dangerous. Unlike an audited annual report, most of them are unaudited, self-selected or forward-looking. A quarterly result carries a limited review, which is a weaker instrument than an audit and says so in its own second paragraph. An investor presentation carries whatever growth rate management prefers, and there are usually three defensible ones. A 13F shows a manager's long US equity positions as they stood at a quarter end that is already forty-five days behind on the day the filing is due, and further behind every day after that. A rating rationale is an opinion with an explicit expiry mechanism written into it. Reading these documents well means holding two facts at once: what the document says, and what the document is permitted not to say.

Two questions run through everything below. Which fact in this document would change a number in my model? And what did the filer choose not to show me, given that they chose the frame? By the end you will have read a prospectus pricing page against the issuer's own restated accounts and reconciled four quarterly results to an audited annual report. You will have mapped an event filing to its item numbers and stripped a portfolio disclosure down to what it cannot say. And you will have converted a rating agency's sensitivities into five things you check every quarter, then classified a post-year-end covenant breach in two accounting regimes that give opposite answers on identical facts.

Learning objectives

By the end you can:

1. Name the eight documents a listed company puts in front of you in a normal quarter, state what each is for, and say which of them is audited, which is reviewed and which is unaudited advocacy.
2. Walk a draft red herring prospectus by section, going straight to Objects of the Issue, Basis for the Offer Price, the Restated Consolidated Financial Information, the Capitalisation Statement, Outstanding Litigation, Related Party Transactions and the Statement of Special Tax Benefits, and say what each one is there to stop.
3. Read a Basis for the Offer Price page adversarially: recompute the issuer's own key performance indicator from the restated accounts, test the definition against a plain one, rebuild the peer set, and quantify what each choice is worth in enterprise value.
4. Reconcile an Indian quarterly result to the audited annual report, deriving the fourth quarter as a balancing figure, separating the standalone and consolidated columns, tying the segment table to profit before tax, and reading the limited-review paragraph for what it does and does not assert.
5. Bridge an earnings release, an investor presentation and a call transcript when the three carry different growth rates, and state which of the three numbers you would put in a model and why.
6. Log a management claim so it can be scored, then score a quarter's worth of claims and turn the hit rate into a number you can quote.
7. Map an 8-K to its item numbers, distinguish filed from furnished items, run an EDGAR full-text hunt for a phrase across filers, and record an accession number so the fact is retrievable.
8. Read a Form 13F and state the four things it cannot tell you, then compute what share of a manager's book it covers and what a listed option position does to the exposure it implies.
9. Classify a Schedule 13D against a 13G from Item 4's stated purpose, the filer's category and the timing, and say what an amendment signals.
10. Convert a credit rating rationale into a five-row monitorable trigger list, computing the headroom to each named sensitivity and naming the quarterly document that refreshes it, and classify a subsequent event as adjusting or non-adjusting under both Ind AS 10 and ASC 855 with the accounting consequence quantified.

Prerequisites & connections

Builds on. M0.05 gave you your first guided walk through an Indian annual report and a US 10-K, and taught the proxy statement as the place the people and the pay live. M1.10 hardened that into a hand-spread of a full filing, with the auditor's report, CARO, the Ind AS and Schedule III presentation rules, the non-GAAP reconciliation discipline and the promoter-pledge mechanics. M8.01 gave you the five questions and the four-layer source stack, plus the exact read order for a filing and the ten Ctrl-F tripwires. M8.02 turned all of it into the twelve-question teardown and the eight-line output. From elsewhere in the spine: M1.04 for the cash flow statement, M2.06 for quality of earnings, M3.03 for the cost of capital and the national-scale rating trap, and M5.02 for how an Indian non-bank finance company funds itself and why its rating rationale reads differently from a manufacturer's.

Feeds forward. M8.03 and M8.04 spend two to three days on a single company, and every document type here is one of the inputs those days consume. M8.05 certifies the whole system under a stopwatch, and the timed sweep below is the drill that makes the document half of that certification survivable. E11.01 owns credit analysis proper and takes the rating rationale from a document you can read to a scorecard you can rebuild. E11.04 takes the shareholding pattern and the pledge line into full India governance forensics. E11.02 takes the event filing into merger and special-situation work, where an 8-K item number is often the first public trace of a deal. M9.01's position sizing consumes the trigger lists built here as monitoring inputs.

Six neighbours own material that touches this one, and each of them keeps it. M0.05 owns the first guided annual-report and 10-K walk, including the proxy statement, and nothing here re-teaches those documents from scratch. M1.10 owns the full-filing capstone, the auditor's report and CARO, the accounting-policy read, and the non-GAAP policing discipline; the prospectus work below applies that discipline to a new document rather than restating it. M8.01 owns the source stack, the five questions, the filing read order and the tripwire list; the sweep below is a different object, being a fixed-time pass across nine document types rather than a read order inside one filing. M8.02 owns the teardown, its twelve questions and its eight-line output. E11.04 owns the India governance forensics: the promoter-group map, the related-party web, Regulation 23 materiality, the subsidiary loan hopscotch and the full pledge cascade with its margin-call arithmetic; what follows reads the quarterly disclosure that feeds that lab and hands the cascade over. E11.01 owns rating methodology, the agency scorecard, covenants and recovery analysis; what follows reads a published rationale as a document and converts its sensitivities into a monitoring list, and stops there.

Core teaching content

8.6.1 The eight documents of a quarter, and where the other six hide

Ask a working analyst what they read about a company in a normal three months and the honest answer is a list of eight things, not two. The annual report or the 10-K is the foundation and it arrives once a year. Everything else arrives on a rhythm, and the rhythm is set by a regulator rather than by the company.

A ninth document sits outside the quarterly rhythm and is read once, at the moment a company first asks the public for money. In India it is the draft red herring prospectus, then the red herring prospectus, then the prospectus. In the US it is the S-1, amended until effective, then the 424B. A prospectus is the single richest document a company ever publishes, because it is written under a liability standard that punishes omission and it has no prior year to hide behind.

Two properties separate these documents from the annual report, and both of them change how you read.

The first is assurance. An annual report carries an audit opinion, which is a positive assertion by a firm that has staked its licence on it. A quarterly result carries a limited review, which asserts only that nothing came to the reviewer's attention. An earnings release, a deck, a transcript and an event filing carry nothing at all. Reading a deck with the credulity appropriate to an audited balance sheet is the single commonest failure in this whole area, and it costs people money every quarter.

The second is selection. An annual report has a mandated table of contents. A deck has none, so what is in it is a choice, and so is what is missing. When a metric that led the deck for six quarters vanishes in the seventh, the vanishing is the disclosure. You will only see it if you kept the sixth deck.

8.6.2 The draft red herring prospectus: the section map

A prospectus is red-herring because the front page carries a warning in red saying the price is not yet fixed. The draft version goes to the regulator and to the public for comment; the regulator issues observations; the issuer files the red herring prospectus with a price band, runs the book, and files the final prospectus with the price. Nothing about the sequence matters as much as this: the draft is public, free and usually available for months before the issue, and it is the most complete description of the company that will ever exist.

Read it in this order, and stop when you have what you came for.

Objects of the Issue is the first stop because it separates two very different transactions wearing one name. A fresh issue creates new shares and the money goes to the company. An offer for sale transfers existing shares and the money goes to the selling shareholders. A company raising a headline Rs 1,250 crore of which Rs 800 crore is fresh receives 64.0% of the gross proceeds; the other 36.0% never touches its balance sheet. The prospectus says this plainly, and the newspaper coverage almost never does.

Inside the fresh issue, each object carries an amount, a schedule and a source of the estimate. The Indian rules cap what can be labelled general corporate purposes at a percentage, and the base that percentage applies to is defined in the regulation's own words. Compute the ratio on both bases every time, because a proportion of the fresh issue and a proportion of gross proceeds are different numbers whenever there is an offer for sale, and only one of them answers the compliance question. In the Rs 800 crore fresh issue used throughout the worked examples below, Rs 90 crore of general corporate purposes is 11.25% of the fresh issue and 7.2% of gross proceeds, so it clears a 25% ceiling on either basis. Compute both anyway, because a company that has quietly filled the general corporate bucket has told you it does not have a use for the money, which is a fact about the business rather than about the issue. (The cap, its base and the related limit on unidentified acquisition objects sit in the issue-capital regulations and have been amended; read the current regulation text and the issuer's own compliance statement in the Objects section before ruling on any issue. Checked September 2026.)

The monitoring agency matters more than its one-line mention suggests. Indian rules require an issuer above a stated issue size to appoint a credit rating agency to monitor the use of proceeds and to report on it, and the reports themselves become quarterly public documents afterwards. That converts the Objects table from a promise into a scorecard you can mark later, which is exactly what the deviation statement covered further down does for you. (Threshold and reporting frequency set by the issue-capital regulations and amended more than once; check the current text before quoting a number. Checked September 2026.)

Restated Consolidated Financial Information is the prospectus's most under-read section and its most valuable. Ordinary annual accounts are prepared under whatever policies applied that year. Restated accounts take three years and force them onto one set of policies, with an explicit reconciliation showing every adjustment made to the originally reported numbers. That reconciliation is a free confession: it lists, line by line, the places where the earlier accounts were wrong, or were prepared on a basis the company no longer wants to defend. Read the restatement adjustments before you read the restated statements.

The Capitalisation Statement answers the question the Objects table only implies. It shows borrowings and equity as at the latest balance sheet date and as adjusted for the issue. If the adjusted column shows total debt falling by less than the debt-repayment object, either the schedule is slow or the company expects to borrow again in between. Both are answerable; neither is visible anywhere else.

The last three stops are fast and disproportionately useful. Outstanding Litigation carries amounts, which lets you compare disputed tax demands against net worth in ten seconds. Related Party Transactions shows the promoter web at three-year depth, before the pre-listing tidy-up, which is the only historical view of it you will ever get. And the Statement of Special Tax Benefits is a chartered accountant's opinion on positions the company is relying on, written in the careful voice of somebody who expects to be quoted back at himself.

8.6.3 Reading a Basis for the Offer Price page adversarially

The pricing section is where an issuer argues that the number in the price band is defensible. Indian rules have made that argument unusually rich since the 2022 amendments to the issue-capital regulations. An issuer must now disclose the key performance indicators it showed any pre-issue investor over the preceding three years, have those figures certified, explain each one, and compare itself against listed peers on the same basis. (Requirement introduced by amendment to the issue-capital regulations in 2022; check the current text and the certification requirement before relying on the detail. Checked September 2026.)

The effect is that the issuer must show its own scoreboard, in its own words, with its own definitions. That is a gift and a trap in the same table.

Four questions turn it into information.

Start by asking whose indicators these are. The regulation forces disclosure of metrics shown to investors, not metrics the company runs on internally. A metric that appears in the prospectus and nowhere in the annual accounts is a metric built for selling. That does not make it wrong. It makes it a claim you should be able to reproduce.

Then ask how each one is defined. The definitions sit in a footnote table beside the metrics and they are where the argument lives. An issuer that reports Adjusted EBITDA and defines it as profit before tax plus finance costs plus depreciation plus share-based payment expense plus one-time offer expenses plus a provision for doubtful receivables has told you exactly which costs it wants excluded. Every one of those add-backs may be defensible. The stack of them is a different claim from any single one, and the stack is what you price.

Recompute it. In the restated FY26 accounts of the synthetic issuer used below, revenue from operations is Rs 2,840.00 crore and the plain operating measure, revenue less materials, employee cost and other expenses, is Rs 332.00 crore, an 11.6901% margin. The issuer's Adjusted EBITDA is Rs 501.00 crore, a 17.6408% margin. The gap of Rs 169.00 crore is 5.9507 percentage points of margin and it is 50.90% of the plain number. Check the gap by an independent route: share-based payment Rs 74.00 crore plus offer expenses Rs 31.00 crore plus the receivables provision Rs 18.00 crore plus other income Rs 46.00 crore equals Rs 169.00 crore, which is the same figure arrived at without going through profit before tax at all.

The third question is which peers the comparison uses. That table names listed companies and their multiples, and the choice of names is the most consequential paragraph in the section. An issuer selects peers whose multiples support the band. You select peers who compete for the same customer with the same economics. The two lists overlap less often than you would expect, and when they do not overlap the difference is not rhetorical. Quantify it.

The last question is what the issuer chose not to show, because absence is disclosure. A prospectus that gives you three years of a growth metric and no cohort retention, or gives you contribution margin and no customer acquisition cost, has made a decision. Write down what a reasonable seller would have included and did not, then look for the same fact somewhere else in the document. It is frequently in the Restated Financial Information, in a note nobody expects you to open.

8.6.6 The release, the deck and the transcript: three true numbers

Every quarter, one company publishes a press release, an investor presentation and a call transcript within about ninety minutes of each other, and the three documents disagree. Usually none of them is lying.

Work the arithmetic on a synthetic US filer. Reported revenue for the quarter is $2,318.0m against $2,030.0m a year earlier, which is growth of 14.1872%, and that is the number in the release, because the release reports what the accounts report. The deck leads with 18.5930%, printed as 18.6%, which is 4.4058 percentage points higher and 1.3105 times the release figure. Both are computed correctly.

The bridge has two steps. Foreign exchange moved against the company by $42.0m in the quarter, so at constant currency revenue is $2,360.0m and growth is 16.2562%, a step of 2.0690 points. Then the prior-year base included $40.0m from a unit sold during that year, so on a like-for-like basis the base is $1,990.0m and growth is 18.5930%, a further step of 2.3368 points. Add them: 14.1872 plus 2.0690 plus 2.3368 equals 18.5930, which is the deck headline to a rounding.

Three numbers, one quarter, all true. The question is which belongs in a model, and the answer depends on what the model is for. If you are forecasting reported revenue, which is what the accounts and the covenant and the market both see, the release number is the one that compounds. If you are asking whether the underlying business grew, the like-for-like constant-currency figure is closer to the truth. What you may not do is take the deck rate and apply it to reported revenue, which is the commonest modelling error in this whole area. Do it here and one year of revenue is overstated by $102.1258m, and five years of compounding overstates by $937.7356m on a base that started at $2,318.0m.

The transcript is what closes the loop, because the divestment and the currency effect are usually explained in the question and answer section rather than in the prepared remarks. The prepared remarks are scripted advocacy. The questions are asked by people who cover the company full time and who noticed the same gap you did.

Read the three in this order: results statement, then release, then deck, then transcript. Numbers first, framing second, defence last. Reading the deck first anchors you on the frame before you have the facts, and once anchored you will unconsciously look for confirmation in the filing rather than for contradiction.

8.6.7 The call as a document: the analyst list, the dodge and the claim log

A transcript is a document with a structure, and three parts of it repay attention out of proportion to their length.

Start with the analyst list: who asked a question, and from where. Coverage changes are informative. A name that has covered the company for four years and stops appearing has usually dropped it, and a sudden influx of new firms usually means an index inclusion, a large placement or a story that has started to sell. Count the questioners, not just the questions.

Then look for the questions that did not get answered. Managements rarely refuse outright. They redirect. The three standard moves are answering a different question, answering with a direction instead of a number ("we are seeing good traction"), and promising to follow up offline. Mark each one. A question that gets deflected two quarters running, by an analyst who keeps asking, is the highest-value item on the whole call, because a professional who covers the company has decided it is worth spending scarce question time on twice.

Last, read the guidance language, and learn its ladder. We expect is a forecast. We are targeting is an aspiration with no commitment. We are confident is a mood. We have line of sight to means it is contracted or nearly so. And we are not providing guidance at this time, on a company that guided last quarter, is a change in the disclosure regime. Treat it as news.

Then the habit that turns listening into evidence: the claim log. Every quarter, write down each dated, checkable statement management makes, in their words, with the date they attached to it. One line each. Next quarter, mark each one met, missed, quietly dropped or restated with a later date.

After four quarters you have a number. A synthetic example: fourteen claims logged, nine met on the stated timetable, three quietly disappeared and two were restated with a later date, which is a hit rate of 64.2857% and a slippage-or-silence rate of 35.7143%. That number is worth more than any adjective you could apply to management, and it is available to anybody willing to keep a text file. Almost nobody does.

8.6.8 The shareholding pattern and the encumbrance line

An Indian listed company files a shareholding pattern every quarter, and inside it sits a table that has no US equivalent at that frequency: promoter shares that are pledged or otherwise encumbered, as a number and as a percentage. The takeover code adds event-driven disclosure when a pledge is created, invoked or released.

Read four things together, and always as a trend rather than a level.

Take four consecutive quarters of a synthetic company. Promoter holding runs 52.40%, 52.40%, 51.10% and 48.90% of capital. The pledged proportion of that holding runs 18.00%, 26.50%, 33.20% and 41.00%. Multiply the pair to get encumbrance as a share of the whole company: 9.4320%, 13.8860%, 16.9652% and 20.0490%. The promoter has sold 3.50 points of the company and pledged 23.00 more points of what remains, and the encumbered block has crossed 20% of total capital, which is one of the two thresholds that trigger detailed disclosure of the reasons under the takeover code. (Encumbrance thresholds and the detailed-reasons requirement sit in the takeover regulations and have been amended; check the current text. Checked September 2026.)

Then size the block against the market that would have to absorb it. With 41.60 crore shares outstanding, the encumbered block is 8.3404 crore shares, worth Rs 2,235.2229 crore at Rs 268. Against thirty-day average traded value of Rs 18.4 crore a day, that block is 121.4795 days of the entire market's volume. A forced sale cannot happen at anything like the screen price, which is why an invocation is a price event rather than an ownership event.

The cascade from there, the loan-to-value, the maintenance cover, the trigger price and the reflexive spiral, is the India governance lab's material and it does the full treatment. The arithmetic is worth carrying in outline, because you will want it in the forty seconds you spend on this table during a sweep. At a 45% loan-to-value the implied promoter borrowing against that block is Rs 1,005.8503 crore; at a lender maintenance cover of 1.60 times the collateral value at which a margin call fires is Rs 1,609.3605 crore, which is a share price of Rs 192.96, or 28.00% below today's. A four-quarter chart of pledge percentage against share price tells you whether that distance is opening or closing.

The US analogue is annual and narrative rather than quarterly and numeric. Pledged shares appear in footnotes to the beneficial ownership table in the proxy statement, and many boards now carry an anti-pledging policy that the proxy also describes. Insider transactions arrive faster than in India, through Forms 3, 4 and 5, with Form 4 due within two business days of the trade. What the US does not give you is India's quarterly, percentage-of-holding fuse gauge.

8.6.10 Hunting on EDGAR: full text, and the accession number you record

Company-by-company browsing finds what you already know to look for. Full-text search finds what you did not know existed, and it is the single most under-used free tool in the profession.

The EDGAR full-text system indexes filing documents from 2001 onward and searches the body of filings rather than the metadata, which means you can search for a phrase and get back every filer who used it. Three shapes of query pay for themselves.

In a phrase hunt you search an exact contractual phrase across all filers to find every company using the same structure, the same lender, or the same auditor's language. Quoted phrases are the point.

In a counterparty hunt you search a private company's name to find it in other people's exhibits. A supplier that never files becomes visible through the material contracts of its customers, and a customer concentration disclosed vaguely by one filer is often named precisely by another.

In a narrow-by-form hunt you restrict to a form type and a date range, then read the hits in filing order. A search restricted to 8-K filings mentioning a specific covenant phrase over eighteen months produces a timeline of an industry's credit conditions in about twenty minutes.

Whatever you find, record the accession number. It is the eighteen-digit identifier stamped on every filing, written as ten digits, then a two-digit year, then a six-digit sequence, and it is the only durable way to name a specific document. A URL can change and a company can re-file; an accession number identifies exactly one submission, forever. Every fact in your notes that came from a filing should carry the filer, the form type, the filing date and the accession number, in that order. This is not administrative fussiness. A thesis you cannot re-trace in eighteen months is a thesis you cannot defend when it goes wrong.

8.6.16 The sweep: nine documents, forty-five minutes, and the disagreement rule

The sweep is a fixed-time pass across nine documents, run when you pick a company up again after a quarter away, or when you are preparing for a call, or as the document half of a certification rep. It produces one page per document, and its output is not a summary. It is a list of things that changed and what each one does to a number.

Nine documents, forty-five minutes, five minutes a document on average. The order is deliberate: audited before reviewed, reviewed before unaudited, company before outsiders, and the subsequent-events note last because it is the only one whose meaning depends on everything above it.

Then the rule that the whole exercise exists to install. When two documents disagree, the ranking is fixed and you do not negotiate with it.

1. The audited annual accounts, because somebody staked a licence on them.
2. The reviewed interim results, because somebody at least looked.
3. The regulatory filing that is filed rather than furnished, because it carries liability.
4. The furnished release or deck, because it is the company's own frame.
5. The transcript, because it is unscripted and therefore honest about tone but unreliable about numbers.
6. Everything else.

The ranking resolves a disagreement about fact. It does not resolve a disagreement about definition, and most disagreements you meet are about definition. Three growth rates that reconcile through a bridge are not a conflict; they are one number seen through three lenses, and the work is to build the bridge and then choose the lens that matches your question. A conflict that will not bridge, where two documents assert incompatible facts about the same period, is a different animal entirely. Write it down, take it to the call, and if it is not resolved within one quarter, treat the unresolved gap as a finding rather than as an inconvenience.

8.6.17 Where a machine helps here, and where it will cost you

Document work is the most obvious place in the whole programme to reach for a model, and it is also where the failure mode is quietest. An extraction that transposes one column of a segment table produces a tie-out that fails by an amount you can see. An extraction that transposes a column of a peer comparison table produces a number that is wrong and looks entirely reasonable, and nothing downstream will ever catch it.

{
  "skill": "Extracting a fixed-schema table from a long filing, then checking it",
  "use": "Have an assistant pull a defined set of fields out of a prospectus, a quarterly result or a rating rationale into a table you specify in advance, and have it flag every field it could not find rather than filling gaps. You then run the arithmetic tie-outs that prove the extraction is faithful.",
  "tools": ["Document or PDF AI grounded on the filing you already hold", "Chat assistants with code execution for the recomputation", "EDGAR full-text search (efts.sec.gov) and the exchange filing feeds for the primary documents", "A spreadsheet for the side-by-side against the source page"],
  "prompt": "Here is a draft red herring prospectus. Extract into one table: (a) every object of the fresh issue with its amount and deployment schedule; (b) every key performance indicator disclosed in Basis for the Offer Price, with the issuer's own definition verbatim and the page number; (c) the peer comparison table exactly as printed, with each peer's name and each multiple; (d) revenue from operations, cost of materials, employee benefits expense, other expenses, finance costs, depreciation and amortisation, profit before tax and tax expense from the restated consolidated statement for each of the three years. Mark any field you cannot locate as NOT FOUND and give the page you searched. Do not compute anything and do not fill a gap with an estimate.",
  "verify": "Recompute the issuer's headline adjusted metric from the extracted line items yourself and confirm it reproduces the printed figure to the rupee; then recompute it a second way, through the add-back stack, and confirm both routes agree. Check that the objects sum to the fresh issue and that the fresh issue plus the offer for sale equals the stated issue size. Open the peer table on the source page and read every multiple against the extracted row, because column transposition in a dense table is the failure this tool makes most often and it is invisible downstream. Trace every extracted figure to a page number and record the document, the filing date and the accession number or the exchange announcement identifier. Where a definition is quoted, confirm it is verbatim rather than paraphrased, because a paraphrased definition is the one thing you cannot audit later.",
  "diy": "You must be able to run the Basis for the Offer Price read by hand: recompute one issuer-defined metric from the restated accounts, build your own peer set, and quantify the enterprise-value effect of the peer choice and the definition choice separately. The gate rewards that, because the judgement about which peers actually compete is exactly the judgement no extraction tool can make, and the arithmetic is the only thing standing between a plausible extraction and a wrong one.",
  "market": "IN"
}

Two further limits are worth stating flatly. A model asked what a regulation currently requires will answer with the version it absorbed, which for anything amended in the last two years is frequently the superseded text; every threshold, deadline and cap in this area moves, and the primary text is the only acceptable source. And a model asked to summarise a call transcript will summarise the prepared remarks, because they are longer and better organised, when the entire value of the document is in the question and answer section.

8.6.18 Verify block: every worked figure at full precision

Every number printed in the worked examples, the practice solutions and the mastery keys is reproduced by the computation script and restated here at full precision, so that a reader who gets a different answer knows immediately whether the difference is rounding or error.

From the prospectus example: issue size Rs 1,250.00 crore, fresh Rs 800.00 crore (64.0% of gross), offer for sale Rs 450.00 crore (36.0%). Objects: Rs 300.00, Rs 260.00, Rs 150.00, Rs 90.00 crore, summing to Rs 800.00 crore; general corporate purposes 11.25% of the fresh issue and 7.2% of gross proceeds. Restated FY26: revenue from operations Rs 2,840.00 crore, other income Rs 46.00 crore, total income Rs 2,886.00 crore, total expenses Rs 2,738.00 crore, profit before tax Rs 148.00 crore, tax Rs 37.25 crore, profit after tax Rs 110.75 crore. Plain EBITDA Rs 332.00 crore (11.6901%); issuer Adjusted EBITDA Rs 501.00 crore (17.6408%); gap Rs 169.00 crore and 5.9507 points, verified independently as 74.00 plus 31.00 plus 18.00 plus 46.00. Post-issue shares 36.0000 crore, market capitalisation Rs 14,400.00 crore at Rs 400, net debt Rs 120.00 crore, enterprise value Rs 14,520.00 crore, which is 28.9820 times adjusted EBITDA and 43.7349 times plain EBITDA, and a price-to-earnings ratio of 130.0226 on restated earnings per share of Rs 3.0764. Issuer peer median 31.0000 times, learner peer median 19.5000 times. Enterprise value in the issuer's frame Rs 15,531.00 crore, in the learner's frame Rs 6,474.00 crore; peer-choice effect Rs 5,761.50 crore, definition effect Rs 3,295.50 crore, total Rs 9,057.00 crore, and 5,761.50 plus 3,295.50 equals 9,057.00. The two mixed frames are 19.5000 × 501.00 = Rs 9,769.50 crore and 31.0000 × 332.00 = Rs 10,292.00 crore. Implied price per share Rs 428.0833 and Rs 176.50; the cap price is 226.6289% of the learner's figure, a 55.8750% discount.

From the US event example: covenant EBITDA $412.60m, funded debt $1,285.00m, cash $214.00m capped at $150.00m, covenant net debt $1,135.00m, leverage 2.7508 times against 2.5957 times if all cash were netted, a difference of 0.1551 turns. At 3.50 times: debt ceiling $1,444.10m, headroom $309.10m, EBITDA floor $324.2857m, decline to breach $88.3143m or 21.4043%. At 3.00 times: ceiling $1,237.80m, headroom $102.80m, floor $378.3333m, decline $34.2667m or 8.3051%. Quarter end 30 June 2026, deadline 14 August 2026, 45 days; a 1 July sale is 44 days stale at the deadline and 69 days stale on 8 September 2026. Manager assets $6.20bn, US long equity $2.84bn, coverage 45.8065%, invisible 54.1935%, short book $0.86bn. Reported line $410.00m less put delta $340.00m equals $70.00m, or 17.0732%. Information table 1,240,000 shares at $47.85 equals $59,334,000.00, a 1.5816% stake on 78,400,000 shares. Schedule 13D 4,312,000 shares, 5.50%, 392,000.00 shares above the trigger, cost $190.1592m at $44.10, mark-to-market gain $16.17m or 8.5034%.

From the quarterly reconciliation: nine-month revenue Rs 2,323.60 crore, annual Rs 3,180.20 crore, fourth quarter Rs 856.60 crore. Nine-month profit Rs 197.40 crore, annual Rs 246.10 crore, fourth quarter Rs 48.70 crore. Margins 8.4954% and 5.6853%, a fall of 2.8102 points. Exceptional item Rs 22.50 crore pre-tax, Rs 16.8372 crore post-tax at 25.168%, adjusted fourth-quarter profit Rs 65.5372 crore and margin 7.6509%, explaining 69.9454% of the fall and leaving 0.8446 points unexplained. Standalone revenue Rs 2,610.40 crore, subsidiary revenue Rs 569.80 crore or 17.9171%; standalone profit Rs 274.60 crore against consolidated Rs 246.10 crore, a subsidiary loss of Rs 28.50 crore. Segment revenue Rs 1,742.30, Rs 1,027.90 and Rs 471.40 crore summing to Rs 3,241.60 crore, eliminations Rs 61.40 crore; segment results Rs 289.80, Rs 121.60 and Rs 74.40 crore summing to Rs 485.80 crore, less unallocable Rs 60.40 crore and finance costs Rs 96.50 crore equals Rs 328.90 crore, less tax Rs 82.80 crore equals Rs 246.10 crore, an effective rate of 25.1748%. Segment margins 16.6332%, 11.8299% and 15.7828%. Proceeds Rs 500.00 crore, utilised Rs 268.00 and Rs 96.00 crore, that is 89.3333% and 48.0000%, overall 72.8000%, unutilised Rs 136.00 crore, shortfall against a 75% schedule Rs 54.00 crore.

From the rating and subsequent-events examples: EBITDA Rs 918.00 crore, net debt Rs 1,652.00 crore, interest Rs 148.00 crore, liabilities Rs 2,410.00 crore, net worth Rs 2,180.00 crore. Ratios 1.7996, 6.2027 and 1.1055 times. Net-debt ceiling Rs 2,295.00 crore, headroom Rs 643.00 crore, EBITDA floor Rs 660.80 crore, fall to breach Rs 257.20 crore or 28.0174%. Coverage: interest ceiling Rs 306.00 crore, EBITDA floor Rs 444.00 crore, fall 51.6340%. Net worth gap Rs 230.00 crore, 0.5721 years at Rs 402.00 crore of retained profit. Liquidity sources Rs 1,578.00 crore against uses Rs 830.00 crore, cover 1.9012 times, surplus Rs 748.00 crore, and 1.3590 times excluding undrawn lines. Current assets Rs 1,480.00 crore, current liabilities Rs 890.00 crore, ratio 1.6629 times, working capital Rs 590.00 crore; after reclassifying Rs 620.00 crore, ratio 0.9801 times and working capital minus Rs 30.00 crore, a swing of 0.6828 turns. Acquisition Rs 1,150.00 crore for Rs 205.00 crore of EBITDA, 5.6098 times; pro forma net debt Rs 2,802.00 crore, pro forma EBITDA Rs 1,123.00 crore, leverage 2.4951 times, 0.0049 turns of cushion, and 0.6955 turns of deterioration. With Rs 38.00 crore of synergies, 2.4134 times. Breach on Rs 5.50 crore more debt or Rs 2.20 crore less EBITDA, that is 0.1959% of pro forma EBITDA or a 0.4783% cost overrun.

From the growth bridge and the pledge trend: reported $2,318.00m against $2,030.00m, 14.1872%; constant currency $2,360.00m, 16.2562%; like-for-like base $1,990.00m, 18.5930%. Steps 2.0690 and 2.3368 points, total gap 4.4058 points, ratio 1.3105 times. Extrapolation error $102.1258m in one year and $937.7356m over five. Claims 14, met 9, hit rate 64.2857%, slippage or silence 35.7143%. Promoter holding 52.40%, 52.40%, 51.10% and 48.90%; pledge 18.00%, 26.50%, 33.20% and 41.00% of that holding; encumbrance 9.4320%, 13.8860%, 16.9652% and 20.0490% of capital. Encumbered block 8.3404 crore shares on 41.60 crore outstanding, worth Rs 2,235.2229 crore at Rs 268, which is 121.4795 days of Rs 18.4 crore average daily traded value. Implied loan Rs 1,005.8503 crore at 45% loan-to-value; trigger collateral Rs 1,609.3605 crore at 1.60 times cover; trigger price Rs 192.96, that is 28.00% below.

Common mistakes & how experts think differently

1. Treating the deck's growth rate as the company's growth rate. The mistake is tempting because the deck number is the one on the slide with the big font, and because it is arithmetically correct. What it is not is the number that compounds in the accounts. An expert asks a single question of any growth rate before using it: what is the denominator? Then they build the bridge from the reported figure and choose a lens on purpose. Applying the deck rate to reported revenue overstates one year by $102.1258m and five years by $937.7356m on the example above.

2. Reading a limited review as though it were an audit. Three reviewed quarters that later disagree with an audited year are not evidence of deceit; they are evidence that a review is enquiry and analytics rather than testing and confirmation. The expert holds interim numbers more loosely by default, watches the fourth quarter as the true-up quarter, and treats a modified review report as a fire alarm arriving nine months early.

3. Quoting a 13F as a portfolio. The filing covers long positions in a defined universe of mostly US-listed equity, shows no shorts, arrives up to forty-five days after a quarter end, and can be inverted by an option position it does not have to reveal. In the example above it sees 45.8065% of a book and reports a $410.00m line whose economic exposure is $70.00m. The expert uses it to generate questions about who owns what, never as evidence of what anybody thinks.

4. Accepting the issuer's peer set. A prospectus peer table is an argument, not a fact, and it is the single highest-leverage paragraph in the pricing section. The expert rebuilds the set from who competes for the same customer, then computes what the substitution is worth. In the worked example the peer choice alone moves enterprise value by Rs 5,761.50 crore and the definition choice by a further Rs 3,295.50 crore, and separating the two is what makes the argument answerable rather than a matter of taste.

5. Adding back everything the issuer adds back, or nothing. Both extremes are lazy. Share-based payment is a real cost of the employees who produced this year's revenue; offer expenses genuinely will not recur; a receivables provision is an operating cost dressed as an event. The expert prices each add-back on its own merits, then states the stack as a total, because Rs 169.00 crore of add-backs on Rs 332.00 crore of plain earnings is a different claim from any single line in it.

6. Missing the fourth quarter because nobody filed it. Since Q4 is a balancing figure, an analyst who reads only what was filed never sees it. The expert subtracts nine months from the audited year as a reflex, computes the derived margin, and asks what landed there. Two lines of arithmetic, and it is where the year's honesty is settled.

7. Reading a rating instead of its sensitivities. The symbol is a compression of an opinion; the sensitivities are the falsifiable part, written as conditionals with numbers in them. The expert converts them into a table with a distance and a refreshing document on each row, computes each distance two ways, and re-runs it every quarter. A trigger without a named source document is a worry rather than a monitor.

8. Skipping the subsequent-events note because the numbers do not change. A non-adjusting event by definition leaves every ratio on the page untouched, which is precisely why it is the last place most people look and the first place a material change hides. In the example, a note describing an acquisition takes real leverage from 1.7996 to 2.4951 times, and the reported balance sheet says nothing at all.

9. Comparing current ratios across accounting regimes. A post-year-end covenant waiver produces a current ratio of 0.9801 under the international rules and 1.6629 under US practice on identical facts, which is 0.6828 of a turn of pure accounting. The expert checks which regime and which covenant test date applies before comparing anything, and treats a cross-border liquidity screen that ignores this as broken.

10. Not recording where a fact came from. A number in a model with no document, date and identifier behind it is a number you will not be able to defend in eighteen months, and eighteen months is roughly when you will need to. The expert writes filer, form, date and accession number beside every extracted figure, and refuses to promote anything into a model until it carries all four.

Worked examples

Each example states every input, verifies the answer by an independent route where one exists, and closes with what the number does not mean. All companies here are synthetic and labelled as such; the document structures, section names and item numbers are drawn from the public texts of the regulations and forms that create them. Every printed figure comes from the module's computation script.

Worked example 1: Reading a Basis for the Offer Price page adversarially (India, ₹)

Vaayu Mobility Limited (synthetic) has filed a draft red herring prospectus. The issue is Rs 1,250.00 crore, of which Rs 800.00 crore is a fresh issue and Rs 450.00 crore an offer for sale by two early investors. The price band is Rs 372.00 to Rs 400.00 per share. Pre-issue there are 34.00 crore shares. Objects of the fresh issue: Rs 300.00 crore to prepay borrowings, Rs 260.00 crore of capital expenditure on a new plant, Rs 150.00 crore of incremental working capital and Rs 90.00 crore of general corporate purposes.

The restated consolidated FY26 statement shows revenue from operations Rs 2,840.00 crore, other income Rs 46.00 crore, cost of materials consumed Rs 1,432.00 crore, employee benefits expense Rs 486.00 crore of which Rs 74.00 crore is share-based payment, other expenses Rs 590.00 crore, finance costs Rs 62.00 crore and depreciation and amortisation Rs 168.00 crore. Borrowings are Rs 640.00 crore and cash Rs 220.00 crore, and the statutory tax rate is 25.168%.

The Basis for the Offer Price section discloses "Adjusted EBITDA", defined as profit before tax plus finance costs plus depreciation and amortisation plus share-based payment expense plus one-time offer-related expenses of Rs 31.00 crore plus a provision for doubtful receivables of Rs 18.00 crore. It compares the company against three listed peers trading at 34.00, 31.00 and 29.50 times enterprise value to EBITDA.

Solution. Step 1, split the issue. The company receives the fresh issue only: 800.00 / 1,250.00 = 64.0% of gross proceeds. The selling shareholders receive 36.0%. Objects sum to 300.00 + 260.00 + 150.00 + 90.00 = Rs 800.00 crore, tying to the fresh issue. General corporate purposes is 90.00 / 800.00 = 11.25% of the fresh issue and 90.00 / 1,250.00 = 7.2% of gross proceeds, inside the cap on either basis.

Step 2, rebuild the accounts. Total income = 2,840.00 + 46.00 = Rs 2,886.00 crore. Total expenses = 1,432.00 + 486.00 + 590.00 + 62.00 + 168.00 = Rs 2,738.00 crore. Profit before tax = Rs 148.00 crore; tax at 25.168% is Rs 37.25 crore; profit after tax Rs 110.75 crore.

Step 3, reproduce the issuer's metric. Adjusted EBITDA = 148.00 + 62.00 + 168.00 + 74.00 + 31.00 + 18.00 = Rs 501.00 crore, a margin of 501.00 / 2,840.00 = 17.6408%.

Step 4, compute the plain version. Revenue from operations less materials, employee cost and other expenses = 2,840.00 − 2,508.00 = Rs 332.00 crore, a margin of 11.6901%. The gap is Rs 169.00 crore, which is 5.9507 percentage points of margin and 50.90% of the plain figure.

Step 5, verify the gap by an independent route. The add-backs, taken directly and never through profit before tax, are share-based payment 74.00 plus offer expenses 31.00 plus the receivables provision 18.00 plus other income 46.00 (which the plain measure excludes and the issuer's measure includes by construction) = Rs 169.00 crore ✓. Two routes, one number.

Step 6, price it both ways. New shares at the cap = 800.00 / 400.00 = 2.0000 crore; post-issue shares 36.0000 crore; market capitalisation Rs 14,400.00 crore. Post-issue net debt = (640.00 − 300.00) − 220.00 = Rs 120.00 crore; enterprise value Rs 14,520.00 crore. That is 28.9820 times the issuer's adjusted measure and 43.7349 times the plain one.

Step 7, rebuild the peer set and decompose. The issuer's median is 31.0000 times. Three companies selling the same product to the same customers trade at 17.00, 19.50 and 22.00 times, a median of 19.5000 times. Enterprise value in the issuer's frame is 31.0000 × 501.00 = Rs 15,531.00 crore; in the learner's frame 19.5000 × 332.00 = Rs 6,474.00 crore. Decompose the Rs 9,057.00 crore gap: the peer choice, holding EBITDA at the issuer's definition, is (31.0000 − 19.5000) × 501.00 = Rs 5,761.50 crore; the definition choice, holding the multiple at 19.5000, is 19.5000 × (501.00 − 332.00) = Rs 3,295.50 crore. Check: 5,761.50 + 3,295.50 = Rs 9,057.00 crore ✓.

Step 8, convert to a price. Issuer's frame: (15,531.00 − 120.00) / 36.0000 = Rs 428.0833, which is 7.0208% above the cap price. Learner's frame: (6,474.00 − 120.00) / 36.0000 = Rs 176.50, so the cap price is 226.6289% of it, a discount of 55.8750%.

Read it. The Rs 9,057.00 crore is not a valuation and it is not a claim that the issue is overpriced by 55.8750%. It is the price of two arguable choices, separated so each can be argued about on its own. A reader who accepts the issuer's EBITDA definition but not its peer set lands at 19.5000 × 501.00 = Rs 9,769.50 crore of enterprise value; one who accepts the peer set but not the definition lands at 31.0000 × 332.00 = Rs 10,292.00 crore. The Rs 6,474.00 crore figure belongs only to a reader who rejects both, applying the lower multiple to the lower base. What the decomposition rules out is the comfortable middle position of accepting a frame without knowing what it costs. Break it if the peers are genuinely not comparable, because a business with 40% revenue growth and one with 6% do not deserve the same multiple regardless of what they sell.

Worked example 2: An 8-K, a 13F and a 13D on one name (US, $)

Halden Industrial Corp (synthetic, NYSE: HLDN) files an 8-K carrying Items 1.01, 2.02, 5.02, 7.01 and 9.01. Item 1.01 describes a new $1,600.00m senior secured credit agreement with a maximum consolidated net leverage covenant of 3.50 times, stepping down to 3.00 times after four quarters, with cash netting capped at $150.00m. Item 2.02 furnishes the earnings release: LTM covenant EBITDA of $412.60m, total funded debt of $1,285.00m and balance-sheet cash of $214.00m. Item 5.02 reports the chief financial officer's resignation effective in six weeks, with no reason stated. Item 7.01 furnishes the investor deck. Item 9.01 lists the exhibits.

Separately, Larkspur Capital Management LLC (synthetic) reports 1,240,000 HLDN shares on its Form 13F information table for the quarter ended 30 June 2026, at a quarter-end price of $47.85. Larkspur manages $6.20bn: $2.84bn of US-listed long equity, $1.55bn of non-US listed equity, $0.62bn of private and illiquid positions and $1.19bn of cash and Treasury bills, plus a short book with $0.86bn of notional. It also holds listed puts on HLDN with a delta-equivalent of minus $340.00m against a $410.00m long line in a related name. HLDN has 78,400,000 shares outstanding.

Meridian Value Partners LP (synthetic) files a Schedule 13D on 4,312,000 HLDN shares acquired at an average $44.10, whose Item 4 states that the shares were acquired in the belief they are undervalued, that the filer intends to engage with the board regarding capital allocation and the composition of the board, and that it may seek representation.

Solution. Step 1, triage by item number. Item 1.01 changes a model input, because a covenant is a hard constraint on future capital allocation. Item 2.02 is the release, furnished not filed. Item 5.02 is a governance signal whose weight depends on what the next 8-K says. Items 7.01 and 9.01 are packaging. Read 1.01 first.

Step 2, compute the covenant position. Cash netting is capped, so covenant net debt = 1,285.00 − 150.00 = $1,135.00m, and leverage = 1,135.00 / 412.60 = 2.7508 times. Netting all $214.00m would give 2.5957 times, so the cap costs 0.1551 turns, and quoting the uncapped number is the single commonest error in reading a credit agreement.

Step 3, size the headroom at both levels. At 3.50 times the debt ceiling is 3.50 × 412.60 = $1,444.10m, so headroom is $309.10m; the EBITDA floor is 1,135.00 / 3.50 = $324.2857m, a decline of $88.3143m or 21.4043%. At the stepped-down 3.00 times the ceiling is $1,237.80m, headroom $102.80m, the floor $378.3333m, and the tolerable decline $34.2667m or 8.3051%. Four quarters from now this company can absorb an eight percent earnings decline, not a twenty-one percent one, and nothing on the face of the release says so.

Step 4, read the 13F for what it is. The information table line is 1,240,000 × $47.85 = $59,334,000.00, a 1.5816% stake. The four limits: the components sum to $6.20bn and the filing covers 2.84 / 6.20 = 45.8065% of the book, leaving 54.1935% invisible; the $0.86bn short book does not appear; the quarter ended 30 June 2026 and the deadline was 14 August 2026, 45 days later, so a 1 July sale is 44 days stale at the deadline and 69 days stale to a reader on 8 September 2026; and where puts are held, a $410.00m reported line with minus $340.00m of delta is an economic $70.00m, or 17.0732% of what the line implies.

Step 5, classify the 13D. The stake is 4,312,000 / 78,400,000 = 5.50%, which is 392,000.00 shares above the trigger, at a cost of $190.1592m and a mark-to-market gain of $16.17m or 8.5034% at the quarter-end price. The Item 4 language runs past "investment purposes" to naming capital allocation and board composition, and to reserving the right to seek representation. That is an activist filing on the intent ladder, and the correct classification for a monitoring list is engaged-with-stated-agenda rather than passive.

Read it. The covenant arithmetic is the only part of this that changes a model, and it changes it in a specific way: it caps leverage rather than forecasting it, so it constrains buybacks, acquisitions and dividends four quarters out. The 13F changes nothing at all and merely tells you who to expect on the shareholder register. The 13D changes the probability distribution of governance outcomes without changing any line of the accounts. What none of these documents does is tell you whether the chief financial officer left for a better job or because of something in the numbers, and the honest entry in your notes is a question with a date on it, not an inference.

Worked example 3: A quarterly result reconciled to the audited year (India, ₹)

Suraksha Engineering Limited (synthetic) reported consolidated revenue from operations of Rs 742.30, Rs 768.90 and Rs 812.40 crore in its first three quarters, with consolidated profit after tax of Rs 61.40, Rs 64.80 and Rs 71.20 crore. The audited annual results show consolidated revenue of Rs 3,180.20 crore and profit after tax of Rs 246.10 crore, with standalone revenue of Rs 2,610.40 crore and standalone profit after tax of Rs 274.60 crore. The fourth quarter carried a pre-tax exceptional charge of Rs 22.50 crore; the statutory rate is 25.168%. Segment revenue was Rs 1,742.30, Rs 1,027.90 and Rs 471.40 crore with segment results of Rs 289.80, Rs 121.60 and Rs 74.40 crore; unallocable expenditure net of unallocable income was Rs 60.40 crore, finance costs Rs 96.50 crore and tax Rs 82.80 crore.

Solution. Step 1, derive the missing quarter. Nine-month revenue = 742.30 + 768.90 + 812.40 = Rs 2,323.60 crore, so Q4 revenue = 3,180.20 − 2,323.60 = Rs 856.60 crore. Nine-month profit = Rs 197.40 crore, so Q4 profit = 246.10 − 197.40 = Rs 48.70 crore.

Step 2, compare the margins. Nine months: 197.40 / 2,323.60 = 8.4954%. Fourth quarter: 48.70 / 856.60 = 5.6853%. The fall is 2.8102 percentage points in the quarter that was audited rather than reviewed.

Step 3, strip the exceptional item. Post-tax effect = 22.50 × (1 − 0.25168) = Rs 16.8372 crore. Adjusted Q4 profit = Rs 65.5372 crore, a margin of 7.6509%. The exceptional item explains (7.6509 − 5.6853) / 2.8102 = 69.9454% of the fall, leaving 0.8446 points unexplained by it.

Step 4, read the two columns. Subsidiary revenue = 3,180.20 − 2,610.40 = Rs 569.80 crore, or 17.9171% of the group. Subsidiary profit = 246.10 − 274.60 = minus Rs 28.50 crore. Subsidiaries contribute nearly a fifth of revenue and subtract from earnings.

Step 5, tie the segment table out. Segment revenue totals Rs 3,241.60 crore against reported Rs 3,180.20 crore, so eliminations are Rs 61.40 crore. Segment results total Rs 485.80 crore; less unallocable Rs 60.40 crore and finance costs Rs 96.50 crore gives Rs 328.90 crore of profit before tax; less tax Rs 82.80 crore gives Rs 246.10 crore ✓, which is the annual profit derived in Step 1. Segment margins are 16.6332%, 11.8299% and 15.7828%, so the profit engine is the largest segment on both measures, and the smallest segment out-earns the middle one by 3.9528 points.

Step 6, read the review paragraph. The first three quarters carry a limited review, which asserts only that nothing came to the reviewer's attention. The fourth quarter carries no review of its own, because it was never a reviewed period; it is the residue of an audited year. The exceptional item therefore arrived in the one period with the most scrutiny and the least separate disclosure.

Read it. The 0.8446 points the exceptional item does not explain is the finding, not the 2.8102. An exceptional charge is disclosed, quantified and easy to adjust for; residual margin compression in the audited quarter, after adjusting, is the thing nobody labelled. It could be a year-end inventory provision, a bonus true-up, or genuine price pressure that started in January. The tie-out proves the arithmetic is sound, which is exactly what makes the residual worth a question rather than a shrug. What this does not tell you is whether the subsidiaries' Rs 28.50 crore loss is a start-up drag or a structural one, because a single year of the standalone-consolidated gap cannot separate the two.

Worked example 4: A rating rationale converted into a trigger list (India, ₹)

Kalyani Industrial Products Limited (synthetic) has a published rationale from a domestic agency. FY26 EBITDA Rs 918.00 crore, net debt Rs 1,652.00 crore, interest expense Rs 148.00 crore, total outside liabilities Rs 2,410.00 crore, tangible net worth Rs 2,180.00 crore, profit after tax Rs 402.00 crore. Liquidity section: cash from operations Rs 742.00 crore, cash and liquid investments Rs 386.00 crore, undrawn committed lines Rs 450.00 crore, against debt repayments of Rs 310.00 crore and committed capital expenditure of Rs 520.00 crore over twelve months. Sensitivities as printed: downgrade if net debt to EBITDA is sustained above 2.50 times or interest coverage falls below 3.00 times; upgrade if total outside liabilities to tangible net worth is sustained below 1.00 times and coverage is sustained above 6.00 times.

Solution. Step 1, compute today. Net debt to EBITDA = 1,652.00 / 918.00 = 1.7996 times. Coverage = 918.00 / 148.00 = 6.2027 times. Liabilities to net worth = 2,410.00 / 2,180.00 = 1.1055 times.

Step 2, distance to the leverage trigger, both ways. Debt route: the ceiling is 2.50 × 918.00 = Rs 2,295.00 crore, so Rs 643.00 crore of extra net debt breaches it. Earnings route: the floor is 1,652.00 / 2.50 = Rs 660.80 crore, so a fall of Rs 257.20 crore, or 28.0174%, breaches it.

Step 3, distance to the coverage trigger. The interest ceiling is 918.00 / 3.00 = Rs 306.00 crore, more than double today's charge. The EBITDA floor is 3.00 × 148.00 = Rs 444.00 crore, a fall of 51.6340%. Coverage is the slack constraint; leverage binds first, and by a wide margin.

Step 4, distance to the upgrade triggers. Net worth must reach Rs 2,410.00 crore at today's liabilities, a gap of Rs 230.00 crore, which is 0.5721 years of profit at Rs 402.00 crore if all of it is retained. The coverage condition is already met at 6.2027 against 6.00 times, but by only 0.2027 turns, so a 3.27% fall in EBITDA loses it. An upgrade that depends on a condition with three percent of headroom is not an upgrade you should plan around.

Step 5, recompute liquidity. Sources = 742.00 + 386.00 + 450.00 = Rs 1,578.00 crore. Uses = 310.00 + 520.00 = Rs 830.00 crore. Cover 1.9012 times, surplus Rs 748.00 crore. Strike out the undrawn lines and cover falls to (742.00 + 386.00) / 830.00 = 1.3590 times, which is the number to carry, because a committed line is committed until the lender finds a reason.

Step 6, name the refreshing document for each row. Leverage and coverage refresh from the quarterly result, with the debt figure only fully visible in the half-yearly balance sheet the listing regulations require. Net worth refreshes half-yearly. Liquidity refreshes from the rationale itself, the cash flow statement and the borrowings note. Any row you cannot refresh on a schedule comes off the list.

Read it. The table says the rating is not close to moving on earnings, since it would take a 28.0174% fall to breach leverage and a 51.6340% fall to breach coverage. It also says the rating is very close to moving on balance sheet events, because Rs 643.00 crore of borrowing does it without EBITDA changing at all. That asymmetry is the finding, and it points at exactly one thing to watch: capital allocation. What the table does not do is predict the agency's behaviour. Sensitivities say "sustained", agencies exercise judgement about what sustained means, and a rationale is a description of the current view rather than a contract.

Worked example 5: Two subsequent events, classified and costed (India ₹, US $)

Kalyani Industrial Products (synthetic), the company above, has a 31 March year end. Its draft balance sheet shows current assets of Rs 1,480.00 crore and current liabilities of Rs 890.00 crore. Two events occur before the board approves the accounts on 20 May.

Event A: the company failed a net-worth covenant on a Rs 620.00 crore term loan as measured at 31 March. The lender granted a waiver on 12 May covering the next eighteen months.

Event B: on 18 May the company agreed to acquire a business for Rs 1,150.00 crore, funded entirely by new debt, bringing Rs 205.00 crore of EBITDA. Management guides to Rs 38.00 crore of synergies.

Solution. Step 1, classify Event A. The covenant was tested at the reporting date and failed at the reporting date. Under the international and Indian classification rules a liability is current unless the entity has a right at the end of the reporting period to defer settlement for at least twelve months, and a waiver granted on 12 May does not create a right that was absent on 31 March. The loan is current in the balance sheet dated 31 March. The event is not "adjusting" in the sense of changing a measurement; it settles a classification that was already determined at the reporting date.

Step 2, cost it. Current ratio as drafted = 1,480.00 / 890.00 = 1.6629 times, working capital Rs 590.00 crore. After reclassification, current liabilities are Rs 1,510.00 crore, the ratio is 0.9801 times and working capital is minus Rs 30.00 crore. The swing is 0.6828 of a turn, and Rs 620.00 crore of working capital.

Step 3, run the same facts under US practice. Where a waiver is obtained before the financial statements are issued and the conditions for it are met, US practice supports non-current classification, leaving the ratio at 1.6629 times. Identical facts, identical dates, a difference of 0.6828 turns created entirely by which rulebook applies.

Step 4, classify Event B. The agreement was reached after the reporting date, so the condition did not exist at 31 March. It is non-adjusting: no number in the accounts changes, and the note discloses the nature of the event and an estimate of its financial effect.

Step 5, cost Event B anyway. Multiple paid = 1,150.00 / 205.00 = 5.6098 times. Pro forma net debt = 1,652.00 + 1,150.00 = Rs 2,802.00 crore; pro forma EBITDA = 918.00 + 205.00 = Rs 1,123.00 crore; pro forma leverage = 2.4951 times against the reported 1.7996 times, a deterioration of 0.6955 turns. Against the 2.50 times downgrade trigger from the rating rationale, the cushion is 0.0049 turns: Rs 5.50 crore of additional net debt, or Rs 2.20 crore less EBITDA, which is 0.1959% of pro forma EBITDA or a 0.4783% overrun on the purchase price. With the guided synergies, leverage is 2.4134 times.

Read it. Event A changes a ratio without changing a rupee of cash, and it is the reason a cross-border liquidity screen has to know which regime it is reading. Event B changes no ratio on the page and takes the company from comfortably inside a rating trigger to 0.0049 turns from it. The two together are the argument for reading the subsequent-events note before you read anything else in the notes. What the pro forma calculation does not do is settle whether the deal is good; a 5.6098 times multiple on Rs 205.00 crore of EBITDA may be excellent value, and the leverage arithmetic is silent on that. It tells you only what the deal does to the constraint.

Worked example 6: Three true growth rates, and which one goes in the model (US, $)

Corbin Materials Inc (synthetic, Nasdaq: CRBM) reports quarterly revenue of $2,318.00m against $2,030.00m a year earlier. The earnings release headline says revenue grew 14.2%. The investor presentation headline says 18.6%. On the call, the chief financial officer says the deck figure is "constant-currency and organic". Currency moved against the company by $42.00m in the quarter, and the prior-year quarter included $40.00m of revenue from a unit sold during that year.

Solution. Step 1, reproduce the release. 2,318.00 / 2,030.00 − 1 = 14.1872%, which rounds to the 14.2% printed. The release reports what the accounts report.

Step 2, add back the currency. Constant-currency revenue = 2,318.00 + 42.00 = $2,360.00m, so growth = 2,360.00 / 2,030.00 − 1 = 16.2562%. Step size 2.0690 points.

Step 3, restate the base. Like-for-like prior-year revenue = 2,030.00 − 40.00 = $1,990.00m, so growth = 2,360.00 / 1,990.00 − 1 = 18.5930%. Step size 2.3368 points.

Step 4, verify the bridge. 14.1872 + 2.0690 + 2.3368 = 18.5930 ✓, which is the deck headline. The total gap is 4.4058 points, and the deck number is 1.3105 times the release number.

Step 5, price the modelling error. Applying 18.5930% to reported revenue of $2,318.00m gives $2,748.9849m next year against $2,646.8591m on the release rate, a difference of $102.1258m in year one. Compounded over five years the two paths diverge by $937.7356m.

Read it. All three rates are correct and they answer three different questions. Reported growth is what the accounts, the covenant and the market see, and it is the only one that compounds into future reported revenue. Constant-currency growth answers whether the business sold more. Like-for-like constant-currency growth answers whether the retained business sold more, and it is the right lens for judging operating momentum. The error to avoid is mixing them: a model that grows reported revenue at the organic rate has quietly assumed both that currency reverses and that the divested unit comes back. What the bridge does not tell you is which lens management prefers because it flatters them, and the only way to find that out is to build the same bridge for eight consecutive quarters and see whether the preferred definition ever changes when the answer stops being convenient.

Worked example 7: Four quarters of a shareholding pattern (India, ₹)

Nandan Agro Industries Limited (synthetic) has 41.60 crore shares outstanding and trades at Rs 268. Its last four quarterly shareholding patterns show promoter holding of 52.40%, 52.40%, 51.10% and 48.90% of capital, with pledged proportions of the promoter holding of 18.00%, 26.50%, 33.20% and 41.00%. Thirty-day average daily traded value is Rs 18.4 crore. Assume a 45% loan-to-value at origination and a lender maintenance cover of 1.60 times.

Solution. Step 1, convert to a share of the whole company. Multiply the two columns: 9.4320%, 13.8860%, 16.9652% and 20.0490% of total capital encumbered.

Step 2, read the two trends together. Promoter holding fell 3.50 points, so shares were sold. The pledged proportion rose 23.00 points, so the remaining shares were borrowed against harder. Selling and pledging in the same four quarters is the pattern that matters, because either alone has innocent explanations and the pair usually does not.

Step 3, check the disclosure thresholds. Encumbrance has crossed 20% of total capital, one of the two levels at which the takeover code requires promoters to give detailed reasons. Pledge at 41.00% of promoter holding has not crossed the 50% level, so a reader watching only that ratio would have seen nothing.

Step 4, size the block against liquidity. Encumbered shares = 41.60 × 48.90% × 41.00% = 8.3404 crore shares, worth Rs 2,235.2229 crore at Rs 268, which is 121.4795 times a single day's traded value of Rs 18.4 crore.

Step 5, locate the fuse. Implied borrowing at 45% loan-to-value = Rs 1,005.8503 crore. At 1.60 times maintenance cover the margin call fires when the collateral is worth Rs 1,609.3605 crore, which is a share price of 1,609.3605 / 8.3404 = Rs 192.96, or 28.00% below today.

Read it. The 28.00% is a distance, not a forecast, and it is built on two assumed parameters that the disclosure does not give you. The loan-to-value and the maintenance cover are lender-specific and confidential, so a different pair moves the trigger price substantially, and the honest way to present this is as a range with the assumptions named. What the disclosure does give you without assumption is the trend, the crossing of the 20% threshold and the 121.4795 days of volume, and all three are facts. The full cascade, including invocation dynamics and what it does to the listed company's own cash, belongs to the India governance lab and should be run there before this becomes a position decision.

Practice set

Work each problem fully before reading its solution. A calculator is assumed; answers within about 2% score as correct unless a tighter tolerance is stated. All entities are synthetic; every regulatory structure is drawn from the public text of the relevant rules, and every threshold cited as current should be verified against the primary source before it is quoted.

P1 (guided). Split an issue and test the general corporate cap on both bases. An issue is Rs 1,000 crore, of which Rs 620 crore is a fresh issue and Rs 380 crore an offer for sale. The objects list Rs 172 crore of general corporate purposes, against a stated ceiling of 25%. State what share of the money reaches the company, compute general corporate purposes as a share of both the fresh issue and gross proceeds, and say what the two answers imply for the compliance question.

Solution. The company receives the fresh issue only, 620 / 1,000 = 62.0%; the selling shareholders take 38.0%. General corporate purposes is 172 / 620 = 27.7419% of the fresh issue and 172 / 1,000 = 17.2% of gross proceeds. A 25% ceiling is Rs 155.00 crore on the fresh-issue base and Rs 250.00 crore on the gross-proceeds base, so Rs 172 crore breaches one and clears the other. The compliance answer therefore turns entirely on which base the regulation names, and the only acceptable next step is to read the current regulation text rather than to assume it. Computing one ratio and stopping produces a confident answer that is wrong half the time.

P2 (guided). Recompute an issuer's adjusted metric. An issuer reports revenue from operations of Rs 1,860 crore, cost of materials Rs 942 crore, employee benefits Rs 358 crore and other expenses Rs 401 crore. Its Adjusted EBITDA adds back share-based payment of Rs 46 crore, restructuring of Rs 29 crore and offer expenses of Rs 17 crore to the plain operating figure. Compute both margins, the gap in percentage points, and the add-backs as a share of the plain figure.

Solution. Plain EBITDA = 1,860 − (942 + 358 + 401) = Rs 159.00 crore, a margin of 8.5484%. Adjusted = 159 + 46 + 29 + 17 = Rs 251.00 crore, a margin of 13.4946%. The gap is 4.9462 percentage points, and the add-backs are 57.8616% of the plain figure. An add-back stack worth more than half the underlying earnings is not wrong by construction, but it is always the first thing to argue about.

P3 (guided). Derive the fourth quarter. A company's audited annual revenue is Rs 4,412.6 crore and its nine-month revenue was Rs 3,208.1 crore. Audited annual profit after tax is Rs 318.4 crore against nine-month profit of Rs 284.9 crore. Compute the fourth quarter on both lines and the margin fall.

Solution. Q4 revenue = 4,412.6 − 3,208.1 = Rs 1,204.50 crore. Q4 profit = 318.4 − 284.9 = Rs 33.50 crore. The nine-month margin was 284.9 / 3,208.1 = 8.8806% and the fourth quarter is 33.5 / 1,204.5 = 2.7812%, a fall of 6.0994 percentage points on the highest-revenue quarter of the year. Revenue rose and profit collapsed, which points at a year-end provision or writedown rather than at trading, and the next place to look is the exceptional-items line and the auditor's key audit matters.

P4. Covenant headroom, both routes. A borrower has covenant net debt of $942.0m and LTM covenant EBITDA of $386.0m, against a maximum net leverage covenant of 3.25 times. Compute current leverage, the EBITDA floor, the tolerable percentage decline, and the debt headroom.

Solution. Leverage = 942 / 386 = 2.4404 times. The EBITDA floor is 942 / 3.25 = $289.8462m, so the tolerable decline is 24.9103%. Debt headroom = 3.25 × 386 − 942 = $312.50m. Both routes matter because they have different lead indicators: the earnings route is watched through the quarterly result, the debt route through announcements about acquisitions and buybacks.

P5. What a 13F covers. A manager reports $5.13bn of US-listed long equity on its Form 13F and states elsewhere that it manages $11.40bn. Compute the coverage and the invisible share, then list the four categories that could account for the difference.

Solution. Coverage = 5.13 / 11.40 = 45.0%; the invisible share is 55.0%. The four categories are non-US listed equity, private and illiquid holdings, cash and fixed income, and positions expressed through derivatives or short sales that the form does not capture. Any inference about the manager's view drawn from the reported 45.0% is an inference about a minority of the book.

P6. Which trigger binds first. A rated company has EBITDA of Rs 640 crore, net debt of Rs 1,536 crore and interest expense of Rs 96 crore. The rationale names a downgrade trigger of net debt to EBITDA sustained above 2.75 times, and a second at interest coverage below 3.5 times. Compute both distances and say which binds first.

Solution. Leverage = 1,536 / 640 = 2.40 times; the EBITDA floor at 2.75 times is 1,536 / 2.75 = Rs 558.5455 crore, a tolerable fall of 12.7273%, and the debt headroom is Rs 224.00 crore. Coverage = 640 / 96 = 6.6667 times; the EBITDA floor at 3.5 times is Rs 336.00 crore, a tolerable fall of 47.5%. The leverage trigger binds first, since its floor of Rs 558.5455 crore is far above the coverage floor of Rs 336.00 crore. A company can be a long way from one trigger and close to another, and quoting only the comfortable one is how rating surprises happen.

P7. Reclassify a breached loan. A company has current assets of Rs 3,140 crore and current liabilities of Rs 2,050 crore, and a Rs 1,180 crore term loan that breached its year-end covenant, waived after the reporting date. Compute the current ratio before and after reclassification under the international rules, and state whether a lender's separate 1.20 times current-ratio covenant is then satisfied.

Solution. Before: 3,140 / 2,050 = 1.5317 times. After: 3,140 / 3,230 = 0.9721 times, a fall of 0.5596 of a turn. Against a 1.20 times requirement the reclassified ratio fails, which means one covenant breach has mechanically produced a second, and the second may accelerate other debt through a cross-default clause. This cascade is why the classification question is not a presentational nicety.

P8. Build a growth bridge. A company reports revenue of $1,845m against $1,704m. At constant currency the current period is $1,879m. The prior period included $52m from a divested unit. Compute all three growth rates, both bridge steps and the total gap.

Solution. Reported = 1,845 / 1,704 − 1 = 8.2746%. Constant currency = 1,879 / 1,704 − 1 = 10.2700%, a step of 1.9953 points. Like-for-like base = 1,704 − 52 = $1,652m, so like-for-like growth = 1,879 / 1,652 − 1 = 13.7409%, a further step of 3.4710 points. Total gap 5.4663 points. Check: 8.2746 + 1.9953 + 3.4710 = 13.7409 ✓.

P9. Read a pledge line against two thresholds. A promoter holds 61.2% of a company and has pledged 47.5% of that holding. There are 22.5 crore shares outstanding and the price is Rs 412. Compute encumbrance as a share of total capital, state which disclosure thresholds are crossed, and value the encumbered block.

Solution. Encumbrance = 61.2% × 47.5% = 29.07% of total capital, which crosses the 20%-of-capital level. Pledge at 47.5% of the promoter's holding does not cross the 50%-of-holding level. Encumbered shares = 22.5 × 0.612 × 0.475 = 6.5408 crore, worth Rs 2,694.789 crore at Rs 412. A reader watching only the headline pledge percentage would conclude nothing had been triggered, which is why both ratios belong in the same row of your tracker.

P10. Score a use-of-proceeds statement. A company raised Rs 640 crore against a stated deployment schedule of 80% by this quarter, and has utilised Rs 391 crore. Compute utilisation, the unutilised balance and the shortfall against schedule, and state what you would ask on the call.

Solution. Utilisation = 391 / 640 = 61.0938%; unutilised Rs 249.00 crore; shortfall against 80% = 0.80 × 640 − 391 = Rs 121.00 crore. The question is not "why is it slow", which invites a process answer, but "which object is behind, by how much, and has the deployment schedule been revised in the monitoring agency's report", which invites a number and can be checked against the next quarter's filing.

P11. Tie a segment table to profit before tax. A group reports segment revenue of Rs 2,214, Rs 1,102 and Rs 184 crore with segment results of Rs 268, Rs 61 and Rs 96 crore. Reported revenue is Rs 3,388 crore, unallocable expenditure net of unallocable income is Rs 74 crore and finance costs are Rs 118 crore. Compute the eliminations, profit before tax and the three segment margins, and name the profit engine.

Solution. Segment revenue totals Rs 3,500.00 crore, so eliminations are 3,500 − 3,388 = Rs 112.00 crore. Segment results total Rs 425.00 crore; less 74 and 118 gives profit before tax of Rs 233.00 crore. Margins are 12.1048%, 5.5354% and 52.1739%. The engine on absolute result is the first segment at Rs 268 crore, but the third segment produces Rs 96 crore of result on 5.4% of revenue, so on incremental economics it is the business worth understanding. Naming only the biggest is the mistake this problem exists to catch.

P12. Correct a 13F for options. A manager's 13F shows $268m in one name against which it holds puts with a delta-equivalent of minus $196m, and $51m in a second name against which it holds calls with a delta-equivalent of plus $84m. Compute the economic exposure in each and the total error in reading the two lines at face value.

Solution. First name: 268 − 196 = $72m, which is 26.8657% of the reported line. Second name: 51 + 84 = $135m, which is 264.7059% of it. Reported total $319m; economic total $207m; reading the lines at face value overstates net exposure by $112m. The two errors run in opposite directions, so a reader who assumes derivatives "roughly cancel out" across a book gets the direction wrong in both names.

Applied mini-project

The timed sweep: nine documents, forty-five minutes, one page each

Pick one listed company you have not studied and that has been public for at least three years. India or the US, either is fine, and a company with a published rating rationale gives you all nine artefacts rather than eight. Set a timer. Produce nine one-page extracts in the order and to the minute budget in the sweep table, then a tenth page that reconciles them.

The deliverable is graded on one thing above all others, and the rubric is built to enforce it: a fact that changes a number scores, and a fact that is merely transcribed does not. Copying a segment table onto a page is transcription. Tying that segment table to profit before tax, finding the eliminations and naming the profit engine is judgement. Both take about the same amount of time.

Artefact 1 is the annual document, and it gets eight minutes. Extract the segment mix, cash conversion for the year, the one note that moved most against last year, and the auditor's key audit matters by title. Test 1: state cash from operations as a multiple of profit after tax, and say what would have to be true for that multiple to be misleading.

Artefact 2 is the interim result, in six minutes. Extract the latest quarter against the same quarter a year earlier, the segment table, and any exceptional item. Test 2: derive the most recent fourth quarter as the audited year less nine months, on revenue and on profit, and state the derived margin. Test 3: tie the segment table to profit before tax, showing the eliminations and both reconciling lines. If it does not tie, say by how much and where you think the difference sits.

Artefact 3 is the release with the deck, in five minutes. Extract the headline growth rate with its exact definition, and list the metrics on the deck's key-numbers slide. Test 4: build the bridge from reported growth to the deck's headline, with each step labelled and the steps summing to the gap. If the deck uses the reported figure, say so, which is itself a finding.

Artefact 4 is the transcript, and it deserves seven minutes. List the questioners and their firms, mark every question that received a direction rather than a number, and extract three dated, checkable management claims. Test 5: each claim must be written so that a reader in three months can mark it met or missed without judgement. A claim you cannot score records a mood rather than a commitment.

Artefact 5 covers the event filings, in three minutes. List every event filing since the last annual report by item number, or by disclosure category for an Indian company. Test 6: name the single filing that changes a model input, and state which input and by how much.

Artefact 6 is the ownership table, in three minutes. Extract promoter or insider holding and the encumbrance line for four consecutive quarters. Test 7: state both as a trend, and compute encumbrance as a share of total capital rather than of the promoter's holding.

Artefact 7 covers the outside ownership filings, in four minutes. List holders above the disclosure threshold and any change of filing type. Test 8: for one institutional holder, state the four things its portfolio disclosure cannot tell you, with the staleness in days computed from the actual filing date.

Artefact 8 is the rating opinion, in five minutes. Extract the rating, the outlook, and every named sensitivity. Test 9: convert the sensitivities into a table with today's value, the breach point computed two ways, and the document that refreshes each row. If the company is unrated, use its most recent bank-facility rating or say plainly that this artefact is unavailable and why.

Artefact 9 is the subsequent-events note, in four minutes. Extract every event disclosed. Test 10: classify each as adjusting or non-adjusting, and for each non-adjusting event compute its pro forma effect on one ratio that matters.

Artefact 10 is the reconciliation page, written off the clock. One page listing every place two documents disagreed, the ranking rule you applied, and whether the disagreement was about fact or about definition. Close with the three questions you would ask on the next call, each phrased so that the answer is a number.

Rubric: twelve rows scored 0 to 3, so 36 points, pass at 28 with no zero on any row

Any artefact whose test fails scores zero on its row regardless of how good the page looks, and row 12 is scored last, after the other eleven, because it is the only one that measures whether the sweep was worth doing. A beautiful nine-page transcription of nine documents is the exact failure this exercise exists to prevent, and it is what most people produce on the first attempt.

Reading & resources

Each entry below points at a primary source or at the place to find one. Nothing below is a place to take a number from without opening the underlying filing yourself, and every threshold, deadline and cap named in the teaching above should be checked against the current text of these documents rather than against this page.

The Indian rulebook, read as documents rather than as law:

• SEBI (Issue of Capital and Disclosure Requirements) Regulations, 2018, with its schedules, on sebi.gov.in. Schedule VI is the contents of the offer document, and the Basis for Issue Price requirements have been amended more than once. Read the schedule beside an actual prospectus and the section map above becomes obvious. [Free] [Intermediate]
• SEBI (Listing Obligations and Disclosure Requirements) Regulations, 2015. Regulation 33 for results and the review, Regulation 32 for the deviation statement, Regulation 30 with Schedule III for material events, Regulation 31 for the shareholding pattern. The single most useful regulation to have skimmed once end to end. [Free] [Intermediate]
• SEBI (Substantial Acquisition of Shares and Takeovers) Regulations, 2011. Regulation 29 for stake disclosures, Regulation 31 for encumbrance. The Indian equivalent of the 13D world, with different thresholds and a different philosophy. [Free] [Intermediate]
• Ind AS 1 and Ind AS 10, notified under the Companies (Indian Accounting Standards) Rules on the Ministry of Corporate Affairs site. The classification rules and the subsequent-events standard, in the text that actually applies in India. [Free] [Intermediate]
• ICAI Standard on Review Engagements 2410, on the review of interim financial information by the entity's auditor. Eight pages, and it settles what a limited review does and does not do. [Free] [Beginner]

The US rulebook:

• SEC Form 8-K and its General Instructions, on sec.gov. The item list, the four-business-day rule and the filed-against-furnished distinction, in the form's own words. Twenty minutes, and it makes every 8-K readable forever. [Free] [Beginner]
• SEC Form 13F, its instructions, and the Official List of Section 13(f) Securities. The list is what defines the universe, and reading it once is the fastest way to internalise what the filing cannot cover. [Free] [Intermediate]
• Schedules 13D and 13G with their instructions, and the SEC's adopting release for the 2023 amendments to the beneficial-ownership timelines. Item 4 is the part to read closely. [Free] [Intermediate]
• ASC 855 Subsequent Events, and ASC 470-10-45 on classification of debt where a waiver is obtained. The pair is what creates the divergence from the international answer. [Paid] [Advanced]
• EDGAR full-text search at efts.sec.gov, and the browse interface on sec.gov. Free, indexed from 2001, and the single most under-used research tool in the profession. [Free] [Beginner]

Rating agencies, as publishers:

• CRISIL, ICRA and CARE Ratings rationale libraries, free on each agency's site. Read six rationales across the rating spectrum in one sitting and the fixed anatomy becomes automatic. Their criteria documents explain how the sensitivities were chosen. [Free] [Intermediate]
• Moody's, S&P Global Ratings and Fitch rating definitions and national-scale mapping documents. The mapping tables are the antidote to the domestic-AAA trap, and each agency publishes its own. [Free] [Intermediate]
• IOSCO's code of conduct fundamentals for credit rating agencies. Short, and it explains why the rationale has the sections it has. [Free] [Advanced]

Do this rather than only reading. Take one company you already know and download its last four quarterly results, its last annual report and its last rating rationale. Derive the fourth quarter on both revenue and profit. Tie one segment table to profit before tax. Convert the rating sensitivities into five rows with distances. That is about ninety minutes, it produces a monitoring sheet you will keep using, and it teaches the three habits this whole area rests on better than any amount of further reading. Then break something on purpose: change one segment revenue by ten crore and watch the tie-out fail by exactly that amount, so you learn what a failed tie-out feels like before one happens to you by accident.

Teach it back & journal

Feynman prompt. Sit with a colleague who reads only annual reports and explain why the fourth quarter is the most interesting three months of an Indian company's year, using nothing but subtraction. Show them the two lines: nine months of Rs 2,323.60 crore against an audited Rs 3,180.20 crore, and nine months of Rs 197.40 crore of profit against an audited Rs 246.10 crore. Make them do the subtraction themselves, out loud, and then ask them what a margin of 5.6853% in a quarter following three quarters at 8.4954% is likely to contain. Do not tell them about the exceptional item. Let them guess, and only then show them that Rs 22.50 crore explains 69.9454% of the fall and leaves 0.8446 points that nobody labelled.

Then run the second half on a different document. Hand them a rating rationale with the sensitivities highlighted and ask them to say, without any further information, what would have to happen for the rating to move. Push until they give a number rather than a direction, and push again until they give the number two ways, once as extra borrowing and once as lost earnings. If they finish by asking which quarterly filing would tell them first, the habit has transferred.

Journal prompt. For one full quarter, keep a single text file for one company you follow. Every time one of the nine documents arrives, write three lines: what the document said, what it changed, and what it was permitted not to say, and stop there. At the end of the quarter, answer these in writing. How many of the nine documents actually arrived, and which one did you have to go looking for? Then split the changes you recorded by source, counting how many came from the two documents you were already reading before and how many came from the six you were not. Which document produced a fact that changed a number in your model, and which produced only reassurance?

Then write a paragraph on the uncomfortable part. Somewhere in that quarter, two documents disagreed. Find the entry, and answer honestly: did you resolve it by the ranking rule, or did you resolve it by preferring the number that fitted the view you already held? Write down what the disagreement was, what you decided, and what evidence would have made you decide the other way. The claim log is the same discipline pointed at management; this one is pointed at you, and it is the only one of the two that nobody else will ever run on your behalf.