Autos & Industrials

Phase 5 · Week 49 · ~20 hours

Where it sits. Truck sales fall 30–50% when GDP growth falls 5%. Almost everything in this week's work follows from that gap.

The eighth of Phase 5's ten sector playbooks, and your first full immersion in cyclicals that sell machines. Everything before this week was mostly about businesses that sell something every day: loans, policies, subscriptions, soap, prescriptions, code. Now the customer buys once every several years (a car, a truck) or once every few decades (a power plant, a factory, a highway). Lumpy, deferrable, financed purchases generate the rest: violent volume cycles, operating leverage that giveth and taketh away, channel inventories that let managements borrow demand from the future, order books that tell you the future before the P&L does, and working-capital dynamics that kill profitable companies. Two playbooks share the week because they share the disease: autos (OEMs and their ancillary suppliers) and industrials / capital goods / EPC (product makers and project executors). M2's working-capital forensics becomes the channel-stuffing and EPC-cash detectors; the mid-cycle valuation discipline from the valuation phase stops you from buying a trailing P/E at the top; the capital cycle from the strategy phase explains why supply, not demand, sets the returns. Next week's commodities playbook takes the cyclicality one step deeper, to businesses with no pricing power at all. On the five-questions spine (from the master map), the work sits in question 2 (whether it is a good business, answered per segment), question 4 (what can kill it?: inventory, discounts, working capital), and question 5 (what is the price implying?: the cyclical P/E trap).

Suggested hour budget. Core content: 8h · worked examples: 1.5h · practice set: 3h · applied mini-project (the two workups): 4.5h · flashcards + mastery check: 1.5h · teach-back + journal: 1h.

App link. This module powers four drills in the sector case engine. Two of them are the channel-stuffing detector (quarterly wholesale vs registration data; compute the inventory build and call the give-up quarter) and the RM pass-through lag builder (a metal-price path in, a quarterly margin path out). The other two are the order-book console (book-to-bill, backlog coverage, and burn rate from a cap-goods disclosure; forecast next year's revenue) and the EPC cash autopsy (a profitable P&L whose balance sheet is dying: find where the cash went). Your completed KPI tables seed the autos and industrials pages of the sector KPI cheat-sheet artifact. Passing the mastery check ("Sector exam ≥85%") plus the mini-project unlocks next week's playbook, Commodities, Real Estate & REITs.

Learning objectives

By the end of the week you can:

1. Explain OEM economics segment by segment from the cost structure up: why two-wheelers earn 14–18% EBITDA margins, passenger vehicles 8–13%, and commercial vehicles 8–11% through a cycle. Then predict which way each band breaks when volumes, mix, discounts, or raw materials move.
2. Detect channel stuffing with public data: reconcile an OEM's reported (wholesale) volumes against VAHAN/FADA retail registrations, convert the gap into channel inventory days, read rising discounts per unit as the corroborating flag, and quantify the operating-leverage hit when the destocking quarter arrives.
3. Read the auto demand cycle as three engines: replacement demand, first-time buyers, and financing availability. Apply the segment-specific versions (freight economics for CV, penetration and model cycles for PV, rural cash flows for 2W), including how to read PV/CV order backlogs honestly.
4. Assess the EV transition without cheerleading or denial: per-vehicle economics (battery cost, contribution margin), the cash-burn-versus-optionality trade, what "EV mix" does and does not tell you, and the Tesla-versus-Toyota lesson about integration, distribution, and margin durability.
5. Analyze an auto ancillary as revenue = industry volume × content per vehicle × share. Compute kit-value growth, judge OEM/client concentration and the aftermarket/OEM/export mix, model the RM pass-through lag quarter by quarter, apply the ROCE discipline test, and classify a component portfolio as ICE-at-risk, EV-agnostic, or EV-upside.
6. Explain the order-book business model (book → execute → collect) and run its KPI panel: order inflow, backlog, book-to-bill (>1× = growing), backlog coverage (EPC ~2.5–3.5× revenue), execution/burn rate, capacity utilization (>80% = pricing power), fixed-asset turns, and ROCE against the ~12% hurdle.
7. Show with numbers why EPC companies die of working capital, not P&L: receivables, unbilled revenue (contract assets), retention money, and the bank-guarantee transmission channel. Then contrast the product model (Siemens/ABB-style, customer advances, negative working capital) with the project model (L&T-style) and a US product-plus-dealer anchor (Caterpillar-style).
8. Choose the right valuation lens: EV/EBITDA and EV/EBIT on mid-cycle earnings for anything cyclical, P/E only for franchise-grade compounders within these sectors, never a trailing P/E at a cycle peak. Separate captive-finance books from industrial economics before computing any multiple.

Prerequisites & connections

Builds on. M1.06–M1.08 on inventory and revenue recognition. You'll now meet over-time, percentage-of-completion revenue in the wild, plus the R&D-capitalization comparability trap. M2.03–M2.05 gave you the core forensic toolkit both playbooks run on: working-capital days, cash conversion, CFO/EBITDA. M2.06 treated channel stuffing as a shenanigan family, and here it gets a sector-specific detector built on public data. M3.07–M3.08 covered multiples and mid-cycle normalization. M4.03 taught Five Forces, which you'll run on the brutal PV industry and the cozy 2W duopoly-plus. M4.06 is the capital cycle, and capacity added at the top is the industrials disease.

Feeds into. M5.09, on commodities, runs the same cycle logic with zero pricing power. Metals are these companies' raw materials, so the RM-lag mechanics carry straight over. M5.10 is the cross-sector multiples map, and autos and industrials supply its cyclical column. In Phase 8 the rapid teardown of any manufacturer leans entirely on these KPI panels. Phase 9 covers position sizing for cyclicals, where you size on mid-cycle value and never on peak earnings momentum.

The one-line bridge from M5.07. Pharma sold molecules with patent clocks. Autos and industrials sell metal with demand clocks. In both, the reported quarter tells you almost nothing until you know where you are on the clock.

Core teaching content

4.1 Two families, one disease: deferrable demand

Start with the customer, as always. Nobody must buy a new truck this year. A fleet operator can run his truck a ninth year; a family can keep the old hatchback; a factory can sweat its machines; a state government can delay the highway tender. Nobody defers insulin, which is what last week's businesses were selling. Out of that difference comes the single most important sentence you will read all week:

When purchases are deferrable and financed, demand doesn't decline in a downturn. It disappears.

A 5% fall in GDP growth doesn't produce a 5% fall in truck sales. It produces a 30 to 50% fall, because everyone defers at once and the people who wouldn't have deferred can't get financing. Recoveries are just as violent the other way, with three years of deferred replacement arriving in eighteen months. Now layer on high fixed costs (plants, dealer networks, engineering staff) and you get operating leverage. Margins expand ferociously on the way up and evaporate on the way down. A business with a 25% contribution margin and heavy fixed costs can swing from a 14% EBITDA margin to 2% on a 30% volume decline. Nobody committed fraud and nobody mismanaged anything. It's arithmetic.

That arithmetic raises three questions, and the rest of the week builds the tools to answer them:

1. Where are we in the cycle? Look at volumes against trend, capacity utilization, fleet age, financing availability, order books.
2. Is reported demand real demand? Look at wholesale against retail, channel inventory, discounts, order-book quality, unbilled revenue.
3. What are mid-cycle economics worth? Never value on trailing peak earnings, and never panic at the trough either.

There's a valuation corollary you'll apply all week. Cyclicals get EV/EBITDA and EV/EBIT on normalized, mid-cycle earnings. The trailing P/E is lowest right at the peak, when E is at its high and the market is already discounting the downturn, and it's highest at the trough. A "cheap" auto stock at 6× peak earnings is often expensive. A "shocking" 40× trough P/E is often cheap. You learned the principle in the valuation phase, and this week you practice it on real furniture.

4.4 Playbook 3: Industrials, capital goods & EPC: the order-book business model

4.4.1 Book → execute → collect

A capital-goods or EPC (engineering-procurement-construction) company sells things that take months or years to build: turbines, switchgear, factories, metros, refineries, highways. That one fact restructures the whole analysis:

1. Book. You win orders through competitive government and EPC tenders or through negotiated equipment sales. The order book, or backlog, is tomorrow's revenue signed today. Think of it as the leading P&L, since revenue is just old orders maturing.
2. Execute. You convert backlog to revenue over the project life. Under Ind AS 115 / IFRS 15 / ASC 606, the same five-step model you learned with revenue recognition, long-duration contracts typically recognize revenue over time. They use percentage-of-completion measured by input, cost-to-cost, so reported revenue and margin embed management's estimates of total contract cost. Estimate revisions arrive as lumpy "one-time" margin hits, usually late in a project's life. That's the notorious 90%-complete margin cliff. The gap between work performed and amounts billed goes on the balance sheet as contract assets, or unbilled revenue, a line you'll learn to fear a few pages on.
3. Collect. You bill per milestone and chase government counterparties. You also absorb retention money, the 5 to 10% of contract value held back until the defect-liability period lapses, often 12 to 24 months after completion. Cash arrives years after profit is booked.

Those three steps dictate the KPI panel. One set of dials covers the book: inflow, backlog, book-to-bill, backlog quality. Another covers execution: burn rate, utilization, margins. A third covers collection: working capital, net debt, CFO/EBITDA. ROCE is the summary statistic over all three.

4.4.2 Product vs project: the most important split in the sector

Before you compute any ratio, classify the company. Here are the two poles, with the anchor companies for each:

Siemens India / ABB India-style economics (illustrative, modeled on ~FY24/FY25 scale, so pull the filings) look like this. Revenue runs ₹12,000 to 22,000 crore on net cash balance sheets. Customer advances are so large that core working capital is negative, and ROCE is north of 25%. A branded, technology-differentiated product sold in weeks doesn't lend its balance sheet to customers. L&T-style economics (illustrative, same caveat) are ~₹2.2 to 2.6 lakh crore of consolidated revenue, an order book of ~₹5 to 6 lakh crore, and a group ROE in the mid-teens after decades of discipline. Watch the consolidation trap. L&T's listed IT-services subsidiaries inflate consolidated margins and depress apparent coverage. So always compute EPC metrics on the core projects business, from its segment note. There backlog coverage runs ~3× core revenue, squarely in the brief's 2.5 to 3.5× band.

The US anchor, Caterpillar-style (illustrative, modeled on ~FY24/FY25 scale; pull the 10-K). This is the third body plan. Caterpillar sells products, machines plus high-margin aftermarket parts, through ~150+ independent dealers. That service network is the moat, because nobody can replicate it. A captive lender, Cat Financial, finances the sales, so the captive-finance split from the OEM accounting traps applies. Revenue was ~$65bn± with ~20% operating margins at the recent peak. The same company printed ~$38 to 39bn of revenue at the 2016 trough, after a ~40% peak-to-trough slide across four straight down years.

So Caterpillar is the permanent reminder that best-in-class and cyclical are compatible descriptions. Its quality showed in how it behaved at the trough. It stayed profitable, kept the dealer network fed and bought back stock, and none of that shows up in a peak margin. Read its backlog (~$27 to 30bn, order of magnitude) the way you read L&T's, on coverage, quality and burn.

4.4.3 The order-book KPI panel

Reading the book's quality as well as its size. Ask five audit questions.

1. Margin at bid. Is the book fixed-price or cost-plus, and does it carry input-indexed escalation clauses or is it naked? A fixed-price book bid during a steel trough is a collection of landmines.
2. Mix. What share is low-margin "turnover orders" taken to keep crews busy? A backlog "padded with low-margin orders" is the brief's flag, and you see it when backlog grows while embedded margin guidance slips.
3. Counterparty. Central government / PSU / private / overseas: who pays on time? Who'll arbitrate disputes for a decade?
4. L1 versus signed, an India convention. Companies announce "L1 positions," meaning they were the lowest bidder and haven't been awarded yet. L1 is a lottery ticket. Leave it out of backlog.
5. Ageing and movement. Orders that neither burn nor cancel are zombie backlog. A coverage ratio inflated by zombies is worse than a smaller, cleaner book, so >4 to 5× coverage should prompt a question before anyone treats it as a boast.

Forecasting revenue from the book takes one line: Next-year revenue ≈ opening backlog × burn rate + short-cycle orders booked-and-billed in year. With a backlog of ₹12,000 cr, burn at ~30% and ~₹1,500 cr of book-and-bill, you're looking at ~₹5,100 cr of high-visibility revenue before the year even begins. So industrials investors read order announcements the way FMCG investors read volume growth. It's the growth line, a year early.

4.4.4 Why EPC dies of working capital, not P&L (worked)

If you remember one thing from the industrials playbook, make it this. An EPC P&L can show growth and profit for years while the balance sheet fills up with claims on other people's cash. There are receivables, billed but unpaid. There's unbilled revenue, the contract assets, which is work done but not yet billable and a pure estimate. And there's retention money, contractually stuck for years. Meanwhile growth consumes cash, because every new project mobilizes labor, materials, and site costs before any milestone bills.

Watch the mechanism first, then the kill shot:

"Stambh Infra" (synthetic EPC). In Year 0 revenue is ₹4,000 cr and EBITDA at 10% is ₹400 cr. Depreciation is ₹60 cr, interest is ₹80 cr on net debt of ₹800 cr @10%, and tax at 25% leaves PAT of ₹195 cr. Net working capital runs at 25% of revenue, or ₹1,000 cr: receivables 75 days + unbilled 55 + retention 25 + inventory 15 − payables 55 − advances 24 ≈ 91 net days. Capex is ~1.5% of revenue. EPC is asset-light, and capex was never the problem. Now let it win. Revenue grows 25%/year for three years while margins hold. Everyone cheers. But the incremental orders come from slower-paying clients, accepted to sustain the growth. So NWC intensity creeps from 25% to 27%, then 29%, then 31% of revenue. That's +22 receivable/unbilled days over three years, a mild case by graveyard standards.

Read the poison in three lines. PAT compounds at ~32% a year, and the company still can't pay for its own growth. Net debt rises 54% across three record years, and cash conversion collapses toward zero. Here's the mechanism. At a flat 25% intensity, a quarter of every incremental revenue rupee gets locked into working capital before anyone is paid. So ΔNWC alone eats ~50% of EBITDA at this growth rate. Once intensity creeps, 70 to 80% of EBITDA never becomes cash (in Yr 1, 350 of 500). Interest compounds along with the debt. The P&L never flags a crisis, because EBIT/interest coverage still reads a comfortable ~7×.

The company is profitable and growing, and it's financing its own customers into its own insolvency.

The kill shot never comes from the P&L. It comes through the bank-guarantee channel. EPC runs on non-fund-based limits: bid bonds, performance guarantees (~5 to 10% of contract value), advance-payment guarantees. Banks size those limits off the balance sheet they see. So picture Year 4. One large government client delays certification, and receivables plus unbilled jump 30 days, which strands ~₹640 cr of extra cash. Net debt crosses ~₹1,475 cr, nearly double what it was three years ago. The rating agency cuts one notch, and the banks freeze incremental BG limits. Now follow the chain. With no BG limit the company can't bid. Order inflow stops and book-to-bill collapses. Fixed costs land on a shrinking revenue base and margins go negative. Then a covenant breaks.

The company dies of cash while its last reported P&L showed record profit.

Compress that into one graveyard sentence and you have IL&FS's construction arm, IVRCL, HCC, Punj Lloyd, Gammon. India's 2010s EPC extinction was a working-capital event. The demand, infrastructure, was fine. The survivors, L&T-style, were the ones who kept working capital on a short leash. They walked away from clients who don't pay, collected advances at signing, and let book-to-bill dip before they'd buy bad orders.

Your standing EPC test. Memorize it: revenue growth is only real if working capital grows slower. Track the NWC-days trend. Track cumulative CFO/EBITDA over 3 to 5 years, which should clear 60 to 70% or come with an explanation. Compare unbilled revenue growth with revenue growth, since unbilled outgrowing revenue means estimate risk or certification disputes piling up. And watch net debt against backlog. Debt should fund execution capability, and it shouldn't be there to disguise uncollected billings. The product, short-cycle model makes the same point in reverse. Run Stambh's exact growth path with advances-funded negative WC at −5% of revenue, and Year-1 free cash flow swings from −₹94 cr to a positive ~₹375 cr. Same P&L, opposite fate. Worked example 3 below does this line by line.

4.4.5 Margins, the cycle, and the capital cycle

Capital-goods margins are a utilization story wearing a pricing story's clothes. Below ~70% utilization every player discounts to cover fixed costs. Gross margins compress and "aggressive bidding" headlines appear. That's the brief's under-bidding flag: winning tenders at margins that only work if nothing goes wrong, and something always goes wrong. Above ~80% utilization, lead times stretch, pricing power returns and margins expand. Then managements reach for the capex chequebook, and the capital cycle plays out on schedule. The industry adds capacity at peak margins, and the capacity lands 2 to 3 years later into softening demand. Utilization and margins crash, and the players who held back pick up the trough plants for cents on the rupee.

So time-stamp any industrials thesis with two questions. Where is utilization, and what's the industry's aggregate announced capex? High utilization with low industry capex is the sweet spot, pricing power without a supply wave behind it. High margins with everyone announcing greenfields means sell the euphoria, whatever this quarter prints. India's 2003–08 cap-goods boom-and-bust and its 2021–25 revival both follow this script. Verify current-cycle capex data in the RBI's OBICUS capacity-utilization survey. It's free and quarterly, and almost nobody reads it.

4.4.6 Valuation across the two playbooks

• The default lens is EV/EBITDA and EV/EBIT on mid-cycle earnings, for OEMs, cyclical ancillaries, and capital goods alike. Prefer EV/EBIT where D&A games or capitalization differences distort EBITDA, as with the R&D trap. Compute mid-cycle EBITDA as normalized volumes × mid-cycle margin. Normalized volumes come from a trend line through 7 to 10 years, and the mid-cycle margin is the midpoint of the segment band or the company's own cycle average. Next week's commodities playbook gives you the full method. Here you apply the light version.
• P/E only for franchise-grade names. That means the 2W premium brand, the negative-WC product multinational, the content-compounding ancillary with 20%+ ROCE. In all of them earnings are less cyclical and cash conversion is proven. Even then, reverse the multiple to see the expectations it implies, and check those against realistic volume cycles.
• The trailing-P/E trap, stated once more because it costs real money. Cyclicals look cheapest at the peak, with record E and a single-digit P/E. The market isn't asleep there. It's forecasting. They look most expensive at the trough, where E has collapsed, the P/E is absurd, and you're often in the buy zone. The global mass-OEM corollary is that European and US volume carmakers have traded at 4 to 8× P/E for a decade and stayed "cheap." They have poor FCF conversion, EV-transition capex, pension/finance black boxes, and China exposure. A low multiple on a structurally FCF-poor cyclical is a fair price for what you get. The value-trap test is always FCF conversion and mid-cycle ROCE, never the multiple alone.
• Valuing EPC. EV/EBITDA means little if EBITDA never becomes cash, so cross-check every EPC multiple against CFO/EBITDA and the NWC-days trend. In distress the market moves to P/B, and then to option value on the BG limits, which is to say survival itself. Use sum-of-the-parts wherever conglomeration muddies the picture. For an L&T-style group that's core E&C + listed IT subsidiaries at market value + the finance book on P/B + development assets at NAV. Then take off the 15 to 25% holding-company discount India habitually applies to the parts, the one covered in the valuation phase.
• ROCE is the through-line. Whatever the multiple says, growth at sub-12% ROCE destroys value, because that hurdle is the program's proxy for WACC. A fast-growing order book at value-destroying bid margins deserves a discount, and certainly no growth premium. The key value-driver formula from the valuation phase shows why: Value = NOPAT × (1 − g/ROIC) / (WACC − g). Raise g when ROIC < WACC and value falls.

4.4.7 Industrials/cap-goods/EPC red flags (the brief's list, operationalized)

1. Book-to-bill <1× for several quarters. The revenue decline is already signed, and the P&L just hasn't printed it yet.
2. Backlog padded with low-margin orders. Inflow goes up while embedded margin goes down. "We bid to keep teams busy" is the sentence that comes before the write-offs.
3. Working capital and net debt rising faster than revenue. That's the Stambh spiral, and you corroborate it with cumulative CFO/EBITDA below 40 to 50%.
4. Under-bidding. The company wins share in an over-supplied tender market. Check competitor commentary for "irrational pricing," and check the margin-at-bid disclosures.
5. Execution slippage. Burn rate falls and projects get stuck waiting on client funding, on land, on approvals. Revenue guidance gets cut while backlog "grows."
6. Unbilled revenue outgrowing revenue. Estimates and disputes pile up ahead of certification. It's the EPC cousin of receivables stuffing.
7. Single-client or single-sector dependence. One government program or one anchor customer makes up most of the backlog, so a policy pause becomes an extinction event.
8. Peak-cycle greenfield announcements. This is the capital cycle's sell signal, dressed up as confidence.

4.5 The sixty-minute teardown map for this module's sectors

When Phase 8's clock starts on a manufacturer, this is the routing table:

The same three questions run across all four columns, and they're the spine of the week: where you are in the cycle, whether reported demand is real, and what mid-cycle economics are worth.

Common mistakes & how experts think differently

1. Buying the single-digit trailing P/E at the cycle peak. The novice screens for "cheap." The expert knows a cyclical's P/E is lowest at the top, because the market is already pricing the downturn. Experts value on mid-cycle EBITDA and read a low peak P/E as a warning sign.
2. Treating wholesale volumes as demand. The company's press release is a dispatch number. Experts keep a standing SIAM-minus-VAHAN spread sheet and never call a demand trend without the retail line and channel-inventory days.
3. Extrapolating the CV upcycle (or trough). Three years of 30% CV growth doesn't mean India has changed. It's deferred replacement being released. Experts date the cycle (fleet age, freight rates, financing) and assume mean reversion in both directions.
4. Reading an ancillary like a mini-OEM. The OEM's story is volumes. The ancillary's story is volume × content × share. Experts want the content or kit-value proof, revenue persistently outgrowing industry volumes, before they pay a growth multiple.
5. Panicking at (or ignoring) the RM squeeze quarter. The lag is mechanical and symmetric. Experts model it and watch the spread per unit. They get suspicious only when the spread itself compresses, because that's pricing power eroding, a different disease with the same symptom.
6. Celebrating order inflow without auditing the book. Headlines report inflow. Experts read margin at bid, counterparty, escalation clauses, L1 versus signed, ageing. A ₹10,000 cr order at a 6% embedded EBITDA margin with no escalation clause is a liability announcement.
7. Believing an EPC P&L. EPS can grow for five straight years while the company heads toward death, as Stambh showed. Experts read EPC statements cash first: CFO/EBITDA, NWC days, unbilled trend, BG headroom. They check the P&L last, as the estimate it is.
8. Scoring the EV transition on mix percentage. Mix is a press-release metric. Experts price the option: the per-unit contribution trajectory, capex intensity, how candid the disclosure is. A celebrated mix with hidden unit economics is a red flag to them, and no thesis at all.

Worked examples

Worked example 1: the channel-stuffing autopsy, completed ("Bharat Wheels", synthetic)

Pick up the Bharat Wheels detection where it left off and follow it into the P&L, with every number shown. This is the position at the Q4 exit. Channel inventory is 1,44,000 units, or 64 days. Retail is flat at ~2.06 lakh a quarter, ASP is ₹6,00,000 and the contribution margin is 25%. Fixed costs run ₹1,800 cr a quarter, and discounts are already at ₹31,000/unit.

Step 1: the required destock. Target inventory of 35 days = 2,264/day × 35 ≈ 79,000 units. Excess = 1,44,000 − 79,000 = 65,000 units to drain. Plan to drain it over two quarters, with retail assumed flat. Per-quarter wholesale = retail − drain/2 = 2,06,000 − 32,500 = 1,73,500 units. That's against 2,40,000 just reported, a fall of 28% QoQ.

Step 2: the destock quarter P&L. Revenue = 1,73,500 × ₹6,00,000 = ₹10,410 cr, down from ₹14,400 cr. Contribution = 25% × 10,410 = ₹2,602 cr. EBITDA = 2,602 − 1,800 = ₹802 cr, a margin of 7.7% against 12.5% before. Volume falls 28% and EBITDA falls 55%. The operating-leverage multiplier is ≈ 2×, just as the opening warning about deferrable demand said it would be.

Step 3: the discount drag persists. To keep retail moving through the destock, discounts have to stay at ₹31,000 against the normal ₹18,000. The extra ₹13,000 per wholesaled unit is 13,000 ÷ 6,00,000 ≈ 2.2% of realization. That takes another ~2.2 points off EBITDA margin relative to the old normal, or ₹226 cr of concession a quarter on 1,73,500 units. So even the recovery quarter starts ~2 points below history, and it stays there until inventory normalizes and the price support is withdrawn.

Step 4: what a careful analyst wrote three quarters early. "Wholesale +9%, VAHAN retail +1%, channel days 30→48, discounts +₹9k/unit: growth is warehoused, not sold. Model a 2-quarter destock: revenue −25–30% QoQ, EBITDA margin 7–8%, and cut the multiple. Stuffers de-rate before they destock." If you can write that memo from public data, you've got what this week is for.

Worked example 2: "Meridian Harness": the pass-through lag meets the content story (synthetic)

Take the pass-through lag table from earlier, stretch it into a full-year read, and then lay the growth story on top.

Step 1: the year of margin optics. Sum the four quarters, leaving out the Q0 baseline. Revenue = 100.0 + 104.8 + 104.8 = ₹309.6 cr for Q1 to Q3. Gross profit = 35.2 + 40.0 + 42.4 = ₹117.6 cr. Blended GM = 38.0% against the 40.0% baseline. A screen sees "margin down 200bps." In truth the spread was fully preserved from Q2 (₹40.0 cr) and over-earned in Q3 (₹42.4 cr), thanks to the falling-copper flatter. If copper simply stays at 110, Q4 prints like this. The revenue reset comes off Q3's 110, so the copper component of price = 24 × 1.10 = 26.4 and revenue = 102.4. RM = 62.4, GP = ₹40.0 cr and GM = 39.1%. The spread has been ₹40 cr in every quarter except the squeeze. Grade the business on the spread, and grade the quarter on the lag.

Step 2: overlay the structural story. Same company, next year. Industry 2W volumes grow 5%. Meridian's kit value per vehicle rises 9%, because it won the LED lamp and the EV high-voltage harness on two new platforms. Its share is up 1%. Revenue growth = 1.05 × 1.09 × 1.01 − 1 = 15.6% in a 5%-volume industry. Design-led parts give an incremental contribution margin of 22% against the company average of 18%. So the incremental EBIT on ₹48 cr of new revenue, on the ~₹310 cr base, is ≈ ₹10.6 cr. Set that against maintenance-plus-growth capex of ₹18 cr and ΔWC of ₹6 cr, which makes incremental capital ₹24 cr. That's ₹10.6 cr of incremental EBIT on that capital, an incremental pre-tax ROCE of ≈ 44%. Content growth through an existing plant is the highest-quality growth in the sector, and this is what passing the ROCE test looks like.

Step 3: the classification footnote. Meridian's book holds harnesses, which are agnostic moving to upside, and lighting, which is agnostic. It also holds a legacy fuel-hose line at 12% of revenue, and that's ICE-at-risk. In a deep dive you'd value the 88% as a compounder and run the 12% as a 15-year annuity fading to zero. Then check that management isn't green-fielding fuel-hose capacity out of habit. Portfolio X-rays are done line by line, never at the company level.

Worked example 3: "Stambh Infra" vs "PrecisionGrid Electricals": same P&L, opposite fates (synthetic)

This is the structural contrast underneath the Stambh spiral, computed side by side. This time there's no intensity creep, so you can isolate the one variable. Both companies start at revenue of ₹4,000 cr and grow 25% for 3 years. Both run a 10% EBITDA margin, depreciation of ₹65/72/80 cr, capex at 1.5% of revenue and tax at 25% of PBT. Stambh opens with net debt of ₹800 cr and pays 10% on it. PrecisionGrid opens with net cash of ₹200 cr and earns 6%. Then comes the one difference that matters. Stambh runs NWC at +25% of revenue, held flat with perfect discipline, while PrecisionGrid runs it at −5%, because its customer advances exceed its receivables plus inventory.

Year 1, line by line (₹ cr):

Run it to Year 3, with the same mechanics each year. Stambh's FCF after interest goes +6, then +21, then +40, and its net debt falls from ₹794 to ₹773 to ₹733 cr. Cumulative free cash across three boom years comes to +₹67 cr, which is 3.5% of the ₹1,906 cr of EBITDA it reported. Its CFO/EBITDA is stuck at ~30 to 32%. PrecisionGrid's FCF goes +375, then +481, then +614, and its net cash builds from ₹575 to ₹1,056 to ₹1,671 cr. Its cumulative free cash is +₹1,470 cr, or 77% of the same EBITDA, and its CFO/EBITDA runs ~86 to 88% because advances arrive with orders. Identical revenue and EBITDA lines, and a ₹1,400 crore cash gap opens in three years.

Now look at what "perfect discipline" bought Stambh. It bought survival, and not much more. Stambh is running hard to stand still. Let NWC intensity slip two points a year and this table turns into the spiral you saw earlier.

Now the ROCE verdict. In Year 3, Stambh's capital employed is NWC 1,953 + NFA ~350 = ₹2,303 cr. EBIT = 781 − 80 = ₹701 cr, so ROCE ≈ 30%, which looks great! Here's the trap. That "ROCE" is earned on capital the balance sheet has to keep borrowing to hold. Every marginal rupee of growth requires a rupee of funded WC, and the BG channel caps how many rupees the banks will stand behind. For PrecisionGrid, NWC −391 + NFA ~350 ≈ −₹41 cr, so its capital employed is negative. ROCE can't be meaningfully computed when customers fund the business, and every growth rupee adds cash. So the market hands Siemens/ABB-India-style names 40 to 60× P/E and hands EPC 8 to 12× EV/EBITDA with a cash-conversion asterisk.

The multiple gap comes from the sign of ΔNWC. Sentiment has nothing to do with it.

So whenever an industrials company reports 25% growth, make one subtraction before anything else: ΔNWC against EBITDA. Everything else in the workup hangs off that ratio.

Worked example 4: "Shakti Cranes": capacity utilization, operating leverage, and reading it off the order book (synthetic)

Shakti makes material-handling equipment. It's a product, short-cycle industrial, so don't treat it as an EPC. Nameplate capacity is 12,000 cranes/year. ASP is ₹20 lakh (₹0.20 cr), held flat across both years so the entire margin move comes from volume. Contribution margin is 30%, and fixed costs (plant, engineering, corporate) are ₹360 cr/year, flat across the operating range. It's the same operating-leverage engine you met with trucks, in a capital-goods body.

Step 1: utilization and revenue. In Year A output is 7,800 cranes, so utilization = 7,800 ÷ 12,000 = 65% and revenue = 7,800 × ₹0.20 cr = ₹1,560 cr. In Year B output is 10,200, utilization is 85% and revenue is ₹2,040 cr. Year B has crossed the ~80% line where pricing power returns, but we hold ASP flat here to isolate the leverage.

Step 2: the operating-leverage arithmetic. EBITDA = 30% of revenue − ₹360 cr fixed.

• Year A: contribution 0.30 × 1,560 = ₹468 cr, and EBITDA 468 − 360 = ₹108 cr, a 6.9% margin.
• Year B: contribution 0.30 × 2,040 = ₹612 cr, and EBITDA 612 − 360 = ₹252 cr, a 12.4% margin.

Revenue rose 2,040 ÷ 1,560 − 1 = +30.8%. EBITDA rose 252 ÷ 108 − 1 = +133.3%. That's an operating-leverage multiplier of 133.3 ÷ 30.8 ≈ 4.3×, far above the ~2× of the Bharat Wheels OEM case. Why so much bigger? Look at the starting point. Year A keeps only ₹108 cr over ₹360 cr of fixed cost, so it's close to breakeven, and that's where the amplification is most violent in both directions. A 30% volume fall from Year B would erase roughly three-quarters of EBITDA. Utilization drives the whole margin here, and the leverage multiplier is largest just where the reported margin looks safest to extrapolate.

Step 3: time the ramp with the order book. In the same year opening backlog is ₹1,800 cr, order inflow ₹2,400 cr and revenue ₹2,040 cr. Book-to-bill = 2,400 ÷ 2,040 = 1.18×, above 1, so the backlog is growing. Closing backlog = 1,800 + 2,400 − 2,040 = ₹2,160 cr. Coverage = 2,160 ÷ 2,040 = 1.06×, squarely in the 0.5 to 1.5× product-business band and well short of the EPC's 2.5 to 3.5×. So the book confirms the utilization ramp is demand-led. Nothing's been parked in a channel.

Step 4: the capital-cycle trap the good year plants. At 85% utilization with a 1.18× book, management will reach for a greenfield. That's the capital cycle's sell-signal setup, because capacity ordered at peak margins lands 2 to 3 years later into whatever demand exists by then. So the verdict line reads: value this on mid-cycle EBITDA, with utilization normalized toward ~75% and a margin of ~9 to 10%, never on the 12.4% Year-B print. Read a peak-margin greenfield announcement as a reason to trim the multiple.

Worked example 5: "Apex Driveline Corp": content per vehicle and the EV-transition X-ray, in dollars (synthetic)

Apex is a US-listed global auto-components supplier of electrical architecture and driveline parts, so this is the ancillary model in dollars. Its addressed light-vehicle market is 16.0 million units/year. Apex holds 12.5% of the platforms it competes for, so it touches 16.0m × 12.5% = 2.0 million vehicles. Its content per vehicle is $400 on an ICE platform and $1,000 on a BEV platform. That's the ~2.5× EV kit-value uplift you saw in the content story (high-voltage harness, power electronics, thermal management). All figures are synthetic.

Step 1: blended content as EV mix rises. Blended content/vehicle = (1 − EV%) × $400 + EV% × $1,000.

• Year 1, EV mix 8%: 0.92 × 400 + 0.08 × 1,000 = 368 + 80 = $448.
• Year 4, EV mix 20%: 0.80 × 400 + 0.20 × 1,000 = 320 + 200 = $520.

Step 2: revenue growth from content alone. Hold industry volumes (16.0m) and share (12.5%) flat to isolate the content multiplier, so revenue = 2.0m × blended content. Year 1 gives 2.0m × $448 = $896m and Year 4 gives 2.0m × $520 = $1,040m. Growth = 520 ÷ 448 − 1 = +16.1% over three years, a 5.1% CAGR, with a flat market and flat share. That's the content story in dollars. Content per vehicle is a structural multiplier that manufactures growth where volumes give none.

Step 3: the full multiplicative overlay. Now let one year fire all three engines. Industry volumes rise 3%. Content per vehicle rises 6% on EV mix plus a safety-sensor mandate. Share gains 1 point, from 12.5% to 13.5%, which is a 1.08× factor. Revenue growth = 1.03 × 1.06 × 1.08 − 1 = +17.9%. The additive shortcut gives 3 + 6 + 8 = 17.0%, and it understates because the factors compound, just as they did in the consumer playbook's price × volume split. A 17.9% grower in a 3%-volume market is a content-and-share machine. It isn't a bet on the US SAAR.

Step 4: the same market, the opposite fate. Take a peer that sells only exhaust and fuel-injection parts, with $300 of ICE content and $0 on a BEV. Its blended content is 0.92 × 300 = $276 in Year 1 and 0.80 × 300 = $240 in Year 4. That's a −13.0% revenue slide from mix alone, and it accelerates toward zero as EV penetration climbs. The industry and the volumes are identical. Apex compounds because electrification arrives, and the exhaust peer melts because it does. You've just run the portfolio X-ray with two numbers. Value Apex as an EV-upside compounder and the peer as a finite-life run-off, and never let one "auto-components" label price them the same.

Practice set

Work every problem before you read its solution. Currency is ₹ crore unless stated, and "days" are on 91-day quarters or 365-day years as stated.

P1 (guided: margin bands). An analyst compares three OEMs. A makes two-wheelers at an EBITDA margin of 16.5%, B makes passenger vehicles at 11.8%, and C makes commercial vehicles at 9.6%. The analyst concludes, "A has the best management because it has the highest margin." Correct the reasoning in three sentences, using the segment margin bands.

Solution. All three are comfortably inside their segment bands (2W 14–18%, PV 8–13%, CV 8–11%). So the margin differences reflect segment physics (ticket size, brand pricing power, competitive intensity, buyer rationality) and say nothing about management quality. Comparing margins across segments is a category error. The right test is each company's position within its band, its trend across the cycle, and its ROCE. On the evidence given, all three are performing well for their segments, and nothing distinguishes management skill.

P2 (guided: channel inventory days). A PV OEM opens with dealer stock of 54,000 units. Quarterly wholesale is 1,95,000 and quarterly VAHAN retail is 1,83,000, over 91 days. Compute closing channel inventory in units and days, and state the first question you'd ask.

Solution. Build = 1,95,000 − 1,83,000 = +12,000, so closing stock = 66,000 units. Retail/day = 1,83,000 ÷ 91 = 2,011. Closing days = 66,000 ÷ 2,011 ≈ 32.8 days, up from an opening 54,000 ÷ 2,011 ≈ 26.9. One quarter's build to ~33 days is inside the 3 to 4-week comfort band. Your first question: is next quarter festive season? Pre-festive builds are normal. The same build in a demand lull, or a third build in a row, changes the reading.

P3 (discounts flow-through). ASP is ₹7,20,000 and EBITDA margin is 9.0%. Management raises average discounts by ₹14,400/unit to defend share, with volumes unchanged. What's the new EBITDA margin? And what volume growth would have been needed to keep EBITDA flat instead, at a contribution margin of 24%?

Solution. Discount = 14,400/7,20,000 = 2.0% of ASP, and all of it comes straight off EBITDA. So the new margin = 9.0 − 2.0 = 7.0%, a 22% cut in EBITDA per unit (2/9). To hold absolute EBITDA with margin down to 7.0% of the new realization, work it strictly. New realization = 7,05,600 and new per-unit EBITDA = 64,800 − 14,400 = ₹50,400 against ₹64,800. Volumes must rise by 64,800/50,400 − 1 = +28.6%. A 2% price giveaway needs ~29% more volume at these margins. No wonder experts treat rising discounts per unit as the single most expensive line in the panel.

P4 (content per vehicle). An ancillary's addressed industry volumes grow 3%. Its content per vehicle rises from ₹4,100 to ₹4,510, and its share of addressed platforms rises from 24% to 25%. Compute revenue growth multiplicatively and identify the growth's largest driver.

Solution. Content growth = 4,510/4,100 − 1 = 10.0%. Share factor = 25/24 = 1.0417. Growth = 1.03 × 1.10 × 1.0417 − 1 = 18.0%. The largest driver is content per vehicle, 10 points of the 18, which is the structural story. Volumes contributed only 3. An 18% grower in a 3%-volume industry is a content-and-share machine. Its durability question is the platform-win pipeline, and the auto cycle is the wrong thing to worry about.

P5 (RM lag). A supplier has baseline revenue of ₹200 cr/quarter. RM is 62% of revenue, and aluminium is 45% of RM. Pass-through is quarterly with a one-quarter lag. The aluminium index jumps from 100 to 115 in Q1 and stays there. Compute gross margin (%) for Q0 to Q2 and the absolute gross profit each quarter.

Solution. In Q0, RM = 124 (aluminium 55.8, other 68.2), GP = 76.0 and GM = 38.0%. In Q1 the cost goes up and the price lags. Aluminium = 55.8 × 1.15 = 64.17, RM = 132.37 and revenue stays at 200. GP = ₹67.63 cr and GM = 33.8%, down 4.2pp. In Q2 the price resets. Revenue = 200 + 55.8 × 0.15 = 208.37 and RM = 132.37, so GP = ₹76.0 cr, fully restored. But GM = 76.0/208.37 = 36.5%, still 1.5pp below baseline even though the rupee spread is whole. While the metal stays high the percentage never fully comes back, though the spread does. Judge the spread.

P6 (ancillary ROCE bridge). Supplier X has an EBIT margin of 11% and capital turns of 2.6×. Supplier Y has an EBIT margin of 15% and capital turns of 0.9×. Compute pre-tax ROCE for both against the ~12% hurdle. Then say which one passes the ROCE test, and what Y's likely disease is.

Solution. X: 11% × 2.6 = 28.6%, comfortably above the hurdle. Y: 15% × 0.9 = 13.5%, barely above it despite the fatter margin. X passes easily and Y is marginal. Y's low turns scream capital indiscipline: capacity built ahead of orders, aftermarket inventory bloat, or stretched OEM receivables. The margin line flatters Y, and the capital line convicts it. Always multiply.

P7 (book-to-bill & coverage). A cap-goods company has trailing-12m revenue of ₹9,200 cr, order inflow of ₹11,500 cr and an opening backlog of ₹21,000 cr, with no cancellations. Compute book-to-bill, closing backlog, and backlog coverage. Then state the expected revenue direction and the quality question you still have to ask.

Solution. Book-to-bill = 11,500/9,200 = 1.25×, so the book is growing. Closing backlog = 21,000 + 11,500 − 9,200 = ₹23,300 cr. Coverage = 23,300/9,200 = 2.53×, at the healthy edge of the EPC band (2.5 to 3.5×). Revenue should rise, since 1.25× sustained implies mid-teens growth capacity at a stable burn. The quality question is the margin embedded at bid and the counterparty mix. A 1.25× won by under-bidding government fixed-price tenders is a de-rating. Don't count it as growth. Check embedded-margin commentary and escalation clauses before you applaud.

P8 (burn rate & revenue forecast). An EPC company has an opening backlog of ₹18,600 cr, last year's burn rate was 28%, and short-cycle book-and-bill typically runs ₹900 cr/yr. Forecast next year's revenue. Then recompute it if two stuck projects, ₹3,100 cr of backlog, burn at zero.

Solution. Base case: 0.28 × 18,600 + 900 = 5,208 + 900 = ₹6,108 cr. With ₹3,100 cr frozen, effective backlog = 15,500, and revenue = 0.28 × 15,500 + 900 = 4,340 + 900 = ₹5,240 cr. That's a 14% forecast cut with zero change in reported backlog or book-to-bill. So burn rate and backlog-ageing disclosures matter more than the headline book. Zombie backlog inflates coverage while it starves revenue.

P9 (the EPC cash test). An EPC's statements show EBITDA of ₹640 cr, revenue growth of ₹1,800 cr and NWC running at 22% of revenue. Capex was ₹95 cr, interest ₹210 cr and tax paid ₹88 cr. Compute FCF after interest and the ΔNWC/EBITDA ratio, and give your verdict in one sentence.

Solution. ΔNWC = 0.22 × 1,800 = ₹396 cr. FCF = 640 − 396 − 95 − 88 − 210 = −₹149 cr. ΔNWC/EBITDA = 396/640 = 62%, so nearly two-thirds of EBITDA is consumed by growth's working capital before a rupee reaches capex or lenders. Verdict: this "profitable, growing" EPC is cash-flow negative and borrowing to grow, which is the Stambh spiral in progress. Check BG headroom and the unbilled-revenue trend immediately.

P10 (mid-cycle valuation: timed, 8 minutes). A CV OEM is at the cycle peak, with current-year EBITDA of ₹3,800 cr on a record 96,000 units. Its 10-year average volume is 72,000. Mid-cycle EBITDA margin is 9.5% against the current 12.8%, and current revenue is ₹29,700 cr. Net debt is ₹1,400 cr, market cap ₹33,600 cr, and the sector's mid-cycle EV/EBITDA is 9×. Estimate mid-cycle EBITDA and the stock's premium or discount to mid-cycle fair value. Trailing P/E is 13×. Is it cheap?

Solution. Mid-cycle revenue ≈ 29,700 × (72,000/96,000) = ₹22,275 cr. Mid-cycle EBITDA = 9.5% × 22,275 ≈ ₹2,116 cr. Fair EV = 9 × 2,116 = ₹19,045 cr, so fair equity = 19,045 − 1,400 = ₹17,645 cr against a market cap of ₹33,600 cr. The stock trades at a ~90% premium to mid-cycle value, with market EV/mid-cycle EBITDA ≈ 16.5×. The 13× trailing P/E is the trap in its purest form. Earnings are at a double peak, in volumes and margin, so the "cheap" multiple rests on EBITDA ~80% above mid-cycle. It isn't cheap. It's priced for the cycle never to turn.

P11 (ICE/EV classification: timed, 5 minutes). Classify each revenue line of "Deccan Autoparts" into ICE-at-risk / EV-agnostic / EV-upside, and compute the at-risk share. The lines are exhaust systems 28%, seating 22%, wiring harness 18%, alloy wheels 12%, fuel injection 11% and battery thermal management 9%. Give a one-sentence verdict.

Solution. ICE-at-risk: exhaust 28 + fuel injection 11 = 39%. EV-agnostic: seating 22 + harness 18 + alloys 12 = 52%. You could argue the harness is upside, so accept either classification for it. EV-upside: thermal 9%, plus the harness if you classed it that way. Verdict: a transition-manageable book. 39% at risk is real, but the majority is agnostic-to-upside. The thesis hinges on what happens to the exhaust/fuel-injection cash flows, a ~15 to 20-year Indian run-off given slow CV/2W electrification. Either they get redeployed into thermal/harness capacity at ROCE, or they get burned defending the melting pool.

P12 (integration: timed, 10 minutes). Skim-read this quarterly disclosure set and list every red flag, naming the test that catches each one. OEM "Vistara Motors". Wholesale +11% YoY, FADA retail +2%. Channel inventory "in line with festive preparation" (third consecutive quarter of this phrase). Discounts undisclosed, but receivables +34% YoY and dealer-inventory funding book at captive NBFC +41%. EBITDA margin 13.9% (band: PV 8–13%) at 91% utilization. EV mix "doubled" (from 1.8% to 3.6%), EV contribution margin undisclosed. ₹6,200 cr greenfield announced.

Solution. There are six flags here.

1. A wholesale-retail gap of 9pp, which raises stuffing suspicion straight from the channel detector.
2. "Festive preparation" three quarters running. A seasonal excuse can't recur like that, which is the festive-season caveat that comes with the detector.
3. Receivables and dealer funding growing at ~4× the rate of wholesale. That's the stuffing's financing shadow, OEM red flag 6.
4. A margin above the PV band at 91% utilization, a cyclical peak reported as normal. That's OEM red flag 5.
5. EV mix trumpeted off a trivial base with no unit economics. That's the "EV cash burn hidden" flag from the EV economics, OEM red flag 4.
6. A greenfield at peak utilization, the capital-cycle top signal borrowed from industrials.

Six flags blow well past the multi-signal standard, and the memo writes itself: reported growth is warehoused, margin is a peak artifact, and capex is being committed at the top.

P13 (guided: capacity utilization & operating leverage). A products capital-goods company has nameplate capacity of 8,000 units. Output rises from 5,200 units in Year A to 6,600 in Year B. ASP is ₹42 lakh/unit and unchanged, contribution margin is 32%, and fixed costs are ₹520 cr/year, flat. Compute utilization and EBITDA margin in each year, and the operating-leverage multiplier. Then say in one line what it implies for extrapolating Year B's margin.

Solution. Utilization is 5,200 ÷ 8,000 = 65.0% in A and 6,600 ÷ 8,000 = 82.5% in B. Revenue is 5,200 × ₹0.42 cr = ₹2,184 cr in A and 6,600 × ₹0.42 cr = ₹2,772 cr in B. EBITDA = 0.32 × revenue − 520. Year A = 698.88 − 520 = ₹178.9 cr, an 8.2% margin. Year B = 887.04 − 520 = ₹367.0 cr, a 13.2% margin. Revenue rose 26.9% and EBITDA 105.2%, so the multiplier ≈ 3.9×. Year B's 13.2% is a high-utilization print sitting on a ~3.9× leverage spring, so normalize to mid-cycle utilization before you capitalize it. The same volume swing that lifted margin ~5 points will take those points away just as fast.

P14 (guided: volume × ASP decomposition). A PV OEM's revenue rose from ₹28,000 cr to ₹32,200 cr, while units sold rose from 5.60 lakh to 5.94 lakh. Decompose the revenue growth into volume and ASP, confirm it's multiplicative, and state the one follow-up question.

Solution. Volume growth = 5.94 ÷ 5.60 − 1 = +6.1%. Revenue growth = 32,200 ÷ 28,000 − 1 = +15.0%. ASP growth = (1.150 ÷ 1.061) − 1 = +8.4%, taking ASP from ₹5.00 lakh to ₹5.42 lakh. Check: 1.061 × 1.084 = 1.150. It's multiplicative, since the additive version gives 6.1 + 8.4 = 14.5 ≠ 15.0. Growth is more ASP-led (8.4 pts) than volume-led (6.1 pts). So the follow-up is the ASP decomposition from the OEM panel: which of three things drives the 8.4% ASP gain? Mix is quality growth, because buyers are choosing SUVs and richer variants. Pricing lasts only if volumes hold. RM or regulatory pass-through is neutral, because it merely recovers cost. "15% revenue growth" means nothing until you know that split.

P15 (independent: book-to-bill). A US capital-goods company has an opening backlog of $5,200m, order inflow of $4,000m and revenue of $4,600m for the year. Compute book-to-bill, closing backlog, and coverage. Then state the reading.

Solution. Book-to-bill = 4,000 ÷ 4,600 = 0.87×, below 1. Closing backlog = 5,200 + 4,000 − 4,600 = $4,600m, down from $5,200m. Coverage = 4,600 ÷ 4,600 = 1.00×. Inflow ran below execution, so the book that feeds future revenue shrank. A sub-1× book-to-bill is a pre-announced growth slowdown that the P&L hasn't printed yet. For a short-cycle product business, 1.0× coverage isn't alarming in itself. The direction is the flag, with book-to-bill at 0.87× and backlog falling. One soft year is noise, and four in a row is a trend.

P16 (independent: the EPC working-capital spiral). An EPC contractor grows revenue from ₹5,000 cr to ₹6,500 cr at an EBITDA margin of 11%. NWC intensity creeps from 24% to 27% of revenue. Capex is 1.5% of revenue, interest is ₹150 cr and tax is ₹120 cr. Compute EBITDA, ΔNWC, FCF after interest, and ΔNWC/EBITDA, then give the one-line verdict.

Solution. EBITDA = 11% × 6,500 = ₹715 cr. NWC goes from 24% × 5,000 = ₹1,200 cr to 27% × 6,500 = ₹1,755 cr, so ΔNWC = ₹555 cr of cash out. Capex = 1.5% × 6,500 = ₹97.5 cr. FCF after interest = 715 − 555 − 97.5 − 120 − 150 = −₹207.5 cr. ΔNWC/EBITDA = 555 ÷ 715 = 77.6%, so over three-quarters of EBITDA is locked into working capital before capex or lenders get paid. Verdict: this is a profitable, fast-growing EPC that's cash-flow negative and has to borrow to grow, the Stambh spiral again. Check BG headroom and the unbilled-revenue trend now.

P17 (independent: EV content-mix transition). An ancillary's ICE content per vehicle is ₹6,000, and its EV content is ₹15,000. The EV mix in its addressed segment rises from 12% to 30% while industry volumes and its share stay flat. Compute blended content each year, the content-driven revenue growth, and classify the supplier.

Solution. EV kit value = 15,000 ÷ 6,000 = 2.5× ICE. Blended content in Year 1 = 0.88 × 6,000 + 0.12 × 15,000 = 5,280 + 1,800 = ₹7,080. In Year 2 it's 0.70 × 6,000 + 0.30 × 15,000 = 4,200 + 4,500 = ₹8,700. With volumes and share flat, revenue growth = 8,700 ÷ 7,080 − 1 = +22.9%. Classification: EV-upside. Content rises as the segment electrifies, so the transition is this supplier's tailwind. It's no threat to terminal value. The reflexive "EVs kill component makers" thesis is backwards for a portfolio positioned in high-voltage/thermal/electrical content.

P18 (stretch: timed, 10 minutes, mid-cycle valuation). A US machinery maker is at the top of its cycle. Current-year revenue is $58,000m at a peak EBITDA margin of 19%, against 10-year average revenue of $46,000m and a mid-cycle EBITDA margin of 14.5%. Net debt is $5,000m, market cap is $132,000m, and the sector's mid-cycle EV/EBITDA is 10.5×. Compute peak EBITDA, mid-cycle EBITDA, mid-cycle fair equity value, and the stock's premium or discount to it. Trailing EV/EBITDA is 12.4×. Is it cheap?

Solution. Peak EBITDA = 19% × 58,000 = $11,020m. Mid-cycle EBITDA = 14.5% × 46,000 = $6,670m. Fair EV = 10.5 × 6,670 = $70,035m, so fair equity = 70,035 − 5,000 = $65,035m against a market cap of $132,000m. That's a ~103% premium to mid-cycle value, and market EV of $137,000m ÷ mid-cycle EBITDA = 20.5×. The 12.4× trailing EV/EBITDA (137,000 ÷ 11,020) is the trap in dollars. It rests on doubly-peak earnings, in volume and margin, so a multiple that screens "reasonable" is really 20.5× mid-cycle. It isn't cheap. It's priced for the peak to persist, which is the trailing-multiple trap all over again.

Applied mini-project: the twin workup (the module gate)

Task. Build both playbooks on real companies, ~4.5 hours total.

Part A: Autos (2.5h). Pick one listed Indian OEM (Maruti Suzuki, Tata Motors, M&M, Eicher, Hero, Bajaj, TVS, or Ashok Leyland). Pull 5 years of annual reports plus the last 4 quarterly investor presentations from the company website or BSE. Add monthly wholesale from SIAM or company releases, and monthly retail from FADA's press releases (fada.in, free), with the VAHAN dashboard for raw registrations. Then build four things:

1. The OEM KPI panel for 8 quarters, wholesale against retail, with a computed channel-days trend.
2. ASP and margin against the segment band, with a commentary on mix.
3. The EV disclosure check, using the disclosure test from the EV economics.
4. A one-page verdict covering where the OEM is in the cycle, whether reported demand is real, and mid-cycle EV/EBITDA against price.

Part B: Industrials (2h). Pick one order-book company (L&T, Siemens India, ABB India, BHEL, Thermax, KEC, or Cummins India), or Caterpillar via the 10-K if you'd rather work on a US name. Pull 5 years of annual reports plus the last 4 quarters' order-inflow disclosures. Then build four things:

1. The order-book panel: inflow, book-to-bill, backlog coverage, burn rate. Add NWC days from the balance sheet, with unbilled revenue and advances included.
2. The cash test, cumulative 5-year CFO/EBITDA and ΔNWC/EBITDA.
3. A product-vs-project classification, applying the contrast between the two poles.
4. One page answering two questions. Is the order book growing and collectible? And what does ROCE against the 12% hurdle say about whether the growth is worth anything?

Scoring rubric. Be strict with yourself, because you need 12/16 to pass the gate.

Reading & resources

• FADA monthly retail data + VAHAN dashboard (fada.in; vahan.parivahan.gov.in). [Free] [Beginner] The retail truth serum. Read three consecutive months alongside SIAM or company wholesale releases and you have the channel-stuffing detector running live.
• SIAM statistics & company monthly releases (siam.in; BSE announcements). [Free] [Beginner] The wholesale line. The gap between this and FADA is the week's signature trade.
• Company annual reports & investor decks. Read Maruti Suzuki and Eicher Motors, Motherson for the 3CX10 discussion, and Uno Minda for its kit-value disclosures. Read L&T for the order book and WC discussion in its MD&A, and Siemens/ABB India for the advances on the balance sheet. Then read the Caterpillar 10-K's segment note, Machinery vs Cat Financial, where you practice the captive-finance split. [Free] [Intermediate].
• Aswath Damodaran, "The Dark Side of Valuation" (cyclical-company chapters) and his blog posts on valuing cyclical & commodity companies. [Paid book / Free posts] [Intermediate] The mid-cycle normalization method P10 practiced.
• Marathon Asset Management / Edward Chancellor, "Capital Returns" (Introduction + capital-cycle essays). [Paid] [Intermediate] The capital-cycle rule of watching supply instead of demand, written up by its best practitioners.
• Womack, Jones & Roos, "The Machine That Changed the World". [Paid] [Beginner] Why Toyota's production system is a cost moat. It's the book that explains the Toyota side of the Tesla-versus-Toyota contrast.
• RBI OBICUS capacity-utilization survey (rbi.org.in, quarterly). [Free] [Intermediate] The utilization data behind the >80% pricing-power threshold, for India manufacturing broadly.
• ACT Research / FTR Class 8 order releases (summaries free in trade press: transportdive.com, freightwaves.com). [Free summaries] [Intermediate] The US CV cycle's leading indicator. Pair it with FRED series TOTALSA (US light-vehicle SAAR). [Free] [Beginner].
• CRISIL / ICRA / CareEdge sector notes on autos, auto components, and construction (free summaries on their sites). [Free] [Intermediate] Where to verify financing-penetration, channel-inventory norms, and WC benchmarks each year.
• Zerodha Varsity, sector-analysis chapters (auto & capital goods). [Free] [Beginner] A gentle India-first second pass over these panels.
• BNEF annual battery price survey (headline number reported widely in free press). [Free] [Beginner] Your yearly $/kWh update for the EV per-vehicle economics. Verify every EV unit-economics claim against it.
• Going deeper: BorgWarner "Charging Forward" strategy disclosures + Aptiv investor-day decks (company investor-relations sites). [Free] [Intermediate] The ancillary portfolio X-ray, managed as live capital allocation. Two components majors reclassify their own revenue pools and move capital from ICE-at-risk toward EV-upside, disclosed quarter by quarter. Read them as running cases.
• Going deeper: Aswath Damodaran, "Narrative and Numbers: The Value of Stories in Business" (the disruption/auto chapters). [Paid] [Intermediate] How to bound an EV-transition or cyclical-recovery story with numbers before it becomes a multiple. It's the thinking behind pricing the EV option instead of the press release, and behind the mid-cycle normalization in the valuation lens.

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": "Running the wholesale-vs-retail channel-stuffing detector (§4.2.5) and the EPC working-capital-spiral test (§4.4.4) across a whole coverage list instead of one company at a time",
  "use": "An LLM turns monthly SIAM/FADA releases or quarterly EPC working-capital disclosures into a clean panel and drafts a first flag for any name whose channel-inventory days or NWC intensity has risen for several consecutive periods, in minutes instead of an afternoon of cross-referencing PDFs.",
  "tools": ["Claude/ChatGPT/Gemini to extract wholesale/retail volumes or EPC working-capital line items from pasted releases into a table", "a notebook (Python/pandas) to hold the panel and compute channel-inventory days, NWC intensity, and CFO/EBITDA yourself", "the VAHAN dashboard's exportable registration data", "company investor-presentation PDFs for backlog and segment disclosures"],
  "prompt": "Paste this company's last six quarters of wholesale (SIAM/company release) and retail (FADA/VAHAN) volumes, or an EPC company's revenue, backlog, receivables, unbilled revenue, and retention money. Build the channel-inventory-days trend or the NWC-intensity trend and flag any period where it has worsened for three or more consecutive quarters, alongside the corroborating signal (discounts/unit, or CFO/EBITDA).",
  "verify": "Recompute at least one quarter's channel-inventory days or one year's ΔNWC from the raw disclosed numbers yourself (§4.2.5's and §4.4.4's formulas) before trusting a flag — LLMs conflate wholesale with retail when a release does not label the figure clearly, and will happily count an unawarded L1 position as backlog unless told not to.",
  "diy": "The mastery gate is unaided: build the wholesale-retail reconciliation or the EPC NWC-spiral test from raw disclosures by hand, and state whether growth is real or borrowed — no tool in the room. The tool above saves the transcription hour; it does not replace knowing why a rising order book can still be a sell."
}

Modern Data Analysis

By hand first. The Bharat Wheels channel-inventory table and the Stambh Infra working-capital spiral are both built by hand a quarter or a year at a time.

Today's workflow. An analyst covering a dozen OEMs, ancillaries, and EPC names doesn't rebuild these tables from scratch every reporting season. In practice you pull SIAM wholesale plus VAHAN/FADA retail, or an EPC company's quarterly working-capital disclosures, into one panel. Then you let code carry the channel-inventory-days trend and the NWC-intensity trend forward for every name at once. The code flags this week's thresholds for you: inventory days above 6 to 8 weeks for three-plus quarters, or NWC intensity rising while cumulative CFO/EBITDA falls below 40 to 50%. That beats eyeballing one company's numbers by hand each month.

Tools & sources (IN + US). For India, start with these:

• The VAHAN dashboard (vahan.parivahan.gov.in, Ministry of Road Transport & Highways), with near-real-time registrations you can export.
• FADA's monthly retail sales data (fada.in).
• SIAM's monthly wholesale statistics (siam.in).
• The RBI's OBICUS capacity-utilization survey (rbi.org.in). It's quarterly and machine-readable, and it's the source behind the 80% pricing-power line.
• L&T's and Siemens/ABB India's segment notes in their investor presentations give you backlog and coverage.

For the US:

• The Census Bureau's Manufacturers' Shipments, Inventories & Orders (M3) survey (census.gov/manufacturing/m3) for US industrial order-book and backlog data, the direct analogue of SIAM's and L&T's disclosures.
• FRED series TOTALSA for the light-vehicle SAAR.
• ACT Research/FTR Class 8 net-order releases (trade-press summaries) for the US truck-cycle leading indicator.
• SEC EDGAR's XBRL company-facts API for Caterpillar's and Ford's historical revenue, backlog, and capex, straight from the filings.

# recomputed in-session -- R3
# verification anchor -- the Sec 4.4.4 Stambh Infra EPC working-capital spiral
revs = [4000, 5000, 6250, 7812.5]
nwc_pct = [0.25, 0.27, 0.29, 0.31]
nwc = [r * p for r, p in zip(revs, nwc_pct)]
dNWC = [nwc[i] - nwc[i - 1] for i in range(1, 4)]
ebitda = [0.10 * r for r in revs]
tax = [65, 89, 116, 150]                 # as disclosed (25% of PBT; PBT itself not separately given)
capex = [0, 75, 94, 117]
net_debt = [800]
for i in range(1, 4):
    interest = 0.10 * net_debt[i - 1]
    cfo = ebitda[i] - dNWC[i - 1] - tax[i]
    fcf = cfo - interest - capex[i]
    net_debt.append(net_debt[i - 1] - fcf)
    print(f"Yr{i}: CFO/EBITDA {cfo/ebitda[i]:.1%}  FCF-after-interest Rs{fcf:.0f}cr  "
          f"closing net debt Rs{net_debt[i]:.0f}cr")
print(f"net debt +{ (net_debt[3]/net_debt[0]-1)*100 :.0f}% over 3 record-profit years")

# NEW -- the same channel-inventory-days flag, run across several names at once
import pandas as pd
panel = pd.DataFrame({
    "name":      ["Alpha Motors", "Beta Auto", "Gamma Wheels"],   # illustrative, not real companies
    "wholesale": [240000, 198000, 205000],
    "retail":    [206000, 201000, 204000],
    "days":      [64, 33, 31],           # this quarter's channel-inventory days, from the by-hand formula
    "days_3q_ago": [30, 34, 29],
})
panel["stuffing_flag"] = (panel.days > 42) & (panel.days > panel.days_3q_ago)
print(panel[["name", "wholesale", "retail", "days", "stuffing_flag"]])

Verify. The Stambh Infra reproduction matches the table above to its own rounding. CFO/EBITDA comes out at ~12%, ~7%, ~3% across the three years. FCF after interest is roughly −₹94, −₹137, −₹199 cr, and closing net debt is roughly ₹894, ₹1,031, ₹1,230 cr. Net debt rises ~54% while PAT compounds at ~32%/year. The Bharat Wheels channel-inventory-days table reproduces too, at 32, 37, 48, and 64 days across the four quarters. So does the discount-sensitivity rule, where a ₹10,000/unit discount at a ₹6,00,000 ASP costs ~1.7 margin points. The multi-name panel above is illustrative data built to show the mechanism. Alpha Motors copies the Bharat Wheels pattern, and none of it is a claim about any real company.

Quantitative lens

Refuse wholesale as the demand signal. Use retail, reconciled. Bharat Wheels printed accelerating wholesale growth all year while retail grew under 1% and channel inventory doubled from 32 to 64 days. Unwinding that stock cuts wholesale by roughly a quarter, and on the contribution-margin math above EBITDA falls at nearly twice that rate. Operating leverage turns a volume story into a much larger profit story, and you can't see any of it in the wholesale line the press release leads with.

Refuse P&L growth as proof an EPC company is creating value. Use CFO/EBITDA and the NWC-intensity trend. Stambh Infra's PAT compounds at ~32% a year for three straight "record" years, while cash conversion collapses from ~12% to ~3% of EBITDA and net debt rises 54%. A company can be profitable, growing, and financing its own eventual default all at once, and the income statement never says so. In the real graveyard (IL&FS's construction arm, IVRCL, HCC, Punj Lloyd), the kill shot always came through the bank-guarantee channel. It never arrived as a profit warning.

Refuse the headline order book. Use burn-adjusted, L1-excluded backlog coverage. Unawarded L1 positions, zombie orders that neither burn nor cancel, and low-margin turnover work booked to keep crews busy all inflate coverage without adding value. The same >4 to 5× coverage that looks reassuring on a screener should raise a question before anyone boasts about it. Divide by the execution (burn) rate as well as trailing revenue before you believe a coverage ratio.

Honest limits. The Bharat Wheels and Stambh Infra tables are synthetic constructions calibrated to the bands stated above. They demonstrate the mechanism, how operating leverage or a working-capital spiral compounds. They don't forecast any named company, whose real contribution margin, fixed-cost base, and NWC-intensity path must come from its own disclosures. The stuffing-flag function above is a first-pass screen. It can't give you a verdict. A flagged name still needs the detector's corroborating checks before anyone acts on it: rising discounts per unit, dealer-financing growth. And three quarters of data is a thin base for any threshold rule.

Do it in code: hold the panel in G2/DA1.02, chart the channel-inventory-days or NWC-intensity trend with G2/DA1.04, and treat Class 8 net orders or e-way-bill volumes as a leading-indicator time series with G1/QM1.03 instead of reading one release at a time.

Where this goes next: galaxy cross-links

• G2 · DA1.02 Pandas data wrangling builds the multi-name panel above properly, joining SIAM/VAHAN/FADA exports or quarterly EPC disclosures into one running series.
• G2 · DA1.04 EDA & visualization: charting channel-inventory days or NWC intensity against revenue growth is how these red flags become visible a quarter or two before the destock or the covenant breach.
• G1 · QM1.03 Time series, Monte Carlo & ML: Class 8 net orders and e-way-bill/FASTag volumes are the kind of leading-indicator series built here. Treat them as a time series with a lead-lag structure instead of a single data point.
• G2 · CS2.05 APIs, HTTP & JSON: the mechanics of pulling VAHAN's registration data or the Census Bureau's M3 survey programmatically instead of reading dashboards by hand.
• G3 · CN4.02 Industry primers & disruption: the EV-transition portfolio X-ray and the Tesla-versus-Toyota contrast are live instances of the disruption-analysis pattern taught there.
• X1 · AI0.07 The AI-augmented analyst: the general version of the AI-Augment box above, covering what to delegate, what to verify, and why the mastery gate stays unaided.

Teach it back & journal

Feynman prompt. Your smart 15-year-old cousin asks: "Dad's friend owns a truck company share because 'profits doubled.' Dad owns shares of a company that builds metro lines because 'they have orders for five years.' Are those good reasons?" Write one page. First explain why a truck maker's doubled profit can be the worst time to buy, covering deferrable demand, the cycle, and why the P/E looks cheapest at the top. Then explain how a car company can "sell" cars that nobody has bought yet: the dealer garage trick, and how the government's own registration website catches it. Finally, explain why a company with five years of orders can still run out of money. Who pays for the cement and workers while the metro gets built, and what happens when the bank stops vouching for you? Use shop-counter language. Arithmetic is allowed, and jargon isn't.

Journal reflection. Find the last vehicle purchase in your family, any vehicle from any year, and treat it as data for this week's ideas. Was it a replacement or a first-time purchase? Note whether it was financed or paid in cash, and whether a 2-point EMI rate change would have delayed it. Then note whether the dealer offered a discount, and what that implies about that quarter's channel. Now the industrial half. Recall any infrastructure project near you that visibly stalled (a flyover, a building). Write three lines on who probably wasn't being paid while it stood still, and which line of the contractor's balance sheet was swelling. File both under "cycles observed first-hand." Phase 8's scuttlebutt work will ask you to do this deliberately.

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