Commodities, Real Estate & REITs

Phase 5 · Week 50 · ~20 hours

Where it sits. A bank sets its lending rates. An FMCG brand raises prices yearly. A SaaS firm reprices at renewal. Now meet the businesses that mostly take the price the world hands them: steelmakers, aluminium smelters, miners, oil producers, refiners. Real estate takes something subtler from the world, the credit cycle and the interest rate, and it fills the second half of the week. Both playbooks share one deep lesson: when the top line is set by a cycle you don't control, ordinary valuation tools invert. The trailing P/E, a serviceable rough guide everywhere else, becomes an active saboteur that screams "buy" at the top and "sell" at the bottom. Developers' P&L revenue becomes a four-year-old newspaper. REITs bring the fix: an asset class deliberately engineered to convert lumpy real estate into bond-like cash flow, with its own KPI stack (NOI, FFO/AFFO, WALE, LTV) that you'll compute from scratch. The capital cycle you met earlier as theory lives here as daily weather. This is the ninth of Phase 5's ten sector playbooks, and the one where the enemy changes. On the master map's five-questions spine, it is question 2 asked honestly (is it a good business? usually not. So which one survives?) and question 5 with the sign flipped (what is the market implying, and is the cheap-looking multiple a trap?).

Suggested hour budget. Core content: 8h · worked examples: 1.5h · practice set: 3h · applied mini-project (the cyclical & yield workup): 5h · flashcards + mastery check: 1.5h · teach-back + journal: 1h.

App link. Three drills in the sector case engine run on this material. The cycle-clock drill plays a full price cycle out quarter by quarter, and you call where trailing P/E misleads and where mid-cycle value is. In the cost-curve placement game you drop producers onto an industry cost curve, then crash the price and see who bleeds. The FFO/AFFO builder has you assemble the walk from net income to distributable cash from shuffled adjustments. Your completed KPI tables seed the commodities and real-estate pages of the sector KPI cheat-sheet artifact. Passing the mastery check and the graded workup ("Cyclical valuation exam") unlocks the phase capstone.

Learning objectives

By the end you can:

1. Explain price-taker economics from first principles: why a commodity producer's profit is (a price it doesn't control) − (a cost it does), why that makes the cost curve the industry's only durable org chart, and why a 1st-quartile cost position is the sole moat available in the sector.
2. Demonstrate the cyclical trap with numbers. Walk a full price cycle and show why the trailing P/E is lowest at the peak (a sell signal) and highest or meaningless at the trough, and why you never value a cyclical on trailing earnings.
3. Compute and interpret the commodity KPI panel: realization vs benchmark, cash cost per unit and cost-curve quartile, spreads (refining GRM, steel spread), EBITDA/tonne, net debt/EBITDA (<2× healthy), capacity utilization, reserve life and reserve replacement ratio (>100%), and strip inventory gains/losses out of reported earnings.
4. Run a mid-cycle valuation end-to-end: normalize EBITDA across a cycle, apply a mid-cycle multiple, subtract current net debt, cross-check against replacement cost per tonne of capacity, and place today's price inside that range using the capital-cycle clock from M4.06.
5. Explain why a developer's P&L revenue is meaningless (recognition lags sales by years under Ind AS 115's point-in-time model) and run the real panel instead: pre-sales/bookings, collections vs bookings, sales velocity, realization per sqft, launches and pipeline, land bank, unsold and aging inventory, net debt/equity (post-RERA <0.5×), and the OCF self-funding test.
6. Build a developer NAV from project cash flows plus land bank minus net debt, explain RERA's transformation of the sector's business model, and contrast the Indian developer with the US homebuilder (P/B lens, land options, the 2008 lesson).
7. Compute FFO and AFFO from scratch, including the maintenance-capex, straight-line rent, and TI/LC adjustments, then judge a REIT on NOI and same-store NOI, occupancy (>93%), WALE, re-leasing spreads, AFFO payout (<80% healthy, >90% cut risk), LTV (SEBI cap 49%), NAV and implied cap rate, and distribution yield versus the government bond.
8. Recite and apply both red-flag panels: the five commodity flags (rising cash cost, debt-funded peak expansion, net debt climbing into a downturn, inventory-flattered earnings, low trailing P/E at the peak) and the five real-estate flags (pre-sales miss, collections lagging bookings, aging unsold inventory, AFFO payout >90%, occupancy slide with negative re-leasing spreads).

Prerequisites & connections

Builds on. M1, three times over. Its inventory accounting matters because FIFO cost layers are where commodity inventory gains hide. Its revenue recognition under Ind AS 115 matters because the "point in time" branch of the five-step model is what makes developer P&Ls lag reality. And its depreciation is the line FFO exists to reverse. M2.02–M2.04 gave you ratios, net debt/EBITDA and cash conversion, and here they become survival metrics. M2.05–M2.07 gave you forensics, and channel stuffing has a real-estate cousin called "bookings without collections". M3.04–M3.05 supply the DCF mechanics, since developer NAV is a portfolio of project DCFs and cap-rate valuation is a perpetuity in disguise. M3.06 covers relative valuation, EV/EBITDA and why the multiple must be paired with normalized earnings. M3.08 (special situations) is where the "normalize to mid-cycle" instruction was issued, and here it gets carried out in full. M4.03 brings Porter, and commodity producers are the textbook case of zero buyer/supplier differentiation. M4.06 is the capital cycle, and this is its home game.

Feeds into. M5.10, the synthesis capstone, mixes cyclical cases into the right-lens exam. Airlines and hotels are cyclicals wearing service uniforms, since RevPAR is a realization and CASK is a cash cost. In Phase 7, commodities become macro transmission made visible (China credit impulses, rate cycles, USD), and M7 supplies the upstream causes of these price charts. In Phase 8, a 1 to 2 hour teardown of any metals, energy or realty name runs on these panels. Phase 9 covers position sizing for cyclicals, where high-uncertainty, high-asymmetry bets demand different sizing than compounders.

Core teaching content

4.1 Price-takers: profit = a price you don't set minus a cost you do

Start with what a commodity is: a product whose buyer doesn't care who made it. One tonne of prime hot-rolled coil, one tonne of LME-grade aluminium, one barrel of Brent-spec crude is interchangeable with any other. The buyer's only question is price. That single fact deletes most of the strategy toolkit you built in Phase 4. No brand, no switching cost, no network effect is available here, by construction. Porter's five forces collapse to two. Rivalry is total, since the products are identical, and the second force is entry and exit, which is the capital cycle.

So write the profit equation of every producer in this sector:

Profit per unit = Price (set by the global marginal producer) − Cash cost (set by your assets)

The company controls one of those two terms, and only one. Everything else in the commodities playbook follows from that:

• The company gives you three things to check: where it stands on the industry cost curve, whether its balance sheet can survive the trough, and whether it expands at the top or at the bottom.
• The price is a question about the cycle, and you approach it with humility. You won't forecast the steel price. Instead you ask where the current price stands against the cost curve and against history, and what the capital cycle says supply is about to do.
• Volume is the third lever. A producer is a volume × spread machine, so capacity growth and utilization, along with reserve life, decide how much of any spread it can capture.

One nuance before the tools. A few businesses in this sector are only partial price-takers. Coal India sells ~80% of volumes at government-notified prices under fuel-supply agreements, and only the e-auction slice is market-linked. That slice has historically fetched premiums of 30 to 100%+ over notified prices. So Coal India is an administered-price monopoly, a long way from a pure cyclical. Novelis (Hindalco's US subsidiary) is a converter. It buys aluminium at LME, sells rolled products at LME plus a conversion premium, and earns a per-tonne conversion margin that's far more stable than the metal price. Refiners don't take a price at all. They take a spread: product prices minus crude. So classify the exposure before you reach for the tools. Is the company a price-taker or a spread-taker? Is it a conversion-fee-taker, or is its price administered? Cyclicality declines in roughly that order, and the right valuation shifts with it.

4.4 The commodity KPI panel

Here's the panel first, then each dial with numbers. Treat the thresholds as analysts' rules of thumb, and calibrate them to the sub-sector and to where the cycle is.

The first dial, realization vs benchmark, compares the company's achieved price per unit with the public benchmark. That's LME for aluminium, a 62%-Fe index for iron ore, the domestic HRC index for steel and Brent or a basket for crude. Compute it every quarter as realization ÷ benchmark average for the period. Gaps are normal and explainable. Value-added rolled products earn LME plus $150 to 400/t in product premia, and high-silica ore takes a grade discount. Freight moves it, and so do contract-price lags and hedges. What you're hunting is unexplained drift. A producer persistently realizing 88% of benchmark, and falling, is losing mix, losing grade or over-hedged at bad levels. In India, ONGC's realization tracks Brent less quality discounts, but the state can step in between benchmark and shareholder. The July 2022 windfall levy on crude did that, and so do fuel-price freezes on OMC marketing margins in election seasons. Verify current status before you use either. For Coal India, realization splits into notified-price FSA volumes and e-auction volumes, and the e-auction premium is your cycle thermometer.

The second dial is cash cost per unit and its quartile, and you've met the cost curve already. Add one habit: track the company's cash cost across 5 years in USD and in local currency, against 2 to 3 peers. If the rupee-cost line rises while the USD line stays flat, that's currency. Operations haven't got worse.

Spreads get worked twice, because two industries live on them.

Refining GRM. A refinery buys crude and sells a basket of products. The gross refining margin is the basket value minus the landed crude cost, per barrel. Sindhu Refining (synthetic) lands its crude at $82/bbl, which is Brent at $80 plus $2 freight. Here are its product yields and prices per barrel of crude processed: diesel 42% × $102 + petrol 24% × $95 + jet/ATF 8% × $100 + LPG 4% × $62 + naphtha 8% × $76 + fuel oil & others 8% × $60. The other 6% is consumed as fuel and loss, and earns nothing:

Basket = 42.84 + 22.80 + 8.00 + 2.48 + 6.08 + 4.80 = $87.00 → GRM = 87.00 − 82.00 = $5.00/bbl

Refining EBITDA ≈ (GRM − cash opex, say $1.8/bbl) × throughput. On 20 mmtpa (× 7.33 bbl/t ≈ 146.6 mn bbl), that's $3.2 × 146.6 mn ≈ $469 mn, or ≈ ₹3,940 cr at ₹84/$. For a benchmark, the Singapore complex GRM is the Asian reference. Complex, high-Nelson-Index refineries earn a durable premium to Singapore, because they can crack cheap heavy-sour crude into light products. Think of a Reliance Jamnagar-style site, with Nelson complexity in the low-to-mid teens against 5 to 6 for simple plants. GRMs cycle hard. They drop to low single digits in gluts and reach $15 to 20+ in product squeezes, as in 2022. So value refiners on mid-cycle GRM, using the same method as the steel valuation below.

Steel spread. Blast-furnace steelmaking consumes roughly 1.6 t of iron ore and ~0.7 t of coking coal per tonne of crude steel. Here are illustrative Indian numbers. HRC sells at ₹58,000/t. Iron ore costs ₹5,500/t × 1.6 = ₹8,800. Coking coal is imported at ~$240/t, so it costs ₹20,000/t × 0.7 = ₹14,000.

Steel spread = 58,000 − 8,800 − 14,000 = ₹35,200/t

Subtract conversion costs of ₹15,000 to 16,000/t for an efficient integrated player (power, labor, refractories, consumables, other raw materials) and corporate overhead of ≈ ₹1,500/t. That leaves EBITDA/tonne ≈ ₹18,000. Now you can see why captive iron ore matters. A Tata Steel-style operation with ~100% captive ore swaps the ₹8,800 market cost for a mining cost of ~₹2,500 to 3,000. That's a structural spread advantage of ~₹6,000/t, and it holds in every price environment. Coking coal is mostly imported, and it's Indian steel's shared weak spot. So watch the spread as well as the steel price, because HRC can rise while the spread falls if coal rises faster.

EBITDA/tonne is the sector's summary statistic: spread minus conversion in one number, and it's comparable across companies and time. Indian flat-steel majors have swung roughly from ₹6,000 to 9,000/t in trough years (FY16, FY20) to ₹25,000 to 30,000/t at the FY22 peak. Those are illustrative, modeled on reported ranges, so pull the actuals from investor presentations, which disclose the figure directly. Novelis-style converters report the stable cousin, conversion EBITDA/tonne, at ~$400 to 525/t through recent cycles. That narrow range is the proof that converters aren't price-takers.

Net debt/EBITDA <2×, and why leverage kills cyclicals. Look back at the Deccan Steel table. Debt is identical throughout, yet it's 0.9× at the peak and 4.0× at the trough. Now add the mechanics that make that fatal.

1. EBITDA can halve twice, and a 75% fall from peak to trough is ordinary here. So a ratio computed on peak EBITDA understates true leverage by a factor of 3 to 4. Always compute the ratio on mid-cycle EBITDA, and demand <2×.
2. Debt service is fixed while cash flow is a sine wave. Interest cover of 13× at the peak became 3.0× at Deccan's trough, and one more bad year takes the dividend, then the capex, then the covenants.
3. Refinancing windows shut just when you need them. Credit markets price steel bonds off current spreads, so the FY2 refinancing comes at punitive rates or not at all.
4. The equity becomes an option. At 4× trough leverage, small moves in the commodity price shift huge value between lenders and shareholders. That's how levered cyclicals triple off the bottom, and it's also how they go to zero.

History has run this experiment at scale, twice, in one country, in one year: Tata Steel's ~$13bn debt-funded Corus acquisition (2007) and Hindalco's ~$6bn Novelis acquisition (2007). Both deals were done near the cycle top, and a decade of deleveraging pain followed each. Corus became a chronic impairment machine, and Hindalco survived 2008-09 by inches. Novelis itself proved a good asset. The error was the timing and leverage, and that distinction is worth writing down. This is red flag #2 in its purest historical form: debt-funded capacity expansion (or M&A) at the peak.

Capacity utilization, industry-wide, is the cycle's speedometer. Above ~85% markets are tight and producers regain pricing power, so new capacity gets announced as the capital cycle turns. Below ~75 to 78% it's a glut and a price war. At company level, utilization is also an operating-leverage dial. A smelter or a blast furnace runs flat-out or not at all, because a restart costs months and crores. Mines and rolling mills can flex. So check who can cut production without destroying themselves. Inflexible supply makes troughs deeper and longer.

Reserves come next. Miners and E&P companies eat their own balance sheet, because every tonne produced is a tonne gone. Three dials tell you how fast. Reserve life (R/P) = proved reserves ÷ annual production. Reserve replacement ratio (RRR) = reserves added (discoveries + extensions + revisions + acquisitions) ÷ production over the period. Above 100%, the company is more than replacing what it mines or pumps. Below 100% for years on end means it's slowly liquidating, and its terminal value is shrinking whatever the P&L says. The third dial is finding & development cost per unit, which is what the additions cost.

Here's a worked example (illustrative, Exxon-scale). Proved reserves ≈ 17.7 bn boe, and production ≈ 3.74 mn boe/day × 365 ≈ 1.37 bn boe/yr, so R/P ≈ 13 years. Additions that year were 1.9 bn boe, so RRR = 1.9 ÷ 1.37 ≈ 139%, which is healthy. Use a 3 to 5 year average, because annual RRR is noisy. SEC rules force price-based revisions: reserves count as "proved" only if they're economic at mandated prices, so a price crash mechanically deletes reserves.

Know the categories: 1P (proved, ≥90% confidence), 2P (proved + probable, ≥50%), 3P (+ possible). SEC filings show 1P plus a discounted cash value called the standardized measure (cousin of "PV-10"). It's crude, but it's a useful floor-value cross-check. One footnote on Coal India: its reserve life runs to many decades, and the constraint there is evacuation and pricing policy. Geology isn't the problem.

Inventory gains and losses: strip them out (worked adjustment). A refiner holds 30 to 60 days of crude and products, and a steelmaker holds months of coal and ore. When the input price moves, the P&L books a gain or loss that has nothing to do with operating skill. Cheap old inventory flows through COGS while products sell at new, higher prices. Those are the FIFO layers you met in inventory accounting. Take Sindhu Refining. Average inventory ≈ 18 mn bbl (~45 days), and crude rose $20/bbl over the year, so the inventory gain ≈ 18 mn × $20 = $360 mn. Spread over 146.6 mn bbl of throughput, that's $2.46/bbl of flattered GRM. Reported GRM was $7.5/bbl, so core GRM ≈ $5.0/bbl, and that's the number you value the company on. Companies usually disclose the figure in press releases or on calls. When they don't, estimate it the same way: inventory volume × price move.

This trap runs both ways. In crashing markets, huge inventory losses make core earnings look worse than they are, and that's the mirror-image opportunity. Red flag #3 is a company celebrating an earnings "beat" that's mostly inventory gains, then going quiet about the same line when prices reverse.

That silence is the tell.

4.6 Mid-cycle valuation, end-to-end

Here's the method that replaces the trailing P/E. It takes six steps, and we'll work them on Deccan Steel at the FY6 peak (price ₹390) and the FY2 trough (price ₹135).

Step 1. Assemble a full cycle of the driver. You want 8 to 10 years of EBITDA/tonne (or GRM, or realized spread). Deccan's series, in ₹/t, runs 9,000 · 6,000 · 8,000 · 12,000 · 18,000 · 26,000 · 13,000 · 8,000.

Step 2. Compute mid-cycle economics. Average EBITDA/t = 100,000 ÷ 8 = ₹12,500/t. Use the average unless the industry has changed structurally. India's post-2016 trade protection or consolidation might justify a mid-cycle modestly above history, and China's export surges might justify a lower one. Adjust consciously and in writing, or don't adjust at all. Check the median against the mean too, since big peak years drag the mean up. Mid-cycle EBITDA = 5 mt × 12,500 = ₹6,250 cr.

Step 3. Apply a mid-cycle multiple. Deep cyclicals such as steel and base-metal miners command 5 to 6× EV/EBITDA on mid-cycle EBITDA. Quality diversifieds and integrated energy get 6 to 7×, and converters and administered-price businesses get more. Use 5.5×, so EV = 5.5 × 6,250 = ₹34,375 cr.

Step 4. Subtract current net debt, never a normalized one. Debt doesn't move with the cycle. It's owed in full at the trough too. Equity value = 34,375 − 12,000 = ₹22,375 cr → ₹224/share (100 cr shares).

Step 5. Cross-check against replacement cost. Greenfield integrated flat steel in India costs roughly ₹8,000 to 10,000 cr per mtpa, or ≈ $1.0 to 1.2 bn, and brownfield ~60 to 70% of that (illustrative, so check against recent announced projects). Deccan's mid-cycle EV of ₹34,375 cr on 5 mt gives EV ≈ ₹6,875 cr per mtpa, about 86% of bottom-of-range replacement cost. You can read that two ways. (a) It's a margin of safety. You're buying capacity below what a rational entrant must pay to build it, so new supply can't undercut you profitably. (b) It's a check on the multiple. If your Step-3 EV had come out far above replacement cost, entrants would eventually build, capping long-run returns, so you'd haircut the multiple. P/B serves the same instinct at the trough. Deep cyclicals bought below tangible book stack the odds your way, as long as the assets aren't obsolete.

Step 6. Compare with price and overlay the cycle clock. Mid-cycle value ≈ ₹224. At the FY6 peak the stock trades at ₹390, 74% above mid-cycle value, on 5.1× trailing P/E while sector capex booms. That's the full sell signature. At the FY2 trough it trades at ₹135, 40% below mid-cycle value, on 90× trailing P/E while supply exits. That's the full buy signature, provided the balance sheet survives. Deccan's numbers say it can. Net debt/mid-cycle EBITDA = 12,000/6,250 = 1.9×, inside the 2× line, and trough interest cover is 3.0×, tight but alive. Check the maturities too. Then write the conclusion as a range: "mid-cycle value ₹200 to 250; buying meaningfully below, selling meaningfully above."

Two caveats. First, mid-cycle valuation assumes mean reversion. That holds for commodities with functioning cost curves, but "this time" occasionally is different in level. China's 2002-08 demand step-change was one such time, and the 2021-22 energy-security repricing was another. So Step 2's adjustments have to be conscious and conservative. Second, the method deliberately refuses to time the cycle, so you'll look wrong for quarters at a time.

Passing the balance-sheet screen, surviving the trough, is what buys you the right to wait.

4.7 Playbook variants across the complex

This is one page of dialect differences, and the grammar above holds everywhere.

• Upstream oil & gas (an Exxon-style major; India: ONGC/Oil India). Watch realization vs Brent after grade and quality discounts. Watch lifting cost/boe (the opex of production) and F&D cost/boe. Reserve life and RRR matter as they do in the KPI panel. Then compare production growth with decline rates: existing fields decline 4 to 7%/yr, so capex must first offset that treadmill. Integrated majors mute the cycle, because upstream, refining and chemicals peak at different times. Restraint with capital is the modern religion here, so judge the 5-year record. Did they cut buybacks to fund peak-priced acquisitions? Or did they return cash and buy in busts? For valuation, use mid-cycle EV/EBITDA (5 to 7×), EV/boe of reserves against peers, and the standardized measure as a floor cross-check.
• Refiners (India: the OMCs, Reliance's O2C). They're spread-takers on GRM. Indian OMCs add marketing margins on petrol and diesel, which the state implicitly caps in high-crude periods, so there's a political overlay on the economics. Model refining and marketing separately. Complexity, measured by the Nelson index, buys a structural GRM premium. Inventory gains and losses are quarterly noise, so always strip them out.
• Miners (BHP/Rio-style diversifieds; India: NMDC, Coal India, Vedanta). The cost curve is almost the whole analysis. Add reserve life and grade trajectory. Jurisdiction and royalty risk matter too. Diversifieds smooth the cycle across commodities, while single-commodity miners are the purest cyclicals listed. Coal India, as you saw at the start, is an administered-price, utility-like machine with e-auction upside. Value it closer to a yield asset, on dividend capacity off FSA volumes, with a cyclical kicker on top. And keep an eye on the energy-transition question for its terminal value.
• Steel & aluminium (Tata Steel/JSW/Hindalco-style). The dials are spread economics, captive-input advantages, China policy as an exogenous driver and trade barriers such as safeguard duties and MIP episodes. Hindalco is a hybrid you have to take apart. Indian smelting is an LME-driven price-taker, and Novelis is a stable conversion-fee-taker. Value the two on different multiples and add them up. This sum-of-parts approach, from special situations, recurs across the sector.

4.8 Commodity red flags (the brief's five, expanded)

1. Rising cash cost / sliding down the cost curve. When costs rise faster than peers and faster than the currency explains, you're looking at grade decline or aging assets, or perhaps slipping cost control. Whichever it is, the sector's only moat is eroding. To check, compare the 5-yr cash-cost CAGR with 2 peers, in USD.
2. Debt-funded capacity expansion at the peak. This is the Corus/Novelis signature. Record EBITDA arrives, a "transformational" acquisition or a mega-capex plan follows, and leverage goes to 3 to 4× peak EBITDA, which is 6 to 10× mid-cycle. Meanwhile the best operators do the opposite, and buy assets from these people five years later.
3. Net debt climbing into a downturn. Watch the direction of absolute net debt as the spread rolls over. A company still spending as EBITDA halves is scheduling its own covenant problem. Cross-check capex against guidance, look for a working-capital build, and see whether dividends are being kept up with borrowed money.
4. Earnings flattered by inventory gains. Make the inventory adjustment from the KPI panel and strip them out. Be suspicious when management's headline numbers stop disclosing the split in down-quarters.
5. Low trailing P/E at the peak marketed as cheapness. This is the cyclical trap again. When the sector leader is the "cheapest stock in the index" on trailing earnings and sell-side notes extrapolate record spreads, the cycle is late. Its mirror flag is writing the sector off at the trough because "there are no earnings."

Practitioners add two extras. The first is hedging opacity. A producer whose realization comes loose from benchmark has sold away the upside you're paying for, so read the derivatives note. The second is government wedge risk, meaning windfall taxes, price caps, export duties. These are how the state inserts itself between benchmark and shareholder, right at the peaks. India's 2022 steel export duty and crude windfall levy are the recent specimens. Verify their current status.

4.10 The developer KPI panel

Some of these dials need more than a line in a table:

First, check that collections track bookings. A booking is a contract, and collections are the cash. Over a trailing period on a mature portfolio, collections should run 85 to 100% of bookings. Timing gaps are normal, since collections follow construction milestones. A persistent, widening gap means one of three things. Bookings may be recognized aggressively, with flimsy 5%-down "sales" that will cancel. Construction may be too slow to trigger milestone demands, so the money is stuck behind execution. Or customer stress and cancellations are brewing. This is the sector's channel-stuffing check, the same forensic instinct as the receivables-vs-revenue divergence from your forensics work.

Then there's sales velocity, which is absorption at the project level. It asks what share of launched inventory sells in the first year, and strong Indian launches sell 40 to 60%+. Its companion is quarters-to-sell (QTS) = unsold inventory ÷ trailing quarterly sales pace. Velocity is pricing power's twin.

A developer holding realization while velocity collapses is choosing optics over cash.

With realization ₹/sqft, split any move into mix before you credit a price increase. Were this quarter's launches luxury or affordable? In MMR or in Pune? It's the same price/mix habit you used on FMCG.

Next comes the supply chain of sales. Pre-sales are manufactured from launches, launches come from the pipeline and the pipeline comes from land. Measure land two ways. One is years of runway, attributable msf ÷ annual sales pace. Then look at how it's held. Land owned outright is capital-heavy and rich in optionality. Under joint development agreements (JDA) and development-management deals, the landowner contributes land for a revenue or area share, and the developer contributes brand and execution. That's the Godrej Properties-style asset-light model, and post-RERA economics favor it. A decade-plus land bank bought with debt looks like strength. In practice it's rent paid to hold optionality, and the RERA era's tighter cash rules no longer reward that.

Watch unsold and aging inventory as well. Unsold stock under construction is normal working stock. Completed unsold inventory is the damning line, because the market has seen the finished product at the asking price and said no. It ties up capital, and its eventual clearing price sets marks for the whole micro-market. Track it in ₹ cr, in msf and in age buckets. Growth in the >12-months-completed bucket is a direct red flag.

Net debt/equity <0.5×, the post-RERA constitution. The pre-2017 model ran leverage of 1 to 2×+, because customer advances arrived upfront and could be diverted. You collected on project A, bought land for project B and borrowed against both. RERA ended the diversion, as you'll see next. The 2018 NBFC crisis ended the refinancing, since IL&FS and DHFL were the sector's lenders. Today's survivors run net D/E of 0.2 to 0.5×, and several are near net cash. So treat >0.5× today as a structural problem, whatever the cycle is doing.

The OCF self-funding test is the one-line model audit. Run it every year:

Collections − construction spend − overheads & marketing − interest ≥ land/BD spend + dividends

Run it on Meridian Estates. Collections ₹3,200 cr − construction ₹1,750 cr − overheads ₹420 cr − interest ₹180 cr = ₹850 cr surplus. Against land and business-development spend of ₹600 cr, that's self-funding, and net debt falls ₹250 cr while the company grows. Now flip the land spend to ₹1,400 cr. The same "growth story" is being bought with ₹550 cr of new debt, and you've rebuilt the pre-RERA machine. Here's a screener shortcut: cumulative CFO over 5 years against cumulative debt change tells you which developer you own.

4.11 RERA: the regime change that rewrote the model

The Real Estate (Regulation and Development) Act, 2016 is why the modern panel looks the way it does. State rules mostly took effect in 2017. Here are the mechanics that matter to you as an analyst:

• Projects above 500 sq m or 8 units must be registered before any marketing or sale. That ended the pre-launch, collect-cash-first game.
• The 70% escrow rule is the heart of the Act. 70% of customer collections must be held in a project-specific account, and money comes out only in proportion to certified construction and land progress (architect + engineer + CA certificates). This one clause killed the float-diversion model you just met, because customer money now physically can't fund the next land deal.
• Delay penalties have teeth. Buyers get interest for late possession (typically ~SBI MCLR + 2% under state rules). They also get refund rights and a 5-year structural-defect liability.
• Developers sell on carpet area, using standardized and audited area definitions. They must also disclose approvals, timelines and encumbrances.

You can see the consequences in the numbers. Small developers financed by float exited, and organized players' share of top-city launches roughly doubled. Balance sheets deleveraged toward the <0.5× norm. JDA and asset-light models spread, because land capital became the binding constraint. And quarterly operational disclosure (bookings, collections, launches) became the sector's lingua franca. That one's a dividend for you as an analyst. RERA turned a working-capital arbitrage business back into a development business, and the KPI panel above is simply the audit trail of that change. State-level enforcement varies. Maharashtra's MahaRERA is the deepest public database, so use it to check a developer's project pipeline independently.

4.12 Developer NAV: the valuation logic

Earnings-anchored multiples fail here for the reason you've already seen: a P/E on meaningless revenue means nothing. So the sector's intrinsic lens is Net Asset Value, the present value of everything the developer will harvest, minus what it owes.

NAV = Σ over projects [PV of (future collections − remaining construction cost − SG&A − taxes)] + land bank at realizable value + other assets (rental/annuity portfolio at cap-rate value) − net debt

Build it project by project, since the disclosures let you. For each ongoing or planned project, take saleable area × expected realization to get GDV, then subtract costs. Phase the net cash flows over the sell-and-build years and discount them at 13 to 15%. That's above a WACC, because approval, execution and market risks stack up. Add sold-but-uncollected receivables net of cost-to-complete. Land bank goes in at cost (conservative) or marked to market with a haircut. Say which you used.

Then comes P/NAV. Developers have historically traded at 0.6 to 1.1× NAV. Discounts reflect leverage and execution record. They also reflect governance and any optimism in the NAV assumptions. Governance is the reason this sector always gets paired with the promoter-forensics toolkit. Premiums go to platforms that reliably add NAV through asset-light deals, since a brand that wins JDAs is worth more than its current project list.

Two sanity cross-checks help. The first is EV/pre-sales, a look-through screen of EV ÷ annual bookings read against embedded project margins. Bookings carry ~20 to 30% eventual PAT margins, so EV/pre-sales of 3× implies ~10 to 15× look-through earnings on the current sales pace. That gives you a fast way to see whether growth hopes are already paid for. For the second, compare market cap with the embedded margin in bookings already sold but not yet recognized, which companies disclose as unrecognized revenue.

4.13 REITs: the machine that converts buildings into cash flow

A REIT is a listed trust that owns completed, rent-producing real estate and is legally obliged to pass the rent through. India's rules are the SEBI REIT Regulations, 2014. The market went live with Embassy Office Parks' 2019 listing, and Mindspace, Brookfield India and Nexus Select followed. Here's what the rules require:

• ≥80% of asset value must be in completed, income-generating properties, with ≤20% in under-construction or other permitted assets.
• ≥90% of Net Distributable Cash Flows (NDCF) must be distributed at least half-yearly, and in practice it's quarterly.
• Aggregate leverage is capped at 49% of asset value. There's a tripwire at 25%, beyond which fresh borrowing needs a credit rating and unitholder approval.

Those are as of mid-2026, so verify them. SEBI periodically consults on relaxations, and in 2024 it added a small-and-medium REIT framework for ₹50 to 500 cr assets. As for structure, a trustee holds the assets through SPVs or a HoldCo, a manager runs them for a fee, and a sponsor seeds them under lock-ins and skin-in-the-game rules. In the US, home of the original REIT, REITs must distribute ≥90% of taxable income, and in exchange they pay no corporate tax on distributed income. It's the same pass-through economics, with four decades more variety.

That variety matters for your map. India's listed REITs are so far concentrated in Grade-A offices, plus one retail-mall REIT, tenanted by IT/GCC multinationals on long leases. That's one asset class with one demand driver. The US menu spans offices, malls, industrial/logistics warehouses (Prologis-style), apartments, single-family rentals, data centers, cell towers, healthcare, self-storage, timber and net-lease (Realty Income-style, WALEs of 9 to 14 years), and each has its own cycle. The framework below transfers to all of them, and to India's InvIT cousins holding roads, transmission lines and pipelines. Only the thresholds move.

4.14 The REIT operating panel: NOI, occupancy, WALE, re-leasing spreads

NOI (Net Operating Income) = property revenue (rent + parking + maintenance recoveries) − property-level operating expenses (maintenance, property tax, insurance, utilities net of recoveries). All of it comes before corporate G&A, management fees, interest and depreciation. Think of NOI as the building's own P&L. It's the number cap rates apply to, and it's the sector's EBITDA-equivalent. Grade-A Indian office NOI margins run ~80 to 85%.

Same-store (like-for-like) NOI growth is NOI growth on the constant portfolio. Acquisitions and new completions are stripped out. It's your organic-health dial. For Indian offices it breaks down into contractual escalations (typically ~4 to 5%/yr, or 15% every 3 years), occupancy change and re-leasing spreads. Picture a REIT growing headline NOI 15% through debt-funded acquisitions while same-store NOI shrinks. That's a treadmill, however good the headline looks. You learned this same-store habit on retail SSSG.

Occupancy >93% is the healthy bar for stabilized portfolios. That's the bar we use here, and pre-COVID Indian office REITs ran 92 to 95%. Below it, two problems compound. You lose rent, and you lose pricing power on every renewal negotiated while vacancy is visible on the floor plate. Read occupancy in three forms: physical (space handed over), committed (leases signed, rent starting soon) and economic (rent actually flowing). Managements quote the flattering one, so reconcile all three yourself. Indian office occupancies sagged to the mid-80s% in 2021-23, on work-from-home plus tenant exits, before the GCC leasing wave rebuilt them. It's a live reminder that the threshold is a stabilized norm. It isn't a law.

WALE (Weighted Average Lease Expiry) is the average remaining lease term weighted by rent. It's your cash-flow visibility dial. Indian office WALEs run ~6 to 8 years, and US net-lease runs 9 to 14. Pair it with the lease-expiry schedule. If 8% of rent expires each year, that's routine maintenance. If 25% is bunched into one year, that's refinancing risk in operating clothes.

Re-leasing spread = (new rent on re-let space − expiring rent) ÷ expiring rent. It's the mark-to-market truth serum, because it tells you whether in-place rents are below or above market. Indian office REITs have enjoyed persistent positive spreads. Their in-place rents run 15 to 25% below market, which is embedded growth you get paid to wait for. Post-2020 US offices printed negative spreads, as leases signed in 2015 rolled down to a weaker 2023 market, a BXP-style headwind. Meanwhile Prologis-style logistics printed +30 to 50%+ on the e-commerce warehouse squeeze.

One number, three different theses.

4.17 Real estate & REIT red flags (the brief's five, expanded)

1. Pre-sales miss. The real top line falls short of guidance or trajectory. Watch especially for volume misses dressed up with realization mix. One soft quarter is weather. Two or more is demand, or the product.
2. Collections lagging bookings. This is the gap from the developer panel: booked "sales" that aren't turning into cash. It points to cancellation risk, execution delay or stuffed bookings. Pair it with rising receivables from customers.
3. Aging unsold inventory. Completed stock that keeps growing, with more of it aged >12 months, is failed price discovery. It will eventually mark the whole book. Watch developers who'd sooner borrow than cut price.
4. AFFO payout >90%, or NDCF propped up by return of capital or refinancing. Either way the distribution has been engineered past its cash engine, and the cut, when it comes, hurts both income and multiple.
5. Occupancy slide + negative re-leasing spreads. This is the lethal REIT combination. Volume falls and price falls on the space you re-let, so NOI declines compound just as the leverage math (fixed debt, shrinking NOI) turns against you. Add the leverage corollaries: LTV creeping toward the cap, bunched lease expiries and a WALE that shortens year after year.

Both halves come with governance extras. Look for related-party construction contracts, where the developer builds through a promoter-owned contractor, and for land bought from promoters at fancy marks. Check whether REIT managers are paid on assets (which feeds acquisition appetite) or on per-unit distributions. And track sponsor stake sales and pledges. Every promoter-forensics tool you've learned applies here.

4.18 Synthesis: what the two playbooks share

Same skeleton, different flesh. Four ideas run through both playbooks.

1. The reported P&L misleads by construction. Commodities have cyclical earnings and development has lagged recognition, so each sector built a truer metric stack. Commodities answer with EBITDA/tonne and mid-cycle spreads. Developers answer with pre-sales and collections. REITs answer with NOI and AFFO.
2. Asset-anchored valuation keeps earnings fantasy in check. Replacement cost per tonne, developer NAV, cap-rate NAV: each is a version of one question. What would a rational private buyer pay for the assets?
3. Leverage is the killer variable. Commodity cash flows swing. Development cash arrives late. REIT cash looks bond-like until occupancy slips. So <2× net debt/mid-cycle EBITDA, <0.5× net D/E and an LTV under 49% (really under 35%) are the same idea in three dialects.
4. The cycle clock outranks the forecast. For commodities that means cost curves and capital cycles. For real estate it means interest rates, credit availability and supply pipelines. Buy either sector when capital is fleeing and the trailing numbers look worst, and be afraid when the trailing numbers look best.

Get that inversion into your bones and you've passed the graduation test here. In the phase capstone, airlines and hotels will test it again, in service-industry costume.

Common mistakes & how experts think differently

1. "It's at 4× earnings, so how much cheaper can it get?" This is the sector's defining retail error, and it comes back every cycle. The novice reads a low trailing P/E as a discount. The expert reads it as the market's memo that earnings are peaking. Experts re-anchor every multiple to mid-cycle earnings first, and they treat single-digit trailing P/Es on record spreads as a timer where the novice sees a bargain. It works in reverse at the bottom too. Experts don't flee the 90× trough P/E. They check the balance sheet and the cost curve, and then they get interested.

2. Valuing the company as if the current price is permanent. Extrapolate this year's HRC price, GRM or LME into a 10-year DCF and you've built a castle on the number in finance that looks least mean-reverting and is most mean-reverting. Experts force every commodity forecast through the cost curve, where long-run price ≈ the marginal producer's full cost. They push it through the capital cycle too, asking what supply has been announced. Then they express conviction as a mid-cycle range.

3. Netting normalized earnings against normalized debt. This one's subtle. Analysts who correctly normalize EBITDA sometimes go on to use an average or "comfortable" net debt figure. But debt is owed at face, today. It's mid-cycle EV minus current net debt, always. A related slip is computing net debt/EBITDA on peak EBITDA and calling 1× "underlevered", when it's 3 to 4× on mid-cycle.

4. Treating all "commodity" companies as the same animal. The mistake is lumping a converter (Novelis), a spread-taker (a refiner), an administered-price miner (Coal India) and a pure price-taker (an aluminium smelter) under one multiple. Experts classify the exposure first, and they often split one company into parts with different multiples. Hindalco's smelter and converter are the classic case.

5. Reading a developer's P&L like a normal company's. Revenue crashed 40%, so sell? Not so fast. That revenue is FY-minus-four's news. Experts read bookings, collections and launches. They treat the P&L as a delayed broadcast, useful mainly to confirm margins on completed projects. The mirror error is crediting a "revenue doubling" that's really one big tower reaching completion.

6. Taking bookings at face value. Pre-sales are contracts. Cash is something else, and a 5%-down booking wave can evaporate through cancellations. Experts run the collections-vs-bookings track and the OCF self-funding test before they believe any growth story.

Cash collected is the only booking that counts.

7. Applying the office-REIT playbook to a developer (or vice versa). A REIT is a yield machine on completed assets. A developer is a manufacturing business with multi-year inventory. P/NAV shows up in both, but the NAV mechanics are different species: a cap rate on in-place NOI for one, project DCFs on future sales for the other. The risk drivers differ just as much: occupancy and rates against execution and demand. Experts also refuse the opposite muddle of valuing a US homebuilder on P/NAV. Its economics are turns-based, so the lens is P/B × through-cycle ROE.

8. Trusting FFO, or worse a company-defined AFFO, without auditing the maintenance-capex line. FFO adds back depreciation, and if you stop there you've priced the buildings as free to maintain. AFFO is self-defined, so a management that routes recurring refurbishment into "growth capex" manufactures distributable cash. Experts rebuild AFFO themselves and compare maintenance capex + TI/LC per sqft across peers. They also read the NDCF composition. How much of the distribution is return of capital?

9. Chasing the highest distribution yield. The REIT with the fattest yield usually carries the >90% payout, the bunched expiries or the 45% LTV. Experts price the yield against its growth and its risk. They read the yield spread to the G-sec together with AFFO payout, occupancy trend, re-leasing spreads and LTV headroom. A 6% yield growing 5% beats an 8.5% yield that's about to be cut.

10. Forgetting that the state is a player in these sectors. Its levers are everywhere here: windfall taxes, export duties, price freezes, mining royalties, RERA, stamp duties, development regulations, SEBI's REIT leverage cap. The expert's habit is to list, before finalizing any thesis, the three government levers that could move it. Then they check the current settings against primary sources (PIB releases, SEBI circulars, state RERA sites). It's the reason several specifics above carry "(as of mid-2026)".

Worked examples

WE1: Full mid-cycle valuation of a steelmaker (integrating §4.3–§4.6)

Vindhya Steel (synthetic) has 8 mt of volumes, 120 cr shares, net debt of ₹18,000 cr, D&A of ₹2,600 cr, interest of ₹1,500 cr and a 25% tax rate. Its ten-year EBITDA/tonne history (₹) runs 8,500 · 6,000 · 7,500 · 9,000 · 11,500 · 21,000 · 28,000 · 14,000 · 9,500 · 10,000. Its current price is ₹310. For latest-year EPS, EBITDA = 8 × 10,000 = ₹8,000 cr → PAT = (8,000 − 2,600 − 1,500) × 0.75 = ₹2,925 cr → EPS ₹24.4 → trailing P/E 12.7×.

Step 1. Mid-cycle EBITDA/t. Mean = 125,000 ÷ 10 = ₹12,500. The median is (9,500 + 10,000)/2, but you only see that once you order the series: 6,000 · 7,500 · 8,500 · 9,000 · 9,500 · 10,000 · 11,500 · 14,000 · 21,000 · 28,000. So median = (9,500 + 10,000)/2 = ₹9,750. The two peak years, 21,000 and 28,000, drag the mean well above the median. It's a classic skew. Now make a judgment call, in writing: use ₹11,000. That's between the median and the mean, and it acknowledges that post-consolidation industry structure is somewhat better than the early-decade troughs.

Step 2. Mid-cycle EBITDA. 8 mt × ₹11,000 = ₹8,800 cr.

Step 3. Multiple. It's an integrated player with half its ore captive, so use 5.5×. EV = ₹48,400 cr.

Step 4. Subtract current net debt. Equity = 48,400 − 18,000 = ₹30,400 cr → ₹253/share. At 5 to 6× EV/EBITDA the range is ₹217 to ₹290.

Step 5. Replacement-cost cross-check. EV/tonne = 48,400 ÷ 8 = ₹6,050 cr per mtpa, against ₹8,000 to 10,000 cr for greenfield. That values the company at ~60 to 75% of replacement cost. The multiple isn't inviting new entry, so it needs no haircut.

Step 6. Verdict. At ₹310 the price is ~7 to 43% above the ₹217 to 290 range, whose midpoint is ₹253. That's moderately expensive. The cycle read gives you no support for paying up either: EBITDA/t of ₹10,000 ≈ mid-cycle, and sector capex is re-accelerating. On the balance-sheet screen, net debt/mid-cycle EBITDA = 18,000/8,800 = 2.05×, right at the healthy limit and fine unless it grows. So the conclusion is a watchlist company. That answer came from the analysis.

The trailing 12.7× P/E looks unremarkable and tells you nothing.

WE2: Refiner: strip the inventory gain, then value the spread

Kaveri Refining (synthetic) runs 15 mmtpa of throughput, and 15 × 7.33 ≈ 110 mn bbl/yr. Its reported "GRM" for the year is $8.2/bbl and cash opex is $1.9/bbl. Average inventory of crude and products is 14 mn bbl, and crude rose $18/bbl over the year. Net debt is ₹9,000 cr, there are 60 cr shares, and the rate is ₹84/$.

Step 1. Inventory gain. 14 mn bbl × $18 = $252 mn. Per barrel of throughput, that's 252 ÷ 110 = $2.29/bbl.

Step 2. Core GRM. 8.2 − 2.29 ≈ $5.9/bbl. Check it against the benchmark. If the Singapore complex GRM averaged, say, $5.1, the complexity premium is $0.8, which is consistent with history and believable.

Step 3. Mid-cycle GRM. The ten-year Singapore average ≈ $5.5, and adding the company's ~$0.8 premium gives $6.3/bbl mid-cycle. The company's own 10-yr core-GRM average is $6.4. They converge, so use $6.3.

Step 4. Mid-cycle refining EBITDA. (6.3 − 1.9) × 110 mn = $484 mn ≈ ₹4,066 cr.

Step 5. Value. A complex refiner gets 6× EV/EBITDA → EV ≈ ₹24,400 cr − net debt 9,000 → equity ₹15,400 cr → ₹257/share, with a range of ₹189 to ₹324 at 5 to 7×. Now see what the reported $8.2 GRM would have done. EBITDA (8.2 − 1.9) × 110 = $693 mn → equity value ≈ ₹432/share. That's a 68% overvaluation built entirely on inventory gains and a peak spread, or red flags #4 and #5 wearing numbers.

WE3: Developer: from bookings to NAV, with the P&L lag on display

Meridian Estates is back (synthetic, the developer from the P&L-lag example). It has 100 cr shares, net debt of ₹1,100 cr and book equity of ₹5,600 cr, so net D/E is 0.20×, healthy by post-RERA standards. In the operating year, bookings were ₹3,900 cr (4.2 msf @ ₹9,286/sqft blended). Collections were ₹3,200 cr, or 82% of bookings. That's acceptable on an expanding book, though the trend is worth watching. Launches came to 5.1 msf. Completed unsold inventory is ₹240 cr, 2.5 quarters-to-sell at the current pace, which is clean. Reported P&L revenue was ₹2,100 cr. Those are towers completed this year and sold largely 3 to 4 years ago at a ₹7,400/sqft average, so you can see the lag right there.

OCF self-funding test. Collections 3,200 − construction 1,750 − overheads 420 − interest 180 = +850. Land/BD spend is 600, so there's a surplus of ₹250 cr and net debt falls. The company is self-funding while growing bookings ~15%, so the model is intact.

NAV build (condensed):

• Ongoing + planned projects: 28 msf sellable at a blended realization of ₹9,000/sqft gives GDV of ₹25,200 cr. Costs (land balance + construction + SG&A) are ₹16,400 cr, leaving a pre-tax project surplus of ₹8,800 cr. Phased over ~6 years and discounted at 14%, the post-tax PV ≈ ₹5,300 cr.
• Sold-but-uncollected receivables net of cost-to-complete on sold area: ₹1,150 cr.
• Land bank beyond the launched pipeline (9 msf at conservative realizable value): ₹1,600 cr.
• Annuity assets (a small office block, NOI ₹120 cr at 8.5%): ₹1,410 cr.
• Less net debt (₹1,100 cr), and less the PV of unallocated corporate overhead (₹700 cr).

NAV ≈ 5,300 + 1,150 + 1,600 + 1,410 − 1,100 − 700 = ₹7,660 cr → ₹77/share. At a market price of ₹66, P/NAV is 0.86, a modest discount for a self-funding, low-leverage operator. Cross-check with EV/bookings = (6,600 + 1,100) ÷ 3,900 ≈ 2.0×. Against ~25% embedded PAT margins, that's a look-through ~8× current bookings' eventual earnings, which isn't demanding. Both lenses agree. What would change the verdict is the panel, and the P&L won't tell you: a collections slide toward 70%, velocity stalling on new launches, or land splurges flipping the OCF test.

WE4: REIT: the full yield-machine workup

Horizon Business Parks REIT (synthetic) carries over from the FFO and valuation work above. Pull everything into the one-screen verdict an analyst actually writes:

Now practise writing the verdict. It's the memo skill of competency C11, communicating like a consultant: "Horizon offers a 6.0% growing yield at 0.70× NAV with conservative leverage. The price already embeds a bear case on office demand (implied cap 10.3% vs 8% private-market). Thesis risks, in order: occupancy stalls below 93% (killing the growth that justifies the negative bond spread), payout at 87% leaves little cushion, and duration, because a 150 bps G-sec selloff mechanically reprices the unit regardless of operations. Position only with the first two dials trending right."

Notice the shape. Every claim traces to a panel dial, and the argument rests on the price-implied view, the implied cap rate, more than on the yield alone.

WE5: US logistics REIT: FFO/AFFO, NAV, and the Treasury spread (a US-dollar mirror of WE4)

Cedar Logistics REIT (synthetic, US industrial/warehouse, Prologis-style) owns 120 msf of e-commerce fulfilment and last-mile warehouses and has 250 mn units. You'll take the operating dials first, then the cash walk and the price-implied view. It's the same shape as the Horizon workup, in dollars, for an asset class India's REITs don't offer yet.

The FFO → AFFO walk:

AFFO/unit = 850 ÷ 250 = $3.40. With $2.72 declared, AFFO payout = 80.0%, right on the healthy line, since <80% is the cushion zone. For a REIT re-leasing at +38%, that retained ~20% funds growth capex without issuing units.

The valuation triangle has three corners.

1. NAV. Completed portfolio = NOI ÷ cap rate = 1,400 ÷ 5.5% = $25,455 mn. Prime US logistics has transacted at ~4.5 to 6% (as of mid-2026). Add the development pipeline & land at $1,500 mn and take off net debt, which is gross $9,000 − cash $300 = $8,700 mn. NAV = $18,254 mn = $73.0/unit, so at a unit price of $62, P/NAV = 0.85.
2. Implied cap rate. EV = market cap (62 × 250 = 15,500) + net debt 8,700 − non-income assets 1,500 = $22,700 mn. Implied cap = 1,400 ÷ 22,700 = 6.2% vs the 5.5% appraisal, so the market is pricing the buildings ~11% below appraisal.
3. Yield vs the bond. Distribution yield = 2.72 ÷ 62 = 4.4%, against an illustrative 10-year US Treasury of ~4.3%. That's not a dated quote, so verify the current yield on FRED DGS10. The result is a +9 bps spread, and with +38% re-leasing spreads and +7% same-store growth behind it, that thin number is doing far more work than it looks.

LTV = 9,000 ÷ (25,455 + 1,500 + 300) = 33%, which is conservative. No SEBI-style cap applies in the US, but the same <35% operating-comfort instinct does.

Write the verdict the way you did for Horizon. This is a low-yield, high-growth industrial REIT priced at 0.85× NAV on a 6.2% implied cap while the private market pays 5.5%. If the discount is real, it's the market extrapolating today's rate fear across a portfolio whose rents are 38% below market. The thesis risk fits in one line: a warehouse supply wave compressing those re-leasing spreads before the in-place discount is harvested. Real estate gets capital cycles too.

Take this as a teaching specimen. It isn't a recommendation.

WE6: Aluminium hybrid: value the smelter and the converter as different animals (sum-of-parts, ₹)

Kanha Metals (synthetic, Hindalco-style) is one listed company with two utterly different machines inside it. One is an Indian primary-aluminium smelter, a pure LME price-taker. The other is a global rolled-products converter, a Novelis-style conversion-fee-taker. The playbook variants and common mistake #4 both warn against one multiple for the whole group. Here's the arithmetic behind the warning.

Segment 1. India smelter (price-taker). It makes 1.3 mt/yr at a cash cost of ~$1,850/t. Against LME aluminium of $2,200 to 2,600, that's low-2nd quartile on the global cost curve: profitable through most troughs, uncomfortable below ~$1,900. Mid-cycle EBITDA/t ≈ ₹54,600 (~$650 at ₹84/$), so mid-cycle EBITDA = 1.3 mt × ₹54,600 = ₹7,098 cr. A pure deep cyclical earns a mid-cycle multiple of 5.0× → EV = ₹35,490 cr.

Segment 2. Global converter (fee-taker). It rolls 4.0 mt/yr of products, and conversion EBITDA/t ≈ ₹39,480 (~$470/t). That's the stable cousin, barely moving with the LME because the metal itself is a pass-through. Mid-cycle EBITDA = 4.0 mt × ₹39,480 = ₹15,792 cr. Because that stream is far less cyclical, it earns a richer 7.0× → EV = ₹110,544 cr.

Sum the parts. Enterprise value = 35,490 + 110,544 = ₹146,034 cr. Take off consolidated net debt of ₹42,000 cr. The converter carries the acquisition debt, and net debt ÷ mid-cycle EBITDA = 42,000 ÷ 22,890 = 1.8×, which is survivable. That leaves equity of ₹104,034 cr ÷ 220 cr shares = ₹473/share.

Why splitting pays. A screen that slaps a single deep-cyclical 5.0× on the group's total ₹22,890 cr mid-cycle EBITDA gets EV ₹114,450 cr → equity ₹72,450 cr → ₹329/share. So at a market price of ₹430, the single-multiple analyst sees a stock 31% above "fair value" and calls it expensive. The sum-of-parts analyst sees ₹430 sitting 9% below a ₹473 value and gets interested. That whole ~₹144/share gap comes from one thing: refusing to value a stable fee business as if it were a steel mill. Classify the exposure before you reach for a multiple, and for a hybrid, value the pieces and add them up.

Take this one as a teaching specimen too. It isn't a recommendation.

Practice set

Work each problem fully before you open the solution. Problems 1 to 4 are guided, 5 to 9 are independent and 10 to 12 are timed, at 12 minutes each with closed notes, in exam conditions.

P1 (guided, realization vs benchmark). An aluminium producer sold 1.30 mt in FY-X. LME averaged $2,420/t and its regional product premium averaged $130/t. The company reported metal revenue of $3,120 mn. (a) Compute achieved realization per tonne and the realization ratio vs benchmark-plus-premium. (b) Name two innocent and two worrying explanations for the gap.

Solution. (a) Realization = 3,120 ÷ 1.30 = $2,400/t. Benchmark + premium = 2,420 + 130 = $2,550. Ratio = 2,400 ÷ 2,550 = 94.1%. (b) Two innocent explanations: the product mix may lean to standard ingot over value-added products, or contract prices may lag, with quarterly-priced contracts trailing a rising spot. Two worrying ones: hedges struck at lower prices may be selling away the upside (check the derivatives note), or grade and quality discounts may be widening, including distress discounting to move volume. A gap on its own isn't the flag. An unexplained or widening gap is.

P2 (guided, cost-curve survival). Industry cash-cost curve percentiles for a metal: 25th percentile $1,650/t, 50th $1,950, 75th $2,250, 90th $2,480. Producer A has cash cost $1,700/t and net debt/mid-cycle EBITDA 1.2×. Producer B has cash cost $2,400/t and 3.1×. A demand shock takes the price to $2,100. (a) Work out what each producer earns per tonne. (b) Where does the price find a floor, and why? (c) Who exits the trough stronger?

Solution. (a) A earns 2,100 − 1,700 = +$400/t. B earns 2,100 − 2,400 = −$300/t, so it's cash-negative on every tonne. (b) The floor forms near the ~90th-percentile cash cost ($2,400 to 2,500). Below that, enough supply is losing cash that closures and curtailments shrink supply until price stabilizes, so the cost curve is the floor's blueprint. State-owned or integrated producers can delay that, but they rarely repeal it. (c) A. Its 1st-quartile cost plus a light balance sheet means it compounds through the trough and can buy B's assets from the receiver. B's 3.1× leverage on mid-cycle EBITDA becomes 6 to 8×+ on trough EBITDA, and its equity holders are hostage to the price recovering before the maturities arrive.

P3 (guided, steel spread). Compute EBITDA/tonne: HRC ₹54,000/t; iron ore ₹6,200/t (1.6 t needed); coking coal ₹21,500/t (0.7 t); conversion costs ₹15,800/t; corporate overhead ₹1,400/t. Then recompute for a producer with captive ore at ₹2,800/t mining cost. What does the difference represent?

Solution. For the market-ore producer, spread = 54,000 − (1.6 × 6,200) − (0.7 × 21,500) = 54,000 − 9,920 − 15,050 = ₹29,030, and EBITDA/t = 29,030 − 15,800 − 1,400 = ₹11,830. For the captive-ore producer, ore cost is 1.6 × 2,800 = 4,480 → spread = 54,000 − 4,480 − 15,050 = 34,470 → EBITDA/t = ₹17,270. The difference of ₹5,440/t is a structural cost-advantage moat, the only durable kind in this sector. It persists at every point of the price cycle, and at the trough it's the difference between bleeding and earning.

P4 (guided, the P/E trap in one company). Go back to the Deccan Steel table. One investor buys at FY6 year-end at ₹390 because "5.1× earnings is the cheapest in the market." Another buys at FY2's close at ₹135 despite "a bubble-like 90× P/E." Using the table, compute each investor's return two years after purchase. Then state, in one sentence each, what the P/E was actually measuring.

Solution. Peak buyer, FY6 to FY8: 165/390 − 1 = −57.7%. Trough buyer, FY2 to FY4: 230/135 − 1 = +70.4%, and +189% if held to FY6. The 5.1× P/E was measuring the market's correct refusal to capitalize peak earnings, which is to say the top of the cycle. The 90× P/E was measuring collapsed trough earnings against a price already looking across the valley, which is the bottom. For deep cyclicals the trailing P/E is a cycle clock read backwards. Never use it as a value gauge.

P5 (mid-cycle valuation). A copper miner has volumes of 400 kt/yr, a 10-year EBITDA/t history averaging $2,100 (current year: $3,600), net debt of $1.9 bn and 800 mn shares. Peers trade at a mid-cycle EV/EBITDA of 5.5×. Its current price is $5.90. (a) Value it mid-cycle. (b) The stock trades at 4.6× trailing EV/EBITDA, a discount to peers' 7.1× trailing. Is it cheap?

Solution. (a) Mid-cycle EBITDA = 400,000 × 2,100 = $840 mn. EV = 5.5 × 840 = $4.62 bn, and equity = 4.62 − 1.9 = $2.72 bn → $3.40/share. (b) No. At $5.90 it trades 74% above mid-cycle value. The "cheap" trailing multiple is computed on peak EBITDA of 400 kt × $3,600 = $1.44 bn, which is 71% above mid-cycle. Both the stock's 4.6× and the peers' 7.1× rest on peak-inflated denominators, so the discount-to-peers framing compares two mirages. Before you short anything, run the legitimate counter-check. Is $3,600 the new mid-cycle, on a structural supply deficit? If so, that case has to be argued in writing through the cost curve and the announcement pipeline. You can't just assume it from the spot price.

P6 (reserves). An E&P company has proved reserves of 2,600 mn boe at year start, production of 240 mn boe during the year and additions from extensions/discoveries of 190 mn boe, plus upward price-driven revisions of 95 mn boe. (a) Compute year-end reserves, R/P, and RRR (with and without revisions). (b) Its 5-yr average RRR ex-revisions is 82%. What is the company, economically?

Solution. (a) Year-end = 2,600 − 240 + 190 + 95 = 2,645 mn boe. R/P = 2,645 ÷ 240 ≈ 11.0 years. RRR including revisions = (190 + 95) ÷ 240 = 119%, and ex-revisions it's 190 ÷ 240 = 79%. (b) It's a slow liquidation. Through the drill bit it replaces only ~4/5ths of what it pumps, and the headline >100% is price-revision noise that reverses when prices fall, under the SEC's price-based rules. Its DCF has to model declining volumes or rising F&D spend, and its dividend is partly a return of capital, whatever the payout ratio says.

P7 (inventory gains). A steelmaker holds ~2.5 months of coking-coal inventory (consumption 8.4 mt/yr of coal). Coal prices fell $60/t over the quarter. Reported quarterly EBITDA: ₹2,150 cr (₹84/$). (a) Estimate the inventory drag. (b) Core EBITDA? (c) Management's release calls it "operational excellence offsetting market weakness." Comment.

Solution. (a) Inventory ≈ 8.4 × (2.5/12) = 1.75 mt, so the drag ≈ 1.75 mt × $60/t = $105 mn ≈ ₹880 cr at ₹84/$. That's high-cost coal bought before the fall, flowing through COGS at old prices. (b) Core EBITDA ≈ 2,150 + 880 = ≈ ₹3,030 cr. The business did better than reported, and the inventory losses hid it. (c) The direction is right, but the release should quantify the effect. The test is whether management is even-handed. One that quantifies inventory losses in bad quarters but stayed silent on inventory gains in good ones is curating its disclosure. Your adjustment is an estimate, since FIFO layers and purchase timing blur it, so companies' own disclosed figures take precedence when they give them.

P8 (developer panel). Here's a developer's quarterly pack. Bookings were ₹1,240 cr against guidance of ₹1,400 cr, and collections were ₹880 cr. Collections/bookings over the last four quarters ran 96%, 91%, 83%, 71%. Completed unsold inventory rose from ₹310 cr to ₹520 cr over the year, and ₹280 cr of it is >12 months old. Net D/E is 0.62×, up from 0.44×. Land spend this year was ₹900 cr against OCF before land of ₹420 cr. Run the panel and write the three-line verdict.

Solution. Here are the flags, in causal order.

1. Pre-sales miss, at −11% vs guidance.
2. Collections lagging bookings. The ratio has slid from 96 to 71% over four quarters, so booked sales aren't converting to cash (cancellations or stalled execution).
3. Aging unsold completed inventory, more than doubling, with over half aged >12 months. That's failed price discovery.
4. The OCF self-funding test fails. Land spend of ₹900 cr against ₹420 cr of internal surplus means ₹480 cr was debt-funded, and net D/E has duly crossed the 0.5× post-RERA line to 0.62× and is still rising.

Verdict: demand is softening while the company accelerates land purchases on debt, which is the pre-RERA failure pattern putting itself back together. Avoid it until collections re-track bookings and the balance sheet comes back below 0.5×. Each of these flags is one of the five real-estate red flags, plus the leverage corollary.

P9 (FFO/AFFO). A REIT reports: net income ₹640 cr; real-estate D&A ₹880 cr; gain on sale of a non-core asset ₹150 cr; straight-line rent (non-cash) ₹95 cr; maintenance capex ₹165 cr; TI + LC ₹120 cr; units 60 cr; distribution declared ₹18.9/unit. (a) FFO and AFFO (total and per unit). (b) AFFO payout and the health call. (c) The REIT's presentation excludes TI/LC from its self-defined "AFFO," claiming they're growth investments. Respond.

Solution. (a) FFO = 640 + 880 − 150 = ₹1,370 cr (₹22.8/unit). AFFO = 1,370 − 165 − 120 − 95 = ₹990 cr (₹16.5/unit). (b) Payout = 18.9 ÷ 16.5 = 114%, far beyond the >90% cut-risk line. The distribution exceeds post-maintenance cash generation, so borrowings, asset sales or return of capital are funding it. Expect a cut or a rising LTV. That yield on screen is fiction. (c) Reject the claim. TI and LC on re-leasing existing space are the recurring cost of keeping the rent flowing, because space doesn't re-let without them. Excluding them is the sector's improper-capitalization move. Growth TI on new development can be excluded, so demand the split and rebuild AFFO yourself. On the proper definition this REIT's payout is 114%. Its FFO-ish framing implies ~83% (18.9/22.8).

P10 (timed, 12 min: implied cap rate & the bond spread). A REIT has NOI of ₹2,150 cr on an appraised completed-portfolio value of ₹26,900 cr, holds ₹1,400 cr of other (non-income) assets, and has 55 cr units outstanding. Its gross debt is ₹8,700 cr against cash of ₹300 cr, and its price is ₹296 with a DPU of ₹19.2. The 10Y G-sec is at 7.1% and contracted escalations run ~4.5%/yr. Compute (a) the appraisal cap rate, (b) the implied cap rate at market price, (c) P/NAV, (d) distribution yield and spread vs G-sec and (e) LTV vs the SEBI cap. Then (f) write the two-sentence verdict.

Solution. (a) 2,150 ÷ 26,900 = 8.0%. (b) EV = 296 × 55 = 16,280 + net debt (8,700 − 300 = 8,400) = 24,680. Less other assets of 1,400, that leaves 23,280 attributable to income assets, so implied cap = 2,150 ÷ 23,280 = 9.2%. (c) NAV = 26,900 + 1,400 − 8,400 = ₹19,900 cr → ₹362/unit, and P/NAV = 296 ÷ 362 = 0.82. (d) Yield = 19.2 ÷ 296 = 6.5%, and spread = 6.5 − 7.1 = −60 bps. (e) LTV = 8,700 ÷ 28,300 = 30.7%, comfortable against the 49% cap (rating/approval tripwire at 25% already applicable). (f) Verdict: the market prices the assets ~13% below appraisal (9.2% vs 8.0% cap) while leverage is conservative, and 4.5% contracted growth covers the −60 bps bond spread ~7× over. So the pricing embeds pessimism about offices, and the thesis stands or falls on occupancy and re-leasing spreads. Your diligence goes there next.

P11 (timed, 12 min: cyclical cross-exam). A commodity producer is at what its CEO calls "a structurally higher plateau". Its trailing P/E is 4.8×, the sector's lowest. EBITDA/t is at a 12-year high of $840 against a 12-year average of $410. Net debt is flat at 1.1× trailing EBITDA. A debt-funded 40% capacity expansion has been announced, sector-wide capex is running at 2.3× depreciation, and three competitor IPOs were filed this quarter. List every red flag with its number from the commodity list, compute net debt/mid-cycle EBITDA, and give the verdict in two sentences.

Solution. Start with the flags. #5 is a low trailing P/E at the peak marketed as cheapness: 4.8× on EBITDA/t at 2× the 12-year average. #2 is debt-funded capacity expansion at the peak, the 40% project. Then there are capital-cycle top signals, namely sector capex at 2.3× depreciation and an IPO cluster, the euphoria-in-supply signature. And there's a leverage optic. 1.1× on trailing (peak) EBITDA restates to net debt ÷ mid-cycle EBITDA ≈ 1.1 × (840/410) ≈ 2.3×, already past the 2× line before the expansion debt arrives. Verdict: this is the full sell signature from the cyclical trap and the capital cycle. Peak spreads are being capitalized as permanent and leverage is rising into what the supply pipeline says is a downturn being built, and the CEO hasn't offered the cost-curve proof the "plateau" claim needs. Watch it, don't own it, and note how closely the pattern matches Corus/Novelis 2007.

P12 (timed, 12 min: the P&L lag, quantified). A developer recognized revenue of ₹2,800 cr this year at a 24% EBITDA margin. Bookings were ₹5,600 cr, up 22% yoy, at an embedded project EBITDA margin you estimate at 30%. An analyst writes: "EPS growth of 8% is pedestrian; at 24× P/E the stock is expensive. Sell." Attack or defend that in ≤150 words using the developer framework, with one computation.

Solution. Attack. Reported revenue and EPS describe FY-minus-3-to-4's sales, because Ind AS 115 recognizes this revenue at a point in time. Today's economics are in bookings: ₹5,600 cr growing 22% with ~30% embedded margins. Now look through. Embedded EBITDA being created ≈ 0.30 × 5,600 = ₹1,680 cr/yr, against recognized EBITDA of 0.24 × 2,800 = ₹672 cr. So the P&L is reporting 40% of the value creation actually occurring (672/1,680). As completions catch up with the booking wave, revenue and EPS will mechanically compound far above 8%. The right lenses are P/NAV, EV/bookings against embedded margins, collections tracking bookings (check that cash truth serum) and the OCF self-funding test. A trailing P/E on lagged revenue is this sector's version of the cyclical trap, an earnings-anchored multiple on an earnings number that doesn't measure the present. The verdict waits on those dials, but the analyst's stated reasoning is invalid.

P13 (guided, sales velocity & absorption). A developer launched a 0.90 msf tower at the start of the year, and by year-end it had sold 0.54 msf of it. Across all projects its unsold inventory is 3.6 msf against a trailing sales pace of 4.8 msf/yr. (a) Compute first-year sell-through on the new tower. (b) Compute quarters-to-sell (QTS) on total unsold inventory. (c) Read both against the healthy bars for developers, and state the one thing that would turn healthy velocity into a red flag.

Solution. (a) 0.54 ÷ 0.90 = 60%, a strong Indian launch inside the 40 to 60%+ band. (b) Quarterly pace = 4.8 ÷ 4 = 1.2 msf, so QTS = 3.6 ÷ 1.2 = 3.0 quarters, clean against the healthy <4 to 6. (c) Both dials are healthy and they reinforce each other, because velocity is pricing power's twin. The red flag would be a developer holding realization while velocity collapses, with QTS climbing and sell-through fading. That's choosing optics over cash, and it comes before the aging-inventory flag (real-estate red flag #3).

P14 (guided, REIT NOI, cap-rate NAV, P/NAV). An office REIT has property revenue of ₹3,600 cr and property-level operating expenses of ₹610 cr. Its appraisal cap rate is 8.25%, under-construction + land is ₹2,000 cr, net debt is ₹9,800 cr, and it has 78 cr units at a unit price of ₹340. Compute (a) NOI and NOI margin, (b) NAV/unit at the appraisal cap rate and (c) P/NAV.

Solution. (a) NOI = 3,600 − 610 = ₹2,990 cr, and margin = 2,990 ÷ 3,600 = 83.1%. That's squarely in the 80 to 85% Grade-A office band, so property-level costs aren't leaking. (b) Completed-portfolio value = NOI ÷ cap rate = 2,990 ÷ 0.0825 = ₹36,242 cr. NAV = 36,242 + 2,000 − 9,800 = ₹28,442 cr ÷ 78 = ₹364.6/unit. (c) P/NAV = 340 ÷ 364.6 = 0.93, a shallow discount. Combine it with the implied-cap-rate and yield-spread lenses from the Horizon workup before you conclude, since the whole NAV rests on the 8.25% appraisal cap holding.

P15 (cost curve & the currency lever). Two aluminium smelters are plotted on a USD cost curve. Smelter A (India) has a cash cost of ₹1,58,000/t. Smelter B (Europe) has a cash cost of $2,050/t. The rupee weakens from ₹80/$ to ₹88/$. Global USD cost-curve percentiles: 25th $1,850/t, 50th $2,000, 75th $2,200. (a) Give A's cash cost in USD before and after, and its quartile each time. (b) Say briefly what the currency move did to A's only moat, and what habit it forces on you.

Solution. (a) Before: 1,58,000 ÷ 80 = $1,975/t, between the 25th ($1,850) and 50th ($2,000) percentiles → 2nd quartile. After: 1,58,000 ÷ 88 = $1,795/t, below the 25th percentile → 1st quartile. B is unmoved at $2,050 (3rd quartile). (b) A weaker rupee mechanically slid A left on the USD curve. That's a currency tailwind, and operations didn't earn it. The rule from the KPI panel is to track cash cost in both USD and local currency, because a line flat-in-USD but "improving" only on FX isn't a real cost-position advantage, and it reverses the moment the rupee strengthens.

P16 (same-store NOI decomposition). An office REIT reports same-store NOI growth of +6.5%. Contractual escalations are +4.5%/yr, occupancy rose from 91% to 92%, and re-leasing spreads were +18% on the 6% of area re-let this year. Decompose the growth into its three engines and say whether it looks organically healthy.

Solution. Escalation ≈ +4.5% on the whole in-place base. Occupancy ≈ (92 − 91) ÷ 91 = +1.10%. Re-leasing ≈ +18% × 6% of area = +1.08%. Gross ≈ 4.5 + 1.10 + 1.08 = +6.68%. Against the reported +6.5%, the residual of ≈ −0.18% is re-let downtime or free-rent timing, the same "− downtime" plug as in the Horizon table. It's healthy. All three organic engines are positive and the growth is same-store, so debt-funded acquisitions aren't manufacturing it. That's the treadmill a headline-NOI number can hide.

P17 (developer NAV & its two cross-checks). Here are a developer's condensed NAV inputs, in ₹ cr. Ongoing + planned projects have a post-tax PV of net cash flows of 6,200. Sold-but-uncollected receivables net of cost-to-complete come to 1,400, and the land bank at realizable value is 2,100. An annuity office block with NOI of 150 is valued at an 8% cap rate. Net debt is 2,600, and the PV of corporate overhead is 800. There are 120 cr shares, the price is ₹58 and trailing-year bookings were ₹4,500 cr. Compute (a) NAV and NAV/share, (b) P/NAV and (c) the EV/pre-sales cross-check, and say what it implies at ~25% embedded PAT margins.

Solution. (a) Annuity block = 150 ÷ 0.08 = ₹1,875 cr. NAV = 6,200 + 1,400 + 2,100 + 1,875 − 2,600 − 800 = ₹8,175 cr → ₹68.1/share. (b) P/NAV = 58 ÷ 68.1 = 0.85, a modest discount inside the ~0.6 to 1.1× developer band. (c) EV = market cap + net debt = 58 × 120 + 2,600 = ₹9,560 cr, so EV/pre-sales = 9,560 ÷ 4,500 = 2.1×. At ~25% embedded PAT margins that's ≈ 2.1 ÷ 0.25 ≈ 8.5× look-through earnings on the current booking pace, which is undemanding. Both lenses agree the price embeds little growth. A shift in the panel, such as a collections slide or a failed OCF test, would break that read. A lagged P&L never would.

P18 (timed, 12 min: the aluminium peak, cross-examined). An aluminium major is mid-boom. It makes 1.5 mt/yr, and latest-year EBITDA/t is ₹72,000 against a 10-year average of ₹42,000. D&A is ₹2,000 cr, interest ₹1,100 cr, tax 25%, net debt ₹13,000 cr and shares 100 cr. The price is ₹360, and the mid-cycle EV/EBITDA is 5.5×. It has just announced a debt-funded 30% capacity expansion, and sector-wide capex runs at 2.1× depreciation. Compute (a) the trailing P/E, (b) mid-cycle value/share, (c) how far above mid-cycle value the price is, and (d) net debt on peak vs mid-cycle EBITDA. Then list the red flags by their commodity-list numbers and give a two-sentence verdict.

Solution. (a) Peak EBITDA = 1.5 × 72,000 = ₹10,800 cr. PAT = (10,800 − 2,000 − 1,100) × 0.75 = ₹5,775 cr, so EPS is ₹57.75 → trailing P/E = 360 ÷ 57.75 = 6.2×. That's the "cheapest in the sector" optic. (b) Mid-cycle EBITDA = 1.5 × 42,000 = ₹6,300 cr. EV = 5.5 × 6,300 = ₹34,650 cr, and equity = 34,650 − 13,000 = ₹21,650 cr → ₹216.5/share. (c) 360 ÷ 216.5 − 1 = +66% above mid-cycle value. (d) Net debt/peak EBITDA = 13,000 ÷ 10,800 = 1.2×, which looks comfortable. Net debt/mid-cycle EBITDA = 13,000 ÷ 6,300 = 2.06×, already past the <2× line before the expansion debt arrives. The flags are #5, a low trailing P/E at the peak sold as cheapness (6.2× on EBITDA/t at 1.7× its 10-year average), and #2, debt-funded capacity expansion at the peak. Add the capital-cycle top signal of sector capex >2× depreciation. Verdict: this is the full sell signature. Peak spreads are being capitalized as permanent while leverage climbs into a downturn the supply pipeline is scheduling. Watch it, don't own it, and re-read the Corus/Novelis-2007 note in the KPI panel.

Applied mini-project: the cyclical & yield workup (the module gate)

The task (≈5 hours, two parts, one real company each). This is the "Cyclical valuation exam" that gates your progress on the master map. You hand in your two completed panels plus verdicts, and they're graded against the rubric below.

Part A. Mid-cycle valuation of a commodity producer. Pick Tata Steel or Hindalco in India (or JSW Steel/Vedanta if you prefer), and keep a US/global shadow comparator in view (BHP for miners, Exxon if you chose energy). Pull 10 years of annual reports and investor presentations from the company's IR page plus BSE/NSE filings, and the 10-year financials from screener.in. For the shadow comparator, pull the latest 10-K/annual report from SEC EDGAR (Exxon's reserves tables, BHP's cost guidance). Then build, in a spreadsheet:

1. The 10-year series of volumes, realization vs benchmark, EBITDA/tonne (disclosed or derived) and net debt.
2. An inventory-gain note for any distorted year. Companies flag these in commentary.
3. Mid-cycle EBITDA on both mean and median, with your choice justified.
4. The six-step valuation with a 5 to 6× multiple band and current net debt.
5. The replacement-cost cross-check, EV per tonne of capacity against recent announced greenfield and brownfield costs.
6. A one-paragraph capital-cycle read (sector capex/depreciation, announced capacity, China overlay).
7. The five red flags checked, one line each.

End with a valuation range and a one-paragraph verdict against the current price. For Hindalco, value India smelting (a price-taker) and Novelis (a converter) separately, then add them up.

Part B. REIT distribution-quality panel. Pick Embassy Office Parks, Mindspace or Brookfield India REIT. Pull the latest annual report, the quarterly investor presentation and the NDCF/distribution statements. They're all on the REIT's website, and the unit price is on NSE. Then build:

1. The operating panel: occupancy (all three definitions if disclosed), WALE, same-store NOI growth, re-leasing spreads, escalation structure.
2. The cash walk from net income to FFO to your own AFFO to the payout ratio on AFFO, and the NDCF composition (the interest/dividend/return-of-capital split). State your maintenance capex + TI/LC assumptions if disclosure is partial.
3. Leverage: LTV vs the 49% cap, and the maturity schedule.
4. Valuation: NAV/unit (their appraisal, and your own with ±100 bps cap-rate sensitivity), P/NAV, implied cap rate at market price, and distribution yield vs the current 10Y G-sec.
5. The five real-estate red flags, one line each.

End with a one-screen table like Horizon's and a five-sentence memo verdict.

Scoring rubric (max 20; ≥16 to pass the gate):

File both workups in your knowledge system, because the KPI tables feed the sector cheat-sheet artifact. Both companies are teaching specimens here, and neither is a recommendation.

Case Lab: run case CL-5.6 in the app's Cases tab, a deep-cyclical producer taken across a full price cycle beside a REIT distribution-quality panel, so you practise calling mid-cycle value where the trailing P/E lies and stress-testing an AFFO payout end to end.

Reading & resources

Commodities & cyclicals

• Edward Chancellor (ed.), Capital Returns, Introduction + Ch. 1–2 ("Capital Cycle Revisited"): the supply-side lens everything above runs on. [Paid] [Intermediate]
• Aswath Damodaran, "Ups and Downs: Valuing Cyclical and Commodity Companies" (2009 working paper, free on damodaran.com/SSRN): the normalization playbook from the source. [Free] [Intermediate]
• McKinsey, Valuation (8e), cyclical-companies chapter: mid-cycle normalization with worked exhibits. [Paid] [Intermediate]
• Peter Lynch, One Up on Wall Street, Ch. 7 & the cyclicals passages in Ch. 17–18: the P/E trap in plain English. [Paid] [Beginner]
• Javier Blas & Jack Farchy, The World for Sale: how physical commodity markets actually clear; reads like a thriller. [Paid] [Beginner]
• World Steel Association (worldsteel.org) statistics; International Aluminium Institute (international-aluminium.org); Joint Plant Committee (India steel data); PPAC (petroleum data India); IEA reports: the free benchmark/supply datasets. [Free] [All levels]
• Company primary sources: Tata Steel & Hindalco investor presentations (EBITDA/t and cost disclosures), Coal India annual report (FSA vs e-auction split), BHP annual report + commodity outlook, Exxon 10-K reserves tables (SEC EDGAR), with the standardized-measure note especially. [Free] [Intermediate]
• Cost curves: CRU, Wood Mackenzie, AME. The professional sources (know they exist; broker notes often reproduce key curves). [Paid] [Advanced]

Real estate & REITs

• RERA, 2016: the Act's summary + your state RERA portal (MahaRERA is the deepest database; check any developer's projects independently). [Free] [Beginner]
• Ralph Block, Investing in REITs (4e), Ch. 1 to 3 (what REITs are) and Ch. 8 to 9 (FFO/AFFO, NAV): the sector's standard text (US-framed, so map it to India the way the FFO/AFFO walk above does). [Paid] [Beginner→Intermediate]
• Nareit (reit.com): the FFO White Paper and "REIT Basics", the definitional source for FFO and AFFO. [Free] [Beginner]
• SEBI (REIT) Regulations, 2014 (sebi.gov.in, consolidated version). Read Reg. 18 (distributions) and Reg. 20 (borrowings) yourself; regulations move, and this is the primary source. [Free] [Intermediate]
• Embassy Office Parks REIT annual report + quarterly deck + NDCF statements; Mindspace and Brookfield India equivalents: India's best-disclosed yield vehicles; also one US 10-K for contrast (Prologis or Realty Income). [Free] [Intermediate]
• Developer disclosure packs: DLF, Godrej Properties, Macrotech (Lodha), Oberoi, Prestige quarterly presentations, for bookings/collections/launches panels. US contrast: D.R. Horton or Lennar 10-K (orders, backlog, land position). [Free] [Intermediate]
• Market data: Anarock, Knight Frank India, JLL office reports (free summaries) for supply, absorption and vacancy by city. [Free] [Beginner]
• Green Street Advisors: the US gold standard for REIT NAV/cap-rate work (know it exists). [Paid] [Advanced]
• Zerodha Varsity, Fundamental Analysis module: the free India-context refresher if any accounting linkage feels shaky. [Free] [Beginner]
• Going deeper: Aswath Damodaran, cyclical-and-commodity valuation lecture + his industry EV/EBITDA, ROIC, and cap-rate datasets (pages.stern.nyu.edu → "Data"; the lecture is free on YouTube) [Free] [Intermediate]. Real cross-sector multiples and reinvestment numbers to calibrate the mid-cycle multiple band and the REIT cap-rate ranges against how the market actually prices cyclicals and property, instead of trusting a textbook 5 to 6×.
• Going deeper: Edward Chancellor (ed.), Capital Account: A Money Manager's Reports from a Turbulent Decade [Paid] [Intermediate]. The Marathon predecessor to the assigned Capital Returns: a decade of capital-cycle case studies (telecom, mining, energy) that drill the supply-side lens until reading a sector's capex-vs-depreciation ratio becomes reflexive.

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 six-step mid-cycle commodity valuation (§4.6) and the FFO -> AFFO -> NAV -> implied-cap-rate chain (§4.15-4.16) across a coverage list instead of one balance sheet at a time",
  "use": "An LLM extracts an 8-10 year EBITDA/tonne (or GRM) series from annual reports, or a REIT's FFO reconciliation, NOI, debt, and unit count from quarterly results, into a structured table in minutes instead of an afternoon of cross-referencing filings.",
  "tools": ["Claude/ChatGPT/Gemini to extract a multi-year driver series or an FFO/AFFO reconciliation from pasted annual-report or REIT-results excerpts", "a notebook (Python/pandas/numpy) to hold the series and compute mid-cycle EBITDA, NAV, and cap-rate sensitivity yourself", "screener.in's exported ten-year financials for the EBITDA/tonne reconstruction", "REIT quarterly investor decks (Embassy, Mindspace, Brookfield India, Nexus Select) for NOI/FFO/AFFO"],
  "prompt": "Paste 8-10 years of a commodity producer's EBITDA/tonne (or realized spread), or a REIT's NOI, FFO reconciliation, net debt, and unit count. Compute mid-cycle EBITDA (mean AND median), a mid-cycle EV/EBITDA valuation, or the FFO to AFFO to NAV chain and the implied cap rate at the current price. Flag if the mean and median mid-cycle estimates diverge by more than 10%, or if AFFO payout exceeds 90%.",
  "verify": "Recompute at least one year of the EBITDA/tonne series, or one quarter's FFO-to-AFFO bridge, from the raw disclosed numbers yourself before trusting the table -- AFFO has no standardized definition (§4.15), so an LLM will silently adopt whichever 'AFFO' a company self-reports without checking whether recurring maintenance capex was reclassified as growth capex.",
  "diy": "The mastery gate is unaided: run the six-step mid-cycle valuation, or build the FFO -> AFFO -> NAV chain and back out an implied cap rate from a market price, by hand -- no tool in the room. The tool above saves the transcription hour; it does not replace knowing why mid-cycle, not trailing, is the right anchor."
}

Modern Data Analysis

By hand first. The six-step Deccan Steel build and Horizon REIT's FFO→AFFO→NAV→implied-cap-rate chain are both hand arithmetic on one multi-year series or one quarter's disclosure.

Today's workflow. An analyst covering several cyclicals and REITs holds each name's full driver history (EBITDA/tonne, or NOI/FFO/AFFO) in one table and recomputes mid-cycle EBITDA, NAV, and implied cap rate the instant a new year of data lands, instead of re-running the six-step by hand each time. Sensitivity to the mid-cycle-multiple or cap-rate assumption becomes a parameter sweep instead of a single point estimate.

Tools & sources (IN + US). For India, the Ministry of Steel's Joint Plant Committee (JPC) publishes data on India steel prices and production. Exported ten-year financials from screener.in let you reconstruct an EBITDA/tonne series. SEBI REIT quarterly and half-yearly disclosures (the Embassy Office Parks, Mindspace, Brookfield India, Nexus Select investor decks) give you NOI, FFO and AFFO. For the policy rate and the G-sec yield that anchor the cap-rate spread, use FBIL's benchmark G-Sec yield curve (Financial Benchmarks India Pvt Ltd, fbil.org.in) and RBI Monetary Policy Committee resolutions (rbi.org.in). For the US, FRED series DGS10 (10-year Treasury yield) feeds the US REIT yield-spread gauge. Nareit's T-Tracker (reit.com, quarterly aggregated US REIT FFO/AFFO data) is the sector-level benchmark that the single-REIT worked examples here scale up to. SEC EDGAR's XBRL company-facts API has Prologis's and Realty Income's historical NOI and debt, and Damodaran's data library has cross-sector cap-rate and EV/EBITDA benchmarks.

# recomputed in-session -- R3
# Part 1: verification anchor -- Sec 4.6 Deccan Steel mid-cycle build, extended with mean vs median
import statistics
series = [9000, 6000, 8000, 12000, 18000, 26000, 13000, 8000]     # Rs/tonne EBITDA, 8 years
mean_ebitda_t, median_ebitda_t = sum(series)/len(series), statistics.median(series)
print(f"mean {mean_ebitda_t:.0f} Rs/t, median {median_ebitda_t:.0f} Rs/t, "
      f"gap {(mean_ebitda_t-median_ebitda_t)/mean_ebitda_t:.1%}")
# -> mean 12,500, median 10,500, an 16% gap driven entirely by the two blow-out years

mid_cycle_ebitda_cr = 5_000_000 * mean_ebitda_t / 1e7      # 5mt capacity, Rs/t -> Rs cr
ev = 5.5 * mid_cycle_ebitda_cr
equity_per_share = (ev - 12000) / 100
print(f"mid-cycle EBITDA Rs{mid_cycle_ebitda_cr:.0f}cr, EV Rs{ev:.0f}cr, value/share Rs{equity_per_share:.0f}")
# -> Rs6,250cr, Rs34,375cr, Rs224/share -- reproduces the module exactly

# Part 2: verification anchor -- Sec 4.15-4.16 Horizon REIT FFO/AFFO/NAV, extended with cap-rate sensitivity
ffo = 850 + 950 - 40
affo = ffo - 140 - 70 - 45 - 85
affo_per_unit = affo / 95
print(f"\nFFO Rs{ffo}cr, AFFO Rs{affo}cr, AFFO/unit Rs{affo_per_unit:.1f}, payout {13.0/affo_per_unit:.0%}")

noi = 3040
for cap_rate in (0.07, 0.08, 0.09):
    value = noi / cap_rate
    print(f"cap rate {cap_rate:.0%} -> completed-portfolio value Rs{value:,.0f}cr "
          f"({(value/38000-1):+.1%} vs the 8% base case)")
# -> -100bps: +14.3% NAV;  +100bps: -11.1% NAV -- convex, not linear (NAV = NOI/cap rate is a hyperbola)

Verify. Every figure reproduces the numbers stated above. For Deccan Steel that's the ₹224/share mid-cycle value, ₹6,875 cr/mtpa against replacement cost and 1.9× net debt/mid-cycle EBITDA. For Horizon it's the ₹14.9 AFFO/unit, 87% payout, ₹305/unit NAV, 0.70× P/NAV and the 10.3% implied cap rate against the appraiser's 8%. That last one was checked carefully, since cap rate is NOI ÷ value (never value ÷ NOI). No discrepancies turned up. Both extensions are new. The mean/median split shows that the "big peak years drag the mean up" caution is worth a real 16% in Deccan's case. The cap-rate sweep shows the NAV sensitivity is asymmetric, larger on the downside for a rate rise than the upside for an equivalent fall. Our prose above states that fact ("REIT prices fall mechanically") but doesn't quantify it.

Quantitative lens

Refuse a single-year ROIC or EBITDA print on a commodity producer. Use the mid-cycle average, and check it against the median. Deccan's own EBITDA/tonne ranges from 6,000 to 26,000 across eight years on the identical asset base, a 4.3× spread with no change in the business. So any ROIC or margin computed off one year tells you where that year was on the price curve. It's no verdict on the business. The median (₹10,500/t) is 16% below the mean (₹12,500/t), and that gap is information in itself. It says two blow-out years, more than the business's normal state, are doing most of the work in the average.

Refuse REIT or developer earnings. Use FFO/AFFO/NAV for REITs and pre-sales/collections/NAV for developers. Working from net income alone, you'd never see Horizon's AFFO of ₹14.9/unit. Ind AS depreciation charges a well-maintained office park as though it were a delivery truck, and so it understates the park's cash power. A developer's point-in-time revenue recognition hides the same thing from the other side of the ledger. Both fixes swap an accrual number that lies for a cash number that doesn't.

A cap rate behaves like a bond's duration, convexly, so don't treat it as a static yield. A 100bps rise in Horizon's cap rate cuts the completed portfolio's value by 11.1%. An equivalent 100bps fall raises it by 14.3%. The asymmetry comes from value = NOI/cap-rate being a hyperbola instead of a line. That's why REITs get called "long-duration yield assets" above. The sensitivity has the same shape as a bond's price-yield curve, and the fixed-income convexity toolkit applies almost unchanged.

Limits. Mid-cycle valuation assumes mean reversion, and the caveats in the valuation section already flag that it's sometimes wrong in level (China 2002-08, the 2021-22 energy repricing). No amount of code resolves whether a given cycle is one of those exceptions. A mean/median split or a cap-rate sensitivity table gives you a range conditional on the assumption. It can't forecast which scenario will play out. The discount rate for developer NAV (13 to 15%) and the cap rate for REIT NAV are themselves judgment calls the data can't make. A sensitivity sweep shows you the consequence of a choice. Making the right choice is still your job.

Do it in code: build the multi-year driver panel with G2/DA1.02, chart the mean-vs-median or cap-rate-sensitivity curve with G2/DA1.04, and read the cap-rate/NOI relationship through the same convexity lens taught in G1/FI1.02.

Where this goes next: galaxy cross-links

• G1 · AL1.02 Commodities & real assets: the direct extension of the cost-curve and mid-cycle framework above into a portfolio-level commodities allocation.
• G1 · FI1.02 Duration & convexity: the NAV-to-cap-rate sensitivity above is the same hyperbolic, convex relationship as a bond's price to its yield; the toolkit transfers almost unchanged.
• G1 · MS1.03 Central bank operations & repo: the cap-rate-to-bond-yield spread only becomes readable once you know what actually moves the policy rate and the G-sec curve it anchors.
• G2 · DA1.02 Pandas data wrangling: the multi-year EBITDA/tonne or NOI/FFO/AFFO panel above, built properly across a whole coverage list.
• G2 · DA1.04 EDA & visualization: charting a commodity's mid-cycle band or a REIT's NAV-sensitivity curve is how the "write the conclusion as a range" advice gets built instead of asserted.
• X1 · AI0.07 The AI-augmented analyst: the general version of the AI-Augment box above: what to delegate, what to verify, and why the mastery gate stays unaided.

Teach it back & journal

Feynman prompt. Your audience is a smart 15-year-old who knows what "price" and "profit" mean but has never heard the word cyclical. Explain why the steel stock with the lowest P/E in the whole market can be one of the fastest ways to lose half your money, while the one with a terrifying 90× P/E can be the better thing to buy. Use no jargon beyond price and earnings. An umbrella seller in monsoon against drought works well, and so does an ice-cream stand across the seasons. Then add one short paragraph that takes the same "the reported number is lying to you, and here's the timing reason why" idea to a property developer whose revenue just "crashed 40%." Keep the whole thing to a single page. To test it, hand it to someone who doesn't invest and see whether they can explain the inversion back to you in their own words. If they can't, you've hidden the mechanism behind a word, so rewrite it.

Journal reflection. Where in your own judgment do you anchor to a single trailing number, the way the market anchors to a peak-cycle P/E? Write for fifteen minutes about a time you, or a business or team you watched closely, mistook a cycle high for a permanent plateau. Reach into your own professional experience if it helps. Billing rates, headcount, attrition and hiring demand all run in cycles, and "we'll always be this busy" (or "this lean") is the peak-earnings extrapolation in another costume. Then answer two questions on paper. What would a deliberately mid-cycle version of that judgment have looked like at the time? And what was your balance-sheet equivalent, the reserve of cash, credibility, optionality or patience that lets you survive being early, that would have bought you the right to act on the truer read instead of the trailing one?

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