IT Services & Telecom

Phase 5 · Week 47 · ~20 hours

Where it sits. One of these two businesses earns 40–60%+ on the capital inside it. The other struggles to clear its cost of capital while printing a 50% EBITDA margin. Both "sell technology," both are giant index weights in India, and they are economic opposites, which is why the sixth of Phase 5's ten sector playbooks is also the sharpest contrast the phase offers. An IT-services company rents out skilled hours: almost no capital, cash pouring out, and a P/E multiple whose whole debate is how long can the growth last? A telecom operator rents out an installed network: staggering capital, and an EV/EBITDA multiple that is meaningless until you rebuild the EV with spectrum dues and leases inside it. Studying them side by side hammers home this phase's master rule, that the business model dictates the KPIs and the multiple, because here the same country, the same decade, and sometimes the same customer produce two completely different economic machines. The last playbook gave you the consumer machines; the capital cycle returns here wearing a telecom hard hat; M2's forensics reappear as unbilled receivables and phantom subscribers. On the master map's five-questions spine, IT services is mostly question 2 and question 5: is it a good business, and what is the price assuming? Telecom is question 2 asked the hard way, and question 4, what can kill it?, asked constantly.

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

App link. Two wings of the sector case engine run on the work below: the constant-currency calculator drill (reported growth appears with currency mixes; you strip the FX and grade the real demand), the pyramid-margin builder (assemble an IT delivery unit's P&L from mix, rates, and utilization), the book-to-bill signal reader (TCV headlines with duration traps seeded in), the ARPU bridge drill (revenue and subscriber panels; you compute ARPU, churn, and market-share quality), and the EV rebuilder (telecom EVs with spectrum, AGR, and lease liabilities hidden in the notes). Your completed KPI tables seed the IT-services and telecom pages of the sector KPI cheat-sheet artifact. Passing the mastery check ("Sector exam ≥85%") unlocks Pharma & Healthcare.

Learning objectives

By the end of the week you can:

1. Explain the people-leverage model from first principles: revenue = billable people × utilization × hours × realized rate; why the pyramid (few seniors, many juniors) and the offshore/onsite mix are where the margin actually lives; and why the model's growth has historically been linear in headcount.
2. Compute constant-currency revenue growth from reported growth and a currency mix, and explain why an Indian IT company's INR growth, USD growth, and CC growth are three different numbers, only the last of which measures demand.
3. Run the full IT-services KPI panel with thresholds: CC growth, EBIT margin (~20–26% for tier-1 India), TCV and book-to-bill (~1.0–1.2×+), attrition (>20% = pressure), utilization (82–87%), headcount net adds (the leading demand signal), revenue per employee, client/vertical concentration, DSO with unbilled revenue, offshore/onsite mix, and discretionary vs cost-takeout demand, computing each and judging it against its band.
4. Work the growth–margin–pyramid triangle: show numerically how hiring freezes broaden margins today by narrowing the pyramid's base, and why that mortgages both future growth and future margin; recognize "margin held only via lower net adds" as harvesting, not efficiency.
5. Assess the GenAI question evenhandedly: state the strongest bear case (productivity deflates billable hours) and the strongest bull case (every prior efficiency wave expanded the work), and name the specific indicators in the filings that will settle it.
6. Explain why telecom returns are structurally hard: the capex treadmill (generational rebuilds before the prior generation pays back), spectrum auctions as upfront scarcity-rent extraction, and the arithmetic by which a 50% EBITDA margin coexists with ROIC below WACC. Then why consolidation to three players (India post-2016, the US) is the condition for repair.
7. Run the telecom KPI panel with thresholds: ARPU (computed correctly off average subscribers), subscribers and net adds, churn (<2.5%/month healthy), data GB/user and MOU, subscriber vs revenue market share, AGR and the legacy of the Indian AGR dispute, net debt/EBITDA (<3× on the inclusive definition), capex and spectrum intensity, and VLR/active-SIM %. Then detect a phantom subscriber base.
8. Apply the right valuation lens to each: P/E for IT services (because earnings ≈ free cash flow) with the growth-durability question made explicit via the key-value-driver formula; EV/EBITDA for telecom with the lease and spectrum adjustments done properly and cross-checked against EV/(EBITDA − capex); then run both sectors' red-flag panels from the brief.

Prerequisites & connections

Builds on. M1 (revenue recognition: Ind AS 115/ASC 606 over-time recognition is why IT firms carry unbilled revenue/contract assets on fixed-price work; Ind AS 116/ASC 842 leases, which sit inside every telecom EBITDA you will compare; intangible assets, where spectrum is a purchased intangible and the US/India difference in amortizing it moves EBIT). M2.02–M2.04 (DSO, working capital, DuPont: the IT DSO-plus-unbilled check and the telecom ROIC decomposition are direct applications). M2.05–M2.07 (forensics: unbilled ballooning, capitalizing network opex, phantom subscribers). M3.04–M3.07 (EV construction, where telecom is the hardest EV build in the market; the key-value-driver formula for reversing the IT P/E; reverse DCF). M4.03–M4.06 (five forces on an undifferentiated product; switching costs as the IT moat and efficient scale as the telecom moat; the capital cycle, of which telecom's 2016–2021 India consolidation is the textbook case). M5.04 (contrast: SaaS sells software at ~zero marginal cost; IT services sells hours at fully-loaded human cost, and the margin structures follow).

Feeds into. M5.07–M5.10 (each playbook sharpens the right-lens reflex; the M5.10 capstone mixes IT and telecom cases into the right-lens exam). Phase 7 (currency: the INR's structural slide is a tailwind for IT exporters and a headwind for dollar-capex telecom). Phase 8 (the 1–2 hour teardown of an IT major runs on the panel below; a telecom teardown without the EV rebuild is worthless). The AGR section is also your first full case of regulatory definition risk, a theme that returns in banks, pharma, and utilities.

Core teaching content

4.1 Two ways to sell technology

Strip both businesses to their skeletons first.

An IT-services company hires an engineer in Pune for ₹12–15 lakh a year, trains her, and bills her time to a bank in New York at $25–40 an hour. Multiply by ~600,000 employees and you have TCS. The capital required is desks, laptops, and some buildings. The balance sheet is mostly cash the business couldn't find a use for. The constraint is people: finding them, training them, keeping them billable, and keeping them at all.

A telecom operator spends ₹2–3 lakh crore over a decade on spectrum, towers, fiber, and radios before the network is fully useful, then sells access to it for ₹200 a month per person. The marginal cost of one more subscriber is nearly zero. The average cost is enormous. Here the constraint is capital: raising it, deploying it, and earning anything on it before the next technology generation demands it all again.

Everything in the two playbooks below follows from the second and third rows of that table. In IT services every unit of growth is another person, so it comes at a high marginal cost, and the business ends up margin-stable but growth-expensive. Telecom has near-zero marginal cost on a huge fixed base. That makes it prone to price wars and hungry for capital.

Learn the physics first, and the dials will make sense.

4.4 The two playbooks in one frame: and the first hour

Set the pair side by side one last time, because together they teach the phase's deepest lesson.

IT services is capital-light and people-heavy, with ROIC ~45%+. Its value comes down to growth durability, so the lens is P/E. The killer questions are is demand rolling over? (CC growth, net adds, book-to-bill) and is the model's linearity breaking? (rev/employee). Telecom is capital-crushing, with ROIC hugging WACC on a good day. Its value = structure + leverage + treadmill position, so the lens is adjusted EV/EBITDA. Here the killer questions are who really pays what? (ARPU × mix × VLR) and can the balance sheet outlast the treadmill? (inclusive leverage, capex vs peers).

Same country, same index, opposite physics. So "what multiple does this sector get?" is never a memorization question. You always get the answer out of the model's economics.

First hour on an IT-services company (Phase 8 preview). Open the quarterly factsheet and pull the CC growth trend for 8 quarters. Set net adds against growth, then build the EBIT margin bridge and ask what held it. Check utilization and attrition against their bands. Read TCV and book-to-bill with the definitions checked, then the BFSI/top-client note, then DSO including unbilled. Finish with the rev/employee trend and P/E against implied g.

First hour on a telecom. Start with the subscriber and ARPU table, VLR included, then churn, then revenue share against subscriber share from TRAI. Find the lease boundary in EBITDA and rebuild EV with spectrum + AGR + leases. Next comes inclusive net debt/EBITDA, then capex intensity against peers, then EV/EBITDA and EV/(EBITDA − capex). Close on the contingent-liabilities note.

Common mistakes & how experts think differently

1. Quoting INR (or even USD) revenue growth as demand. The rupee's structural slide manufactures 3–4pp of INR "growth" in an average year, and cross-currency moves distort the dollar number. Experts quote CC growth only, and rebuild it when the disclosure looks convenient.

2. Treating utilization as a margin dial that goes to 100. The bench is growth inventory, the capacity to staff the next deal, and it carries training throughput too. Calling it waste misses both. An expert seeing 88–89% utilization doesn't cheer the margin. She asks which deal the firm won't be able to staff next quarter, and whether the margin "beat" was borrowed from that very deal.

3. Cheering headcount cuts as "efficiency" at an IT firm. People are the productive asset, and net adds are the demand forecast management actually believes. Falling headcount with decelerating CC growth is the master red flag, however much it looks like cost control. An expert entertains one exception: rev/employee inflecting upward through the cuts, which is the GenAI productivity signature. Even then she wants several quarters of it before she believes it.

4. Worshipping TCV headlines. Book-to-bill of 1.2× means nothing if deal duration stretched from 3 to 5 years, so annualize it. It means nothing if the book is mostly renewals, which bring no new revenue. And it means nothing if the definition changed, because Infosys's large-deal TCV isn't TCS's total TCV. Experts convert to ACV, ask for the renewal split, and watch conversion into revenue 2–4 quarters out.

5. Reading telecom EBITDA margin as business quality. A 50% margin can coexist with 7% ROIC, because the capital shows up below the EBITDA line. Experts run the "great margins, poor returns" ROIC arithmetic first. To them EBITDA is a financing-and-multiple convention. It doesn't hand down a verdict.

6. Comparing EBITDA across the lease boundary. Post-2019 Indian EBITDA (Ind AS 116) vs US EBITDA (ASC 842) vs pre-2019 history are three different animals. Experts fix the pairing rule (reported EBITDA with leases in EV, or EBITDAaL with leases out) and re-do any time series that crosses FY2019–20 in India.

7. Reading ARPU without the base. Rising ARPU on a collapsing subscriber base is arithmetic. Pricing power has nothing to do with it. Flat ARPU after a 20% tariff hike is down-trading, however stable it looks. Experts read ARPU × net adds × VLR as one joint signal and never read ARPU alone.

8. Treating spectrum dues, AGR installments, and leases as "not really debt." They carry interest, they have schedules, they rank ahead of equity, and in 2019 one court order turned a contingency into ₹1.4 lakh crore of them. Experts compute leverage on the inclusive definition and size every regulatory contingency against net worth.

9. Extrapolating the GenAI answer, either way. The bear who "knows" hours-based pricing is doomed and the bull who "knows" every efficiency wave expands demand are both trading a coin flip on conviction. Experts hold the indicator set (rev/employee inflection, net-adds decoupling, pricing-model mix, renewal economics) and update it quarterly. The market's implied answer is already in the multiple. Your edge comes from grading it against the evidence.

10. Using one habitual multiple for both sectors. P/E works for IT because earnings ≈ FCF (no leverage, no capex, full payout). It fails for telecom because D&A policy and leverage swamp EPS. EV/EBITDA fails for IT because it ignores the cash pile and flatters nothing that matters. The multiple is downstream of the model, and Phase 5 is built around that idea.

Worked examples

Worked example 1: Kaveri Technologies: the full IT panel read (is the demand rolling over?)

Kaveri Technologies (synthetic tier-1 exporter, reports in USD) prints its Q4. Work the panel top to bottom with every number shown, and do the arithmetic alongside.

Step 1: strip the currency. Revenue is $2,120m against $2,000m a year ago, so reported growth is +6.0%. The $2,120m splits into USD $1,166m (55%), EUR $530m (25%), GBP $254.4m (12%), and others $169.6m (8%, rates flat). Over the year the EUR appreciated 4% against the USD and the GBP depreciated 2.5%. Restate at prior-year rates. EUR-billed = 530/1.04 = $509.6m, and GBP-billed = 254.4/0.975 = $260.9m. CC revenue = 1,166 + 509.6 + 260.9 + 169.6 = $2,106.1m, so CC growth = 2,106.1/2,000 − 1 = +5.3%. This time currency flattered the headline, by 0.7pp. Now log the CC trend across the last four quarters as disclosed: +9.2% → +7.5% → +6.4% → +5.3%.

Demand is decelerating, smoothly and unmistakably.

Step 2: the leading signal. Headcount has gone from 254,000 a year ago to 249,500 now, with net adds of −2,100 and −1,800 in the last two quarters. Management's own hiring engine is telling you it doesn't expect the deceleration to reverse soon.

Step 3: how is the margin being held? EBIT margin went from 23.0% to 23.6% (+60bps) through the slowdown. The HR table explains it. Utilization (ex-trainees) rose from 84.0% to 87.5%, at the very top of the 82–87% band. Attrition went from 14.2% to 16.8%, and it's still rising. Management is defending the margin by sweating the bench and not replacing leavers. Both levers run out, and the second is already starting to fight back. Rising attrition on a stretched workforce is how delivery accidents happen.

Step 4: the order book, de-spun. LTM TCV of $9.8bn on LTM revenue of $8.3bn gives a headline book-to-bill of 1.18×, which looks great. But disclosed average deal duration rose from 3.2 to 4.4 years. Annualize it. ACV = 9.8/4.4 = $2.23bn, against 8.6/3.2 = $2.69bn the prior year. So new annual business is down ~17% while the headline ratio "improved." Duration is doing the flattering through mega-deal cost-takeout signings. That fits clients cutting discretionary spend.

Step 5: hygiene. DSO including unbilled is 71 days and stable, so there's no revenue-quality issue. Top-10 clients are 21% and BFSI 32%. That's the usual concentration, and now it's a cyclical liability.

Verdict. Three master flags fire together. CC is decelerating while headcount falls. The margin is held by stretching utilization and cutting net adds. And book-to-bill quality is decaying under a strong headline. Nothing here says fraud. Everything says demand rollover being managed for optics. The one question to answer next (a Phase 8 habit): is Kaveri losing share, or is this the industry cycle? Pull two peers' CC growth and net adds. If theirs look the same, it's the cycle. Then the debate moves to the stock's P/E. The business itself isn't what's in question. If peers are growing, Kaveri has a company problem.

Worked example 2: Deccan Digital: the margin bridge and the triangle

Deccan Digital (synthetic) reports EBIT margin up from 21.0% to 23.4% (+240bps) in a slowing year. CC growth fell from 8% to 3%, net adds were −3,200, fresher hiring was paused. The investor deck calls it "operational excellence." Rebuild the bridge and see for yourself:

Now project the freeze forward with the cost-index arithmetic from the triangle. A 55/30/15 mix (junior/mid/senior, cost indices 1.0/2.2/4.0) has an average cost of 1.81. Two years without fresher intake ages it to 45/35/20, and the average rises to 2.02. That's +11.6% cost per head with zero wage inflation, roughly a 300–400bps future gross-margin headwind. And it arrives just when the firm must also rehire at market rates to serve returning demand.

Verdict. Of the +240bps, ~340bps came from levers that run out or that management doesn't control (utilization, currency), against ~−100bps of underlying drift. It's the "margin held only via lower net adds" pattern with its full mechanism on show. The print is real. The quality is poor. And the triangle guarantees the bill arrives in years 2–3. Value this company on a normalized margin of 21–22%, because the 23.4% print overstates it. Write down why in your journal before the market does it for you.

Worked example 3: IndusWave Telecom: the forensic rebuild (the death-spiral checklist)

IndusWave is a synthetic India #3 operator, modeled on the distressed-operator pattern and purely illustrative. It reports a quarter, and the press release headlines "ARPU up 3%, EBITDA margin 39%, net debt/EBITDA a comfortable 2.3×." Rebuild everything.

Step 1: ARPU, done properly. Quarterly mobile-services revenue is ₹11,700 cr. Subscribers went from 246m opening to 238m closing, an average of 242m. ARPU = 117,000m ÷ (242m × 3) = 117,000/726 = ₹161/month. A year earlier, revenue was ₹12,300 cr on average subs of 262m, so ARPU = 123,000/786 = ₹156.5. ARPU comes out +3.0%, in the year the industry took a ~20% tariff hike. Flag: the hike is not sticking. The base is down-trading to cheaper packs as fast as prices rise.

Step 2: the base itself. Net adds were −8m this quarter, from gross adds of 22m and disconnections of 30m. Churn = 30/726 = 4.1%/month, above the 4% crisis line. Then the TRAI VLR print shows only 84% active, against 99%+ for peers. The real base ≈ 238 × 0.84 = 200m.

Sixteen percent of the "subscribers" are dead SIMs.

Step 3: share, quality-adjusted. Industry quarterly revenue is ₹63,000 cr and industry subs are 905m. IndusWave's subscriber share = 238/905 = 26.3%, while its revenue share = 11,700/63,000 = 18.6%. With revenue share far below subscriber share, this is a low-value base, monetizing its share at 0.71× the market rate. The premium operators' revenue shares exceed their subscriber shares. Every rupee of industry pricing repair flows disproportionately to them as a result.

Step 4: the balance sheet, inclusive. Market cap is ₹95,000 cr. Conventional net debt is ₹42,000 cr, deferred spectrum payables ₹1,05,000 cr, AGR dues ₹52,000 cr, lease liabilities ₹28,000 cr. LTM EBITDA is ₹18,500 cr (post-Ind AS 116). The press release's 2.3× is just 42,000/18,500. Count everything instead: (42,000 + 105,000 + 52,000 + 28,000) = ₹2,27,000 cr → 12.3× EBITDA. Naive EV/EBITDA = (95,000 + 42,000)/18,500 = 7.4×. True EV = 95,000 + 227,000 = ₹3,22,000 cr, which gives 17.4×. That's more than double the screener number, for a shrinking business.

Step 5: the treadmill. Capex intensity is 11% of revenue, against ~26% for peers. The network-quality gap that created 4.1% churn isn't being closed. Capex is the only discretionary line left after interest and installments, so it's the line that gets raided.

Verdict: run the death-spiral checklist. Dues ↑ → capex ↓ → network gap → churn >4% + port-outs → revenue share ↓ → EBITDA ↓ → leverage ↑ → repeat. Every box ticks. The 39% EBITDA margin and the "2.3×" leverage were true statements about a fictional company. So what is the equity at ₹95,000 cr? It's a leveraged option on (a) further tariff hikes, (b) government relief/conversion, (c) capital raises that dilute but fund capex. Nobody should mistake it for a claim on a self-sustaining business. Analyze it as an option, or pass. This is education. It isn't advice. But the method is non-negotiable: in telecom, nothing above the EV rebuild is real.

Worked example 4: Summit Wireless: what a US subscriber is worth (ARPU, ARPA, and lifetime value)

Summit Wireless is a synthetic US #3 postpaid operator that reports in USD. Take one quarter of its postpaid phone book. Mobile-services revenue is $10,080m, average phone lines are 70.0m, average accounts are 28.0m. Monthly postpaid phone churn is 0.90%. Service contribution margin, the share of ARPU that survives network opex before capex, is 40%. Blended cost to acquire one gross add is $650, and the cost of capital is 9%. Value the subscriber, and mind the denominator.

Step 1: ARPU, then ARPA, and never the twain. ARPU = service revenue ÷ average lines ÷ months = 10,080 ÷ (70 × 3) = $48.00/month. ARPA = service revenue ÷ average accounts ÷ months = 10,080 ÷ (28 × 3) = $120.00/month. The account holds 2.5 lines, so ARPA = 2.5 × ARPU. The telecom KPI panel flagged this multiple. Quote ARPA against an Indian operator's per-subscriber ARPU and you're wrong by 2.5× before the currency even enters.

Step 2: how long does the subscriber stay? Average lifetime = 1 ÷ monthly churn = 1 ÷ 0.009 = 111.1 months ≈ 9.3 years. The contract-and-bundle stickiness of the mature US structure, with postpaid churn of ~0.9%, is what makes the customer valuable. At India's ~2.5% prepaid churn the same formula gives a 40-month life, as P15 shows.

Step 3: lifetime value, first rough, then discounted. Monthly contribution = ARPU × margin = 48 × 0.40 = $19.20. The quick, undiscounted LTV treats the customer as a decaying annuity: contribution ÷ churn = 19.20 ÷ 0.009 = $2,133. Now discount it properly. At 9% a year that's 0.75% a month, and LTV = contribution ÷ (churn + monthly discount) = 19.20 ÷ (0.009 + 0.0075) = $1,164. Discounting nearly halves the number, because dollars nine years out are worth far less than the raw sum pretends.

Step 4: worth acquiring? LTV ÷ CAC = 1,164 ÷ 650 = 1.79× discounted, or 3.28× undiscounted. Payback = CAC ÷ monthly contribution = 650 ÷ 19.20 = 33.9 months ≈ 2.8 years. Yes, the customer is worth acquiring. Now watch the trap that inflates it. Use ARPA ($120) where ARPU belongs and monthly contribution "becomes" $48. LTV balloons to $5,333, and every ratio triples. That one denominator error separates a sober number from a pitch deck.

Verdict. The subscriber economics are fine: a 2.8-year payback on a 9-year customer. Yet this is the "good" mature-market structure, and its returns are at best adequate, as the US comparison showed. LTV/CAC values the customer. It says nothing about the business. The spectrum, towers, fiber, and 5G densification the subscriber rides on are the capital that must clear the WACC, and the "great margins, poor returns" arithmetic shows it frequently doesn't.

So here's the telecom paradox in one company: a healthy customer P&L stacked on a value-destroying capital base. Never let a strong LTV/CAC talk you out of running the ROIC.

Worked example 5: Sutlej Systems: who keeps the GenAI productivity (realized rate and the pass-through fork)

Sutlej Systems is a synthetic Indian IT-services vendor. One of its delivery pods has 200 billable engineers, a standard year of 2,000 hours, utilization of 84%, a blended realized rate of ₹2,400/hour, direct delivery cost of ₹50 crore (₹25 lakh per engineer), SG&A at 13% of revenue. Build the pod, then renew it into the GenAI era.

Step 1: the pod today. Billed hours = 2,000 × 0.84 = 1,680 per engineer, or 336,000 across the pod. Revenue = 336,000 × ₹2,400 = ₹80.64 crore. Gross profit = 80.64 − 50 = ₹30.64 cr, a gross margin of 38.0%. SG&A = 13% × 80.64 = ₹10.48 cr, which leaves EBIT of ₹20.16 cr, a 25.0% margin, the top of the tier-1 band. Revenue per employee = 80.64 ÷ 200 = ₹40.32 lakh. At ₹83 that converts to ~$48.6k, the India-heritage signature.

Step 2: the renewal, and the fork. At renewal the client points at GenAI tooling and demands the same scope for 10% less. Contract value falls to 80.64 × 0.90 = ₹72.576 crore. Now everything turns on who actually captured the productivity, and the two possible answers are a chasm apart.

Step 3, Case A: the client captures it. Sutlej hasn't changed how it delivers, so it still runs 200 engineers at ₹50 cr of cost. Only the price fell. SG&A = 13% × 72.576 = ₹9.43 cr, so EBIT = 72.576 − 50 − 9.43 = ₹13.14 cr, an 18.1% margin. A 10% price pass-through with no cost response wiped out ~690bps of EBIT margin. That's the bear case, in one pod.

Step 4, Case B: the vendor captures it. This time the tooling really does let the pod deliver the same outcome with 170 engineers (−15%), so direct cost falls to 170 × ₹25 lakh = ₹42.5 cr. SG&A is still ₹9.43 cr, and EBIT = 72.576 − 42.5 − 9.43 = ₹20.64 cr, a 28.4% margin. Margin ends up higher than before the cut, because the 30 freed engineers redeploy onto new work (Jevons in services). Same 10% price cut, and a 10.3pp swing in EBIT margin between the two worlds.

Verdict. The productivity gain itself is neutral. The split is everything, and the contract sets the split. Time-and-materials or hours-benchmarked pricing hands the gain to the client, as in Case A. Fixed-price and outcome-based contracts let the vendor keep it, as in Case B. The split explains why the GenAI indicator set puts pricing-model mix and renewal economics beside revenue-per-employee.

When an IT major asserts "AI is margin-accretive," use this pod as the test. Is rev/employee inflecting up? Then Case B is real. Or is it flat-to-down while the disclosures still say T&M? Then you're being sold Case B while living Case A.

Practice set

All companies are synthetic unless noted, and figures are in ₹ crore unless stated. Work each one before you read the solution.

P1 (guided): The INR illusion. Sharada InfoTech reports INR revenue growth of +13.2%. The average INR/USD rate moved from 83.0 to 86.5 over the year. (a) Compute USD revenue growth. (b) Is that the demand signal? One sentence.

Solution. (a) USD growth = 1.132 × (83.0/86.5) − 1 = 1.132 × 0.9595 − 1 = +8.6%. Nearly 5 points of the INR headline was rupee depreciation. (b) It's closer, but it still isn't demand. USD growth remains distorted by EUR/GBP cross-currency moves, and only constant-currency growth (revenue restated at prior-period average rates) measures what clients actually bought. Quote CC or nothing.

P2 (guided): CC growth, two currencies. Current-year revenue is $1,040m: USD-billed $624m, EUR-billed $312m, GBP-billed $104m (as translated). Prior year: $1,000m. Over the period the EUR depreciated 6% against the USD, and the GBP appreciated 3%. Compute reported and CC growth.

Solution. Reported = 1,040/1,000 − 1 = +4.0%. Restate at prior rates. EUR 312/0.94 = 331.9 and GBP 104/1.03 = 101.0, so CC revenue = 624 + 331.9 + 101.0 = $1,056.9m and CC growth = +5.7%. The falling euro hid 1.7pp of real demand. The mechanical rule is to divide each currency bucket by (1 + that currency's move vs the reporting currency). In plain terms, you value the bucket at last year's rate.

P3 (guided): The revenue machine and the utilization lever. Vindhya Systems: 1,000 billable FTEs, standard year 1,840 hours, utilization 84%, blended realized rate $31/hr, blended direct cost $29,500 per FTE. (a) Compute revenue and gross margin. (b) Recompute at 80% utilization. (c) State the sensitivity rule.

Solution. (a) Billed hours = 1,840 × 0.84 = 1,545.6 per FTE, so revenue = 1,000 × 1,545.6 × 31 = $47.9m. Direct cost is $29.5m, which gives gross margin = 18.4/47.9 = 38.4%. (b) At 80%, hours drop to 1,472 and revenue to $45.6m (−4.8%). Costs don't change, because the bench is paid regardless, so gross margin = 16.1/45.6 = 35.4%. (c) 4 points of utilization moved gross margin 3.0pp. That's ~0.7–0.8pp of margin per point of utilization. It makes utilization both the fastest margin-defense lever and the first place to look when a margin "beat" needs explaining.

P4 (guided): Pyramid aging, quantified. Cost indices: junior 1.0, mid 2.2, senior 4.0. Mix today: 55% junior / 30% mid / 15% senior. A two-year hiring freeze ages the mix to 45/35/20. (a) Compute average cost per head, both dates. (b) If revenue per head rises only 4% (seniors bill somewhat more), what happens to a 40% gross margin? (Assume baseline revenue/head 100, cost/head 60.)

Solution. (a) Today: 0.55×1.0 + 0.30×2.2 + 0.15×4.0 = 1.81. After: 0.45 + 0.77 + 0.80 = 2.02. That puts cost per head up +11.6% with zero wage inflation. (b) Revenue/head goes 100 → 104 and cost/head 60 → 67.0 (+11.6%), so gross margin = (104 − 67)/104 = 35.6%, down ~4.4pp from 40%. It's the triangle's deferred bill. Margin "saved" during the freeze is repaid later through a structurally costlier pyramid, usually just as demand returns and lateral hiring gets expensive too.

P5 (independent): Book-to-bill, de-spun. Godavari Tech signs LTM TCV of $11.5bn against LTM revenue of $10.0bn. Last year it signed $9.9bn on revenue of $9.5bn. Average disclosed deal duration rose from 3.1 to 4.6 years. Compute both years' book-to-bill and ACV, and give the verdict.

Solution. Book-to-bill is 11.5/10.0 = 1.15× now, against 9.9/9.5 = 1.04× last year, so the headline is improving. ACV = TCV/duration, which gives 11.5/4.6 = $2.50bn now and 9.9/3.1 = $3.19bn last year. So annualized new business fell ~22% while the headline "improved." Duration stretch from mega-deal, cost-takeout signings is manufacturing the optics. Expect weak revenue conversion over the next 2–4 quarters, because demand mix has shifted defensive. Rising book-to-bill with falling ACV is one of the sector's most reliable spin patterns.

Whatever the headline says, always annualize.

P6 (independent): The toxic pair, costed. Narmada Digital has 250,000 employees. LTM voluntary attrition rises from 15% to 23%, utilization falls from 84% to 79%, and subcontractor costs rise from 7.2% to 9.0% of revenue (revenue ₹80,000 cr). (a) Estimate the annual backfill cost if each exit costs ~₹4 lakh in recruiting, training, and ramp. (b) What does attrition rising while utilization falls tell you that neither number alone does? (c) Where does the subcontractor line fit?

Solution. (a) Exits = 23% × 250,000 = 57,500, so backfill cost ≈ 57,500 × ₹4,00,000 = ₹2,300 cr ≈ 2.9% of revenue. That's a ~290bps drag, scattered across cost lines. (b) Each number alone has an innocent explanation. Together they're damning. People are leaving while the bench grows. The firm is losing the people it needs as hot skills walk out, while it holds on to people it can't bill because their skills don't match. You're looking at a demand-and-relevance problem, often with delivery escalations underneath. Healthy firms show the opposite correlation, where high demand puts pressure on utilization and retention at once. (c) Subcontractors at +180bps of revenue are the P&L shadow of attrition. The firm is buying delivery capacity at retail to protect committed projects.

The three numbers tell one story.

P7 (independent): Margin quality audit. Tapti Consulting holds EBIT margin at 23.5% (up from 22.0%) while CC growth decelerates 8% → 3% and net adds run negative. The disclosed bridge: utilization +80bps (85.0 → 86.2%), currency +60bps, pyramid (hiring freeze) +70bps, pricing +30bps, wage inflation −90bps. (a) Verify the bridge sums. (b) Sort the levers into earned/exogenous/borrowed. (c) Name the red flag and state what normalized margin you'd underwrite.

Solution. (a) 80 + 60 + 70 + 30 − 90 = +150bps, which matches the move from 22.0% to 23.5%. (b) Earned: pricing +30. Exogenous: currency +60, which isn't management's to keep. Borrowed: utilization +80 (band-top, and it reverses when growth returns) and pyramid +70 (repaid with interest, as P4 showed). (c) It's "margin held only via lower net adds," the harvesting flag, with CC deceleration alongside. Underwrite ~22%. That strips out the borrowed 150bps and keeps pricing. Treat any model extrapolating 23.5% as optimistic for the very quarters when rehiring costs will land.

P8 (independent): The P/E as a testable claim. An IT major: core ROIC 45%, cost of equity 11.5% (INR). (a) Fair steady-state P/E at g = 7% and g = 4.5% (key-value-driver formula, on NOPAT). (b) The stock trades at 26×: solve for implied perpetual growth. (c) One sentence connecting (b) to the GenAI indicator set.

Solution. (a) (1 − g/ROIC)/(r − g): g 7% → (1 − 0.156)/0.045 = ~19×; g 4.5% → 0.90/0.07 = ~13×. (b) 26 = (1 − g/0.45)/(0.115 − g) → 2.99 − 26g = 1 − 2.222g → g = 1.99/23.778 = ~8.4%. (c) At 26× the market is asserting ~8%+ growth forever. It's saying the current 3–5% CC prints are cyclical and that GenAI lands on the bull side and expands the work. Your quarterly read of rev/employee, net adds, and pricing-model mix is the running audit of that claim. This is single-stage teaching math, so take the direction and the method from it.

P9 (independent): ARPU and churn, computed. Krishna Mobile reports quarterly mobile-services revenue of ₹16,240 cr. Subscribers went from 302m opening to 308m closing, with gross adds of 21m. (a) ARPU. (b) Monthly churn, against the threshold. (c) Next year the industry hikes tariffs ~12%, and Krishna's ARPU prints +2.5%. Diagnose.

Solution. (a) Average subs = 305m, so ARPU = 162,400m/(305 × 3) = 162,400/915 = ₹177.5/month. (b) Net adds = +6m, so disconnections = 21 − 6 = 15m. Churn = 15/915 = 1.64%/month, comfortably healthy (<2.5%). (c) A 12% hike that delivers 2.5% ARPU means the hike is not sticking. Subscribers are down-trading to cheaper packs, dropping add-ons, or letting second SIMs lapse. It's on the red-flag list. Check whose ARPU did capture the hike, which will be the premium-share operators. The gap is the market telling you where the pricing power actually lives.

P10 (independent): Share quality and the phantom base. A three-player market has 900m subscribers in total and quarterly industry revenue of ₹60,000 cr. K: 350m subs, ₹27,000 cr. L: 340m, ₹24,000 cr. M: 210m, ₹9,000 cr. VLRs: K 99%, L 97%, M 78%. (a) Compute each operator's subscriber share, revenue share, and ARPU. (b) Recompute M's ARPU on active subs and its active-subscriber share. (c) Verdict.

Solution. (a) K: subs 38.9%, revenue 45.0%, ARPU = 270,000/(350×3) = ₹257. L: 37.8%, 40.0%, ARPU = 240,000/1,020 = ₹235. M: 23.3%, 15.0%, ARPU = 90,000/630 = ₹143. (b) M active = 210 × 0.78 = 163.8m, so active-basis ARPU = 90,000/(163.8×3) = ₹183. The total active market = 346.5 + 329.8 + 163.8 = 840.1m, which puts M's active share at 19.5%. The reported 23.3% overstates it. (c) K owns the premium base. Its revenue share is 6.1pp above its subscriber share, so every tariff hike flows to it disproportionately. M's reported base overstates reality by a fifth, and even its true customers pay 29% below K's.

Revenue share is the league table you can trust. Subscriber share is marketing.

P11 (independent): The EV rebuild. SundarTel has a market cap of ₹3,80,000 cr. Conventional net debt is ₹55,000 cr, deferred spectrum ₹78,000 cr, AGR dues ₹22,000 cr, lease liabilities ₹45,000 cr, and it holds a listed tower stake of ₹25,000 cr. LTM EBITDA (post-Ind AS 116) is ₹58,000 cr, lease payments ₹8,200 cr, capex ₹33,000 cr. Compute (a) naive and adjusted EV/EBITDA, (b) the EBITDAaL pairing, (c) EV/(EBITDA − capex), and (d) inclusive net debt/EBITDA against the threshold.

Solution. (a) Naive EV = 380,000 + 55,000 = 4,35,000 → 7.5×. Adjusted EV = 380,000 + 55,000 + 78,000 + 22,000 + 45,000 − 25,000 = ₹5,55,000 cr → 555/58 = 9.6×. (b) EBITDAaL = 58,000 − 8,200 = 49,800, and EV ex-lease-liabilities = 5,10,000, giving 10.2×. Remember the pairing rule: leases go in the numerator or in the denominator's add-back, never neither and never both. (c) 5,55,000/(58,000 − 33,000) = 555/25 = 22.2×. That's the pre-capex illusion quantified, since a "9.6×" business is 22× on cash operating terms. (d) (55 + 78 + 22 + 45) = 2,00,000/58,000 = 3.45×. The ratio is above the <3× line, even though the conventional print (55/58 = 0.9×) looks pristine. Next quarter, the direction matters more than the level.

P12 (timed, 25 minutes): Two disguised companies. A (IT services): CC growth over the last four quarters ran +9.0% → +6.2% → +3.8% → +2.1%. Headcount moved −0.8% in each of the last two quarters. Utilization is 88.2% (ex-trainees), attrition 18.9% and rising, EBIT margin steady at 23.8%, book-to-bill 0.94×. B (telecom): EBITDA margin is 51%, and reported net debt/EBITDA is 1.7× (EBITDA ₹46,000 cr). Deferred spectrum of ₹68,000 cr, AGR dues of ₹18,000 cr and leases of ₹30,000 cr aren't in that ratio. ARPU rose +1.8% in the year of a ~15% tariff hike, churn is 3.1%/month, and capex intensity is 14% against ~27% for peers. For each: name the flags, compute one decisive number, state the valuation lens, and give the one question you'd answer next.

Solution. A: Three flags fire. There's CC deceleration with falling headcount (the master pair), book-to-bill <1×, and a margin held by utilization at 88% (above band) with rising attrition. Together that's the full demand-rollover-managed-for-optics pattern from the Kaveri example. The decisive number is utilization at 88.2%. It means the margin lever is exhausted, and next quarter the P&L must tell the truth. Lens: P/E, and whatever multiple it trades at, reverse it (P8) before you believe it. Next question: what do peers' CC growth and net adds show, cycle or share loss? B: Inclusive leverage = (1.7×46,000 + 68,000 + 18,000 + 30,000)/46,000 = 194,200/46,000 = 4.2×, far over the 3× line despite the "1.7×" print. ARPU +1.8% against a 15% hike means the hike isn't sticking. Capex at half of peers is underinvestment, and churn at 3.1%, drifting toward the 4% crisis line, is the lagging confirmation. Lens: adjusted EV/EBITDA + EV/(EBITDA − capex). Treat the equity as an option on repair. It's no compounder. Next question: how does the installment schedule compare with projected FCF? How many quarters of runway are left before capex gets cut again? Time check: spend ≤8 minutes on the two leverage/CC computations and the rest on verdicts.

P13 (guided): ARPU, ARPA, and the denominator trap. Cascade Mobile (synthetic US operator) reports quarterly postpaid service revenue $12,150m, average accounts 25.0m, average phone lines 75.0m. (a) Compute ARPU and ARPA. (b) An analyst writes: "Cascade's ARPA of $162 dwarfs an Indian operator's ARPU of ₹200. US customers are ~65× as valuable." Name the two errors.

Solution. (a) ARPU = 12,150 ÷ (75 × 3) = $54.00/month, and ARPA = 12,150 ÷ (25 × 3) = $162.00/month. That's 3.0 × ARPU, because each account carries three lines. (b) Two errors compound. Denominator: ARPA is per account and ARPU is per line or subscriber, so the like-for-like figure is Cascade's $54 ARPU. The $162 doesn't belong in the comparison. Currency: ₹200 ≈ $2.4, so the fair comparison is ~$54 against ~$2.4 per subscriber, about 22×. Even that gap reflects income levels and a postpaid-vs-prepaid structure. It doesn't measure "customer quality." Mixing the denominator and the currency is how a real ~22× gap gets mis-sold as 65×.

P14 (guided): The productivity pass-through, costed. Brahmani Consulting (synthetic) renews a ₹500 crore annual book at a 12% lower price. Its cost structure: direct delivery cost 62% of revenue, SG&A 14% of revenue (so EBIT margin 24%). (a) If it delivers the renewed scope with the same team (rupee cost unchanged), what is the new EBIT margin? (b) By what percentage must direct cost fall to hold the 24% margin, and what does that require operationally? (c) One sentence tying this to the GenAI indicator set.

Solution. (a) New revenue = 500 × 0.88 = ₹440 cr. Direct cost stays at ₹310 cr, and SG&A = 14% × 440 = ₹61.6 cr. So EBIT = 440 − 310 − 61.6 = ₹68.4 cr = 15.55%, down ~845bps from 24%. (b) To restore EBIT = 24% × 440 = ₹105.6 cr, direct cost must fall to 440 − 61.6 − 105.6 = ₹272.8 cr. The required cut is (310 − 272.8)/310 = 12.0%. The vendor must convert the entire price concession into an equal-percentage delivery-cost cut, with fewer or cheaper people per unit of scope. (c) It's the bull-vs-bear fork made mechanical. The price cut lands immediately, while the cost cut appears only if productivity is real. Watch revenue-per-employee and the T&M-vs-fixed-price mix. When a vendor signs fixed-price, it's betting it can hit that 12%.

P15 (independent): Prepaid customer economics, and what they hide. Yamuna Telecom (synthetic Indian prepaid operator): ARPU ₹200/month, service contribution margin 45%, monthly churn 2.5%, cost to acquire a gross add ₹150, cost of capital 12% a year. (a) Compute average lifetime, monthly contribution, and undiscounted and discounted LTV. (b) LTV/CAC (both) and payback. (c) The ratio looks spectacular. Why does it not rescue the investment case?

Solution. (a) Lifetime = 1 ÷ 0.025 = 40 months. Contribution = 200 × 0.45 = ₹90/month. Undiscounted LTV = 90 ÷ 0.025 = ₹3,600. Discounted at a monthly rate of 1%, it's 90 ÷ (0.025 + 0.01) = ₹2,571. (b) LTV/CAC = 3,600 ÷ 150 = 24× undiscounted and 17× discounted, and payback = 150 ÷ 90 = 1.67 months. Prepaid CAC is tiny (SIM + channel commission, no device subsidy), so per-customer economics look extraordinary. (c) Because LTV/CAC values the customer. It doesn't value the network. The spectrum, towers, and fiber the customer uses are the capital that must earn above WACC, and the "great margins, poor returns" arithmetic shows it typically earns ~7.5% against ~11–12%. The customer P&L is fine. The capital is what destroys value. High LTV/CAC sitting on a sub-WACC ROIC is the sector's defining illusion.

Always run both before you decide anything.

P16 (independent): The three-number cascade. Chambal Infotech (synthetic Indian major, reports in USD). Current-year revenue $4,000m: USD-billed $2,200m, EUR-billed $1,000m (as translated), GBP-billed $480m, others $320m (rates flat). Over the year the EUR depreciated 3% vs the USD and the GBP appreciated 4%. Prior-year revenue was $3,800m, and average INR/USD moved from 82.5 to 86.0. Compute reported USD growth, CC growth, and the INR headline, and say which is demand.

Solution. Reported USD = 4,000 ÷ 3,800 − 1 = +5.26%. For CC, restate at prior rates. EUR 1,000 ÷ 0.97 = $1,030.9 and GBP 480 ÷ 1.04 = $461.5. So CC revenue = 2,200 + 1,030.9 + 461.5 + 320 = $4,012.5 and CC growth = +5.59%. Here CC comes out above reported, because the falling euro hid a little demand that the rising pound didn't fully offset. For INR, prior revenue was 3,800 × 82.5 = ₹31,350 cr and current is 4,000 × 86.0 = ₹34,400 cr. INR growth is therefore +9.73%, of which ~4.2pp is pure rupee depreciation (86.0/82.5 − 1). One business gives three headlines, +9.7% / +5.3% / +5.6%, and only the CC +5.6% measures what clients bought.

P17 (independent): Leverage direction, not level. Ravi Telecom (synthetic), two consecutive years (₹ crore). Year 1: conventional net debt 40,000, deferred spectrum 60,000, AGR dues 20,000, leases 25,000, EBITDA 42,000. Year 2: conventional net debt 38,000, deferred spectrum 72,000, AGR dues 21,000, leases 27,000, EBITDA 40,000. (a) Compute conventional and inclusive net debt/EBITDA for both years. (b) What story does the conventional number tell, and what do the inclusive number and its direction tell?

Solution. (a) Conventional: Y1 = 40,000/42,000 = 0.95× and Y2 = 38,000/40,000 = 0.95×, flat and pristine. Inclusive, adding spectrum + AGR + leases: Y1 = (40,000 + 60,000 + 20,000 + 25,000)/42,000 = 145,000/42,000 = 3.45×, and Y2 = (38,000 + 72,000 + 21,000 + 27,000)/40,000 = 158,000/40,000 = 3.95×. (b) The conventional ratio says "barely levered, nothing to see," and it's fiction. Management repaid a little bank debt (40,000 → 38,000) while deferred spectrum rose from 60,000 to 72,000 after a fresh auction. Meanwhile EBITDA slipped on share loss. Inclusive leverage climbed from 3.45× to 3.95×. It was already past the <3× line, and it's rising while spectrum dues mount, which means the treadmill is winning. You're standing on the entry ramp to the death spiral. Over two years the direction is the signal, and the point-in-time conventional ratio actively conceals it.

P18 (timed, 18 minutes): Value a US carrier ($). Cascadia Telecom is a synthetic US national operator, in $bn. Market cap is 175, net debt including finance leases 150, a non-core media stake 8. EBITDA is 48, reported under ASC 842, where operating-lease rent is already in opex. Capex is 18, and average postpaid accounts are 40m. Compute (a) adjusted EV and adjusted EV/EBITDA, (b) EV/(EBITDA − capex), (c) inclusive net debt/EBITDA, (d) EV per postpaid account, and (e) one line on why "6–7× looks cheaper than India's ~9–10× but is not the bargain it seems," including the cross-standard catch.

Solution. (a) Adjusted EV = 175 + 150 − 8 = $317bn, so EV/EBITDA = 317/48 = 6.6×. Note the US lease treatment. Under ASC 842 operating-lease rent is already expensed. So, unlike with an Ind AS 116 telco, you don't add operating-lease liabilities into EV. Only debt-classified or finance leases count, and they're already inside net debt. That's the pairing rule in its US form. (b) EV/(EBITDA − capex) = 317/(48 − 18) = 317/30 = 10.6×. It's the pre-capex illusion again: a "6.6×" business is ~10.6× on cash operating terms. (c) Inclusive net debt/EBITDA = 150/48 = 3.13×, above the <3× line even in the mature structure. (d) EV per account = 317,000/40 = $7,925. Use it as a within-market sanity check only, and never set it against a ₹-ARPU base. (e) The 6.6× undercuts India's ~9–10× largely because the Ind AS 116 lease add-back inflates Indian EBITDA, so the two aren't comparable without alignment. Meanwhile the 10.6× capex cross-check and the 3.13× leverage show that the treadmill physics survive consolidation. Cascadia is priced as the dividend-anchored utility it economically is. It isn't a bargain. Time check: (a)–(c) ≤10 minutes, (d)–(e) the rest.

Applied mini-project: the two-sector workup

Task. Build both playbooks on real companies: one IT-services pair (an Indian major vs Accenture) and one telecom pair (an Indian operator vs a US major). Defend a comparative verdict in each sector. The work feeds the sector exam and seeds your sector KPI cheat-sheet pages.

Pick. For IT, take TCS or Infosys (India) and Accenture (US). For telecom, take Bharti Airtel (India) and Verizon or T-Mobile (US). Substitute equivalents if those are easier for you to source. If you want the distressed case, Vodafone Idea makes a powerful third telecom column.

Data to pull (all primary, all free):

• IT: start with the company's quarterly fact sheet or press release from its investor-relations site. TCS and Infosys disclose CC growth, TCV, utilization, attrition, headcount, client-bucket counts every quarter, so the KPI panel is essentially published for you. Add the annual report or 20-F (Infosys files on SEC EDGAR), Accenture's 10-K and quarterly new-bookings releases from EDGAR, and screener.in or stockanalysis.com for the 10-year spine.
• Telecom: start with Bharti Airtel's quarterly report, which has some of the best KPI disclosure anywhere (ARPU, churn, data usage, subscriber detail). Then read the annual report's borrowings, leases, deferred-spectrum and contingent-liability notes. TRAI's Quarterly Performance Indicator Reports and monthly subscriber releases (trai.gov.in) give you revenue market share via AGR, plus VLR data. Verizon's 10-K covers the rest (segment KPIs, the indefinite-lived-license note, lease schedules). Label anything you estimate.

Deliverables (≤6 pages + one workbook):

1. Two IT KPI panels (India major vs Accenture), with every metric from the IT KPI panel that you can compute or source, marked healthy/watch/stress. Verify one quarter's CC growth arithmetic yourself from the currency-mix disclosure.
2. The IT diagnostics. Build a 2-year EBIT-margin bridge sorted earned/exogenous/borrowed, in the Deccan Digital format. Reconcile the book-to-bill definitions (TCS total TCV vs Infosys large-deal TCV vs Accenture bookings), state what each includes, and ACV-adjust where duration is disclosed. Plot net adds against CC growth for 8 quarters. Then write your one-paragraph GenAI indicator reading, with the rev/employee trend included.
3. Two telecom KPI panels: ARPU, churn, data/user, net adds, subscriber vs revenue share (TRAI for India), VLR, capex intensity. Recompute ARPU from revenue and average subs. Don't transcribe it.
4. The telecom balance sheet, rebuilt: full EV (market cap + net debt + deferred spectrum + AGR/regulatory dues + leases − non-core stakes), the pairing-rule multiple both ways, EV/(EBITDA − capex), inclusive net debt/EBITDA vs 3×, for both companies, with the Ind AS 116/ASC 842 boundary handled explicitly.
5. Red-flag sweeps: run both companies in each sector against the two red-flag tables. Document what you checked, even where it came up clean.
6. Reverse both multiples: turn the IT P/E into implied perpetual growth (P8 method), and the telecom multiple into the EBITDA growth/capex path that would justify it. Write one paragraph each on whether the implied story is believable.
7. A two-page memo, Minto style with the conclusion first. In each sector, say which is the better business, which is the better risk-adjusted proposition at the current price, and which three numbers you'd monitor quarterly. This is education. It isn't advice, and you're graded on process.

Rubric (score each 0 = missing/wrong, 1 = partial, 2 = complete & correct; pass = ≥12/16 with no criterion at 0):

Reading & resources

• Pat Dorsey, The Five Rules for Successful Stock Investing. The Business Services and Telecom industry chapters: the published ancestors of both playbooks, including why he treats telecom moats skeptically. [Paid] [Beginner→Intermediate]
• TCS quarterly press release + fact sheet (tcs.com, investor relations). The sector's reference KPI disclosure: CC growth, TCV, headcount, attrition, client metrics. Read four consecutive quarters and the panel assembles itself. [Free] [Intermediate]
• Infosys quarterly release + Form 20-F (infosys.com; SEC EDGAR). Note how its large-deal TCV definition differs from TCS's total TCV; the 20-F's risk factors are a checklist of the red flags above. [Free] [Intermediate]
• Accenture Form 10-K and quarterly new-bookings releases (SEC EDGAR). The contrast model: consulting-heavy mix, ~15% operating margin by design, bookings disclosed with consulting/managed-services split. [Free] [Intermediate]
• Bharti Airtel quarterly report (airtel.in, investor relations). Best-in-class telecom KPI disclosure: ARPU, churn, data per user, 4G/5G mix, per-segment capex. The mini-project's Indian backbone. [Free] [Intermediate]
• Vodafone Idea annual report. Read the borrowings, deferred-spectrum, AGR, and contingent-liability notes as a live distressed-balance-sheet case study. [Free] [Advanced]
• Verizon Form 10-K (SEC EDGAR). Segment KPIs (note ARPA vs ARPU), the wireless-licenses note (indefinite-lived, the cross-standard trap), lease schedules. [Free] [Intermediate]
• TRAI Quarterly Performance Indicator Reports + monthly subscriber data (trai.gov.in). The sector's primary dataset: AGR-based revenue market share, subscriber share, VLR. No other market gives you this for free. [Free] [Intermediate]
• The Supreme Court AGR judgment (Oct 2019): any detailed contemporaneous summary, plus the AGR-liability notes in Bharti's and Vodafone Idea's annual reports. Read the definition fight itself; it is the template for regulatory-definition risk everywhere. [Free] [Advanced]
• Edward Chancellor (ed.), Capital Returns. The capital-cycle framework; telecom and other capex-heavy industries as recurring cases. Re-read it after the India-vs-US capital-cycle section. [Paid] [Advanced]
• Aswath Damodaran's data library (pages.stern.nyu.edu/~adamodar). Sector-level ROIC, margins, and multiples for software/IT services and telecom, updated each January; use it to benchmark both panels. [Free] [Intermediate]
• Business Breakdowns podcast, the Accenture episode (and any tier-1 Indian IT or US telecom episode in the feed). Practitioners running these panels aloud. [Free] [Intermediate]
• Zerodha Varsity, Fundamental Analysis module (free web/PDF). Refresher on Indian AR navigation while you do the mini-project. [Free] [Beginner]
• Earnings-call transcripts for your four project companies (company IR sites; screener.in links them for India). Listen for "productivity pass-through," "vendor consolidation," "tariff repair," and "network investment" language against the numbers. [Free] [Intermediate]
• Going deeper. Alfred Rappaport & Michael Mauboussin, Expectations Investing (revised edition), the price-implied-expectations chapters (Columbia Business School Publishing). The full method behind the IT valuation section: reverse the multiple into the market's implied growth, then locate the operating "value triggers" that would confirm or refute it. The natural next step after the key-value-driver reverse of the IT P/E. [Paid] [Intermediate→Advanced]
• Going deeper. Aswath Damodaran on user- and subscriber-based valuation (his subscriber-value working paper and companion blog posts, free at pages.stern.nyu.edu / aswathdamodaran.blogspot.com). Builds the ARPU → lifetime → LTV/CAC → firm-value bridge formally, with the discounting and renewal-curve refinements that worked examples 4–5 deliberately simplify. Read it to see where the ARPU/churn/LTV arithmetic goes next. [Free] [Advanced]

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": "Reading an IT-services or telecom KPI panel across many quarters and many names without losing the by-hand discipline of §4.2–4.3",
  "use": "An LLM turns ten quarters of scattered fact-sheet PDFs into one clean KPI table (CC growth, TCV, book-to-bill, attrition, utilization, net adds for IT; ARPU, churn, VLR, inclusive net debt/EBITDA for telecom) in minutes instead of an afternoon of copy-pasting, and drafts a first pass of the red-flag checklist against it.",
  "tools": ["Claude/ChatGPT/Gemini to extract figures from a pasted quarterly fact sheet or 10-K/20-F excerpt into a table", "a notebook (Python/pandas) or spreadsheet to hold the panel and compute CC growth and inclusive EV yourself", "screener.in's exported financials for the India names", "SEC EDGAR full-text search for the US 10-Ks/20-Fs"],
  "prompt": "Paste in this quarter's and the last four quarters' IT-services (or telecom) fact-sheet excerpts. Extract into a table: revenue, CC growth %, TCV, net headcount adds, utilization %, attrition % (or ARPU, net adds, churn %, capex/revenue for telecom). Flag any quarter where CC growth decelerated while headcount fell, or ARPU stayed flat despite an announced tariff hike. Show your extraction, not just the flags.",
  "verify": "Recompute CC growth and ARPU/churn yourself from the raw disclosed numbers for at least one quarter (§4.2.2 and §4.3.4's formulas) and check the tool's table against it — LLMs mis-copy digits and conflate TCV definitions (TCS's total TCV vs Infosys's large-deal-only TCV) more often than they get the mechanics wrong. Never accept a flagged red flag without opening the source page.",
  "diy": "The mastery gate is unaided: build the eleven-KPI IT panel and the nine-KPI telecom panel from a raw fact sheet by hand, compute CC growth and ARPU from their definitions, and state which red flags are firing and why — no tool in the room. The tool above saves the transcription hour; it does not replace knowing what CC growth means or why book-to-bill needs an ACV check."
}

Modern Data Analysis

By hand first. The people-leverage machine and the CC-growth restatement above are hand arithmetic on one quarter of one company. The EV rebuild is the same thing for one telecom balance sheet on one day.

Today's workflow. A coverage analyst tracking six IT-services names and three telcos doesn't re-key each fact sheet every quarter. In practice you pull the last 8–12 quarters of each company's disclosed KPIs and hold them in a small table. Then you let code carry every quarter's CC-growth-vs-net-adds check and every telco's inclusive-EV rebuild forward automatically. The "master red flag" becomes a computed column, and you stop re-running it from memory every quarter.

Tools & sources (IN + US). For India, use TCS's and Infosys's quarterly investor fact sheets, which sit on company IR pages as machine-readable XLS. TRAI's Telecom Subscription Data reports and Quarterly Performance Indicator Reports (trai.gov.in) publish AGR-based revenue share and VLR monthly, for free. And screener.in's "export to Excel" gives you a ten-year ratio history per company. For the US, SEC EDGAR's XBRL company-facts API (data.sec.gov/api/xbrl/companyfacts/CIK##########.json) gives machine-readable historical revenue, capex, and debt straight from the filings. Accenture is CIK 0001467373 and Verizon is CIK 0000732712. FRED series DEXUSEU and DEXUSUK (or DTWEXBGS, the broad dollar index) let you rebuild a CC-growth restatement with a real quarter's exchange-rate moves, so you aren't stuck with the synthetic ones used here. Damodaran's data library has the sector ROIC/margin benchmarks cited in the valuation section.

# recomputed in-session -- R3
# Part 1: verification anchor -- the Sec 4.2.1 IT-services machine
hours_std, util = 1800, 0.85
billed = hours_std * util
layers = {
    "onsite_senior": dict(fte=15, rate=82, cost=115_000),
    "offshore_mid":  dict(fte=25, rate=36, cost=40_000),
    "offshore_jr":   dict(fte=60, rate=24, cost=13_000),
}
revenue = sum(v["fte"] * billed * v["rate"] for v in layers.values())
cost = sum(v["fte"] * v["cost"] for v in layers.values())
gross_margin = (revenue - cost) / revenue
print(f"revenue ${revenue/1e6:.2f}m  cost ${cost/1e6:.2f}m  gross margin {gross_margin:.2%}")
# -> revenue $5.46m  cost $3.51m  gross margin 35.83%

# Part 2: a KPI panel that flags the module's own red flags across many quarters at once
import pandas as pd
panel = pd.DataFrame({
    "quarter":       ["Q1", "Q2", "Q3", "Q4", "Q5"],
    "cc_growth_pct": [8.1, 6.4, 4.0, 2.1, 1.8],       # illustrative, not a real company
    "net_adds":      [4200, 1800, -900, -2600, -3100],
})
panel["it_master_red_flag"] = (panel["cc_growth_pct"].diff() < 0) & (panel["net_adds"] < 0)
print(panel[["quarter", "cc_growth_pct", "net_adds", "it_master_red_flag"]])

Verify. The code reproduces the $5.46m revenue and $3.51m cost from the worked build. The gross margin comes out 35.83%, against the "35.7%" in the text. That gap comes from the text's own intermediate rounding, which sums figures already rounded to two decimals before dividing. It isn't a disagreement, and repeating the same rounding sequence gives 35.7%. The CC-growth worked example (Meridian InfoTech) and the DeccanTel EV rebuild both reproduce every stated figure to the text's own rounding: CC growth +5.0%, naive EV/EBITDA 7.5×, true EV/EBITDA 9.5×, inclusive net debt/EBITDA 3.5×. The panel above is illustrative data built to show the mechanism, and Q3 shows the master-red-flag pattern. It makes no claim about any named company.

Quantitative lens

Refuse reported revenue growth. Use constant-currency growth. INR growth flatters with every year of rupee slide (~3–4%/yr), and USD growth still carries cross-currency noise. The worked cascade earlier showed one business printing three different growth headlines (+8.5% INR, +4.3% USD, ~+5% CC) from one underlying demand. CC growth is the only one of the three that isolates volume-and-price. So refuse the other two as a starting point for any IT-services thesis, however prominently management leads with them.

Refuse naive EV/EBITDA on telecom. Use the inclusive rebuild. DeccanTel's naive multiple of 7.5×, on market cap plus conventional net debt, understates the true 9.5× by two full turns. Deferred spectrum, AGR dues, and lease liabilities never entered the EV. The understatement runs the same direction for every Indian telco carrying spectrum debt. An unadjusted EV/EBITDA screener column is therefore biased cheap across the whole sector, and nothing on the screen warns you. Rebuild EV by hand, or in code, before you trust a screen.

The two playbooks need opposite valuation instincts, and mixing them up is the most common practitioner error in this pair. IT services has real, cash-backed earnings (capex 1–2% of revenue, ~80–100% FCF payout), so P/E is correct, and the debate is entirely about growth durability. Reverse the key-value-driver formula and a 26× multiple at ROIC 45% and r 11.5% implies roughly 8.4% perpetual growth. That's a falsifiable claim you can set against the CC-growth trend and the GenAI indicator set. Telecom's earnings are polluted by policy-sensitive D&A and leverage choices, so adjusted EV/EBITDA is correct, and P/E there is close to meaningless. Apply IT's lens to telecom, or the reverse, and you'll misprice both.

What code can't settle. The GenAI productivity question is unresolved as of mid-2026, and no amount of code changes that. The indicator set (revenue/employee inflection, net-adds decoupling, fixed-price mix) is a monitoring tool. It won't work as a forecasting model, and a simulation dressed around it would manufacture false precision on a structural question the market itself hasn't settled. The illustrative panel above demonstrates the mechanism of automated red-flag detection. Don't point it at a real ticker and trust it until you've independently re-derived CC growth and the inclusive EV from that company's own filings, definition quirks (TCS vs Infosys TCV) and all.

Do it in code: pull the panel with G2/DA1.02, chart the CC-growth-vs-net-adds divergence with G2/DA1.04, and regress margin against utilization and net adds with G1/QM1.02 to turn the "~0.7–0.8pp per point of utilization" rule of thumb into a coefficient with a confidence interval.

Where this goes next: galaxy cross-links

• G2 · DA1.02 Pandas data wrangling. The KPI panel above, built properly: joining quarterly fact-sheet exports into one clean time series instead of five separate spreadsheets.
• G2 · DA1.04 EDA & visualization. Charting CC growth against net adds (or ARPU against churn) is how the "master red flag" becomes visible before a quarterly print states it in words.
• G1 · QM1.02 Hypothesis testing & regression. Turns "each point of utilization is worth ~0.7–0.8pp of gross margin" from a rule of thumb into a regression coefficient with a confidence interval, across many quarters or many companies at once.
• G2 · CS2.05 APIs, HTTP & JSON. The mechanics of pulling EDGAR's XBRL company-facts API and TRAI's published datasets programmatically instead of downloading PDFs by hand.
• G3 · CN4.02 Industry primers & disruption. The GenAI bear/bull framing above is a live instance of the disruption-analysis pattern taught there; run the indicator set above through that framework.
• 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. A smart 15-year-old asks: "My cousin joined Infosys, and our family pays Airtel ₹300 a month. Which company has the better deal?" Write one page that explains four things. First, why the company that rents out people earns fabulous returns on almost no money down. Use the idea of billing your cousin's hours for more than her salary, many thousands of times over. Second, why the company that owns the towers struggles even though nobody ever cancels their phone. The network had to be built first, the government auctions the airwaves, and the whole thing must be rebuilt every few years. Third, why "the phone company has 50% margins" and "the phone company barely pays its way" are both true at once. Fourth, what single question decides each company's future. For Infosys, will clients keep buying more hours in the AI era? For Airtel, will the three players keep prices sane? Keep it free of jargon, with pocket-money arithmetic only.

Journal reflection. Two entries. (1) Your own telecom scuttlebutt: what do you pay per month, what did you pay three years ago, and did you down-trade or upgrade after the last tariff hike? Which KPI in the telecom panel registered your behavior, and with what lag? (2) Your GenAI position, pre-registered: write three sentences on which way you currently lean in the GenAI debate. Then write the two indicator readings (specific numbers, specific companies, specific quarters) that would force you to change your mind. File both in your knowledge system, and re-open the GenAI entry every results season. It's calibration practice (Phase 9) on the sector's biggest open question. The value comes from having written it down before the answer arrives.

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