IT services business model

4.2.1 The machine: rate × utilization × pyramid

The entire revenue line of an IT-services company is four numbers multiplied:

Revenue = billable headcount × standard hours × utilization × realized rate per hour

and the entire margin story is the gap between the realized rate (what clients pay per hour) and the loaded cost (what the employee costs per hour). Three structural levers set that gap.

1. The pyramid. Projects are staffed like a pyramid: one onsite architect, a few offshore leads, and a broad base of 2–6-year engineers and freshers. The junior layers bill at rates far above what they cost. Seniors are closer to breakeven, since their value is in winning and holding the account. A fatter base means a cheaper average employee, and a cheaper average employee means a wider margin. Fresher hiring, in other words, is a strategic variable. It deserves more than an HR footnote.

2. The offshore/onsite mix. An onsite employee, working in the client's country, might bill $80/hour and cost the firm $110,000+ a year. An offshore engineer bills $25–35/hour and costs $12,000–45,000 a year depending on seniority. Offshore revenue per head is lower, but offshore margin per head is far higher. The global delivery model puts a thin onsite layer near the client for relationship and requirements, over a deep offshore factory for execution. That arbitrage built the Indian industry. Track two mixes here. The effort mix is the share of hours delivered offshore, typically ~75–80% for tier-1 India. The revenue mix is offshore's share of revenue, and it's much smaller than its share of effort because offshore rates are lower.

3. Utilization. This is the share of available employee time actually billed to clients. Firms disclose it both including and excluding trainees, so read the definition. The unbilled remainder is the bench: people between projects, in training, or hired ahead of demand. Hold one band in your head. Utilization of 82–87%, excluding trainees, is healthy. Below ~80%, the firm is paying for idle capacity and margins bleed. Above ~87–88%, there's no slack to staff new deals. The firm is eating its growth inventory, and delivery quality risk rises.

Utilization is also the one dial managements can turn instantly, so keep a close eye on it. A margin "beat" that came from stretching utilization is borrowed margin, and you should read it that way.

Worked example: margin in a 100-person delivery unit

Now build the machine once, with numbers. The unit is synthetic and in USD for clarity. Take a delivery unit of 100 billable FTEs, a standard year of 1,800 hours and utilization of 85%, which gives 1,530 billed hours each:

Gross margin = (5.46 − 3.51)/5.46 = 35.7%. Take off ~13% of revenue for SG&A and delivery overheads and you're left with EBIT ≈ 22–23%, the middle of the tier-1 band. Revenue per employee comes to $54.6k, squarely in the Indian-heritage range. For comparison, TCS runs ~$48k and Infosys ~$58k, while Accenture reaches ~$84k on a consulting-heavier, onsite-heavier mix. All of those are illustrative, modeled on ~FY24/FY25 scale, so pull the actual filings.

Now move the dials and watch the sensitivity:

• Drop utilization from 85% to 78%. Billed hours fall to 1,404 and revenue falls 8.2% to $5.01m. Costs don't move, because the bench still gets paid, so gross margin drops to 30.1%. That gives you a rule of thumb: each point of utilization ≈ 0.7–0.8pp of gross margin.
• Broaden the pyramid, taking offshore mid from 25 to 15 and junior from 60 to 70. Revenue slips 3.4% to $5.28m, since juniors bill less, but cost falls by more. Gross margin rises to 38.7%, up ~3pp. The pyramid lever trades a little revenue for a lot of margin. It's also the easiest lever to abuse, as the growth-margin-pyramid triangle further down will show you.

One more structural fact. Every unit of growth requires hiring, training, and benching a human, so growth has historically been linear in headcount. That linearity is what makes headcount net adds a leading indicator. And the GenAI debate is about that very linearity.

4.2.2 Constant-currency growth: the only growth number that measures demand

An Indian IT major earns roughly half its revenue in USD and ~20–30% in EUR/GBP, with the rest in other currencies, and it reports in INR. That leaves you with three growth numbers, and they can diverge wildly:

1. INR revenue growth, flattered every year the rupee depreciates. That's most years, since the INR's long-run slide against the USD is ~3–4% a year. As a demand signal it's meaningless.
2. Reported USD growth, which is better but still distorted by cross-currency moves. If the euro falls 6% against the dollar, euro-billed work translates into fewer dollars even when clients bought more work.
3. Constant-currency (CC) growth: this period's revenue restated at the prior period's average exchange rates, compared with the prior period. It gives you the volume-and-price truth, meaning how much more work clients actually bought and at what rates. It's the sector's headline number, and the only one you should quote.

How do you calculate constant-currency growth?

Worked FX adjustment. Meridian InfoTech (synthetic) reports in USD. This quarter's revenue is $1,000m, of which $600m is billed in USD, $250m in EUR (as translated) and $150m in GBP. Over the year the EUR depreciated 5% against the USD and the GBP appreciated 2%. The prior-year quarter brought in $962m.

• Reported growth = 1,000/962 − 1 = +4.0%.
• Restate at last year's rates. EUR-billed revenue would have translated to 250/0.95 = $263.2m, and GBP-billed to 150/1.02 = $147.1m. CC revenue = 600 + 263.2 + 147.1 = $1,010.3m.
• CC growth = 1,010.3/962 − 1 = +5.0%. The euro was hiding a full point of real demand.

Now for the INR cascade, the version that fools retail investors every quarter. Say the same firm's revenue is $2,400m for the year against $2,300m last year, so USD growth is +4.3%. Meanwhile the average INR/USD rate moved from 83.0 to 86.3, a rupee depreciation of ~4%. In rupees, revenue goes from 2,300 × 83.0 = ₹19,090 cr to 2,400 × 86.3 = ₹20,712 cr. That's +8.5% "growth."

Half that headline is currency.

If cross-currency moves also flattered the dollar number, CC growth might be 3.5–4%. So one business gives you headlines of 8.5%, 4.3% and 3.9%, and only the last one is demand.

Companies disclose CC growth in every quarterly release. Your job is to refuse every other version. When you check their arithmetic, rebuild it the way you just did, with current revenue at prior-period average rates.

4.2.3 The KPI panel

Run these eleven dials on any IT-services company, each against its band. The thresholds are tier-1-India rules of thumb, so calibrate them to size and mix. Mid-caps run hotter growth and often thinner margins, and Accenture's consulting-heavy model runs ~15% operating margin by design.

How do you read the main IT services KPIs?

Six of these deserve a paragraph each.

TCV and book-to-bill. TCV (total contract value) is the full value of deals signed in the period, and book-to-bill = TCV ÷ revenue for the same period. Above ~1.0× the order book is at least replacing what's burned, and 1.1–1.2×+ signals acceleration. Sustained sub-1.0× means revenue is eating backlog.

Watch for three traps. First, definitions differ. TCS discloses total TCV, but Infosys discloses only large-deal TCV, which is a subset. Divide Infosys's number by revenue and you get a meaningless "book-to-bill," so never compare the two raw. Second, duration inflates TCV. A $500m deal over 5 years adds the same TCV as a $500m deal over 2, but only 40% as much annual revenue. When average deal duration stretches, convert to annualized contract value (ACV = TCV ÷ average duration) before celebrating. Third, renewals vs new. A book full of renewals defends revenue, and only net-new TCV grows it. Ask for the split. If you can't get it, watch whether big TCV quarters actually convert into revenue acceleration 2–4 quarters later.

Attrition. LTM voluntary attrition above ~20% is the pressure line. The industry lived through it in FY22, when many firms printed 20–28%, and the P&L showed the full anatomy: wage hikes off-cycle, retention bonuses, subcontractor costs spiking to backfill seats, fresher over-hiring, delivery wobble. Subcontractor expense is the tell, so watch it as a % of revenue. Each exit costs roughly 30–50% of the seat's annual cost in recruiting, training, and ramp time, plus unbillable transition weeks. When attrition normalizes back to ~12–15%, as it did by FY24–25, you get a margin tailwind worth 100–200bps. Keep that in mind when you read a margin bridge, before you credit management with structural improvement.

Headcount net adds, the leading signal. Because the model is linear in people, firms hire ahead of demand they can see in their pipelines. When net adds rise, management sees growth coming. When net adds stay flat or negative for 2–3 quarters, it doesn't, whatever the commentary says. So the most revealing line in any IT quarterly is the net-adds number read against the CC growth number. CC growth decelerating while headcount falls is the sector's master red flag, with the present and the future rolling over at once.

One GenAI caveat. If revenue per employee is inflecting upward at the same time, the linkage itself may be loosening, and that's the open question the GenAI section takes up. Check for it before you declare the signal broken.

Client and vertical concentration. BFSI (banking, financial services, insurance) is ~30% of tier-1 Indian IT revenue. That figure is illustrative, so check each firm's segment note. Either way, it makes the sector a leveraged bet on bank IT budgets, and the 2023 US regional-bank stress promptly showed up as guidance cuts. Top-client concentration matters most in mid-caps. There one account can be 10–20% of revenue, and losing it is an earnings event.

DSO with unbilled. Compute DSO on receivables plus unbilled revenue/contract assets. Unbilled arises legitimately from fixed-price percentage-of-completion accounting, where revenue is recognized as effort is incurred and invoiced at milestones. But it's also revenue that rests on management judgment. If unbilled grows much faster than revenue for several quarters, you have a revenue-quality question, the kind your forensics work trained you to chase. Treat it that way long before anyone calls it a "growth" story.

Discretionary vs cost-takeout demand. Two engines pull the sector, and they run on different fuels. Discretionary work (transformation programs, new digital builds, experimentation) expands when client confidence is high and gets cut first in downturns. This is also the shorter-cycle revenue. Cost-takeout work (outsourcing, managed services, vendor consolidation) is counter-cyclical. CFOs under pressure sign bigger, longer deals to cut run-costs, so mega-deal announcements cluster in downturns even as revenue growth slows.

Read every demand commentary through this split. "Strong TCV, weak revenue conversion" usually means cost-takeout is replacing discretionary. Such a shift is defensive, deflationary for growth, and margin-mixed, since takeout deals often start dilutive.

4.2.4 The growth–margin–pyramid triangle

Every margin print in this sector gets judged against one tension. Growth, margin, and pyramid health pull against each other, and a management can optimize any two at the expense of the third.

• Growth needs hiring ahead. Freshers get benched and trained before they bill. That depresses utilization and margin today.
• Margin can be defended instantly. Stretch utilization, freeze hiring, broaden nothing. That starves growth capacity.
• The pyramid keeps score in the background. Stop hiring freshers for two years and the base narrows. The same people, now two years more senior and more expensive, deliver the same work. Average cost per head ratchets up even with zero wage inflation.

Make the third point concrete. Index loaded costs: junior 1.0, mid 2.2, senior 4.0. A healthy mix of 55% junior / 30% mid / 15% senior has an average cost index of 0.55 + 0.66 + 0.60 = 1.81. Now freeze hiring for two years and let the mix age to 45/35/20. The index becomes 0.45 + 0.77 + 0.80 = 2.02, so cost per employee is up +11.6% with not one salary raised.

That's real money, and it comes due later. The margin "defended" in years 1–2 (no bench, no training costs, utilization stretched) is repaid in years 3–4, as a structurally costlier pyramid meets returning demand. By then the firm must hire laterals at market premiums and rebuild the fresher base at the same time.

The whole industry lived through this in FY22.

So when a company holds EBIT at 24% while CC growth halves and net adds go negative, don't write "impressive cost control." Write this: "margin held via lower net adds: future growth and future margin are being sold to protect this year's print." It's one of the sector's listed red flags, and now you know how it works. Margin defense done properly (pricing, pyramid broadening, offshore shift, SG&A leverage) shows up with stable or rising net adds and utilization inside the band.

4.2.5 The GenAI question: treated evenhandedly

The sector runs on a linear equation: more revenue requires more people. That equation now faces its first credible challenge since offshoring built it. As of mid-2026 the outcome is unresolved. Your job is to know the indicators and keep your own view modest. Here are both cases at full strength.

How could GenAI change IT services?

The bear case. Code generation, testing automation, and agentic tooling raise output per engineer materially, plausibly by 20–40% on the large share of work that is maintenance, testing, migration, and routine development. IT services is priced substantially in hours, either as time-and-materials or as fixed prices benchmarked to hours. Clients will therefore demand the productivity back as price cuts at every renewal: "do the same for 25% less." The pyramid gets hit at its money-making base first, because the junior, high-markup work is what the tools do best. Revenue deflates even if workloads grow, because the industry's pricing anchor deflates. In this world headcount-led growth is over, and the sector re-rates from "compounder" P/Es toward utility P/Es.

The bull case. Every prior efficiency wave was supposed to shrink the industry, and each one expanded it instead. Y2K built the client base. Offshoring cut prices 40–60% and demand exploded, because cheaper IT meant more IT (Jevons' paradox in services). Cloud was going to eliminate infrastructure work and instead spawned a decade of migration programs. RPA "automated away" BPO, and BPO grew. GenAI creates enormous new work as well: data-estate cleanup (models are useless on messy data), integration, model deployment and operations, governance and compliance layers, and the change-management that enterprises can't staff internally. Yes, there's deflation per unit of output. There are also more units, and the integrators capture the assembly work as they always have. In this world the leaders gain share, since only scale players can retrain 300,000 people, and the multiple holds.

So who's right? Nobody can say yet, but these are the indicators in the filings that will settle it:

1. Revenue per employee. This is the decisive chart. If firms grow revenue with flat headcount, rev/employee inflects upward and the linear model is productively broken, which is the bull case. If revenue and headcount fall together, it's just a downturn. The cycle explains it. If revenue falls with flat headcount, pricing is deflating. Call that the bear case.
2. Net adds decoupling from CC growth. Sustained growth without hiring means productivity capture, but watch several quarters before you conclude anything.
3. Pricing-model mix. When disclosures shift from T&M toward fixed-price and outcome-based contracts, productivity gains move from the client's pocket to the vendor's. Track the fixed-price share.
4. Renewal economics. In earnings calls, listen for "productivity pass-through" and for deal-size compression on like-for-like renewals.
5. AI-related revenue disclosures. Bring the same skepticism you'd bring to any forensic read, because relabeling is a real risk. Even so, the direction across 2–3 years matters.

Hold both cases and grade the evidence quarterly. And remember a lesson you'll meet again and again: when a structural debate is open, you can read the market's implied answer in the multiple, as the valuation section below shows. Your edge comes from comparing that implied answer with the indicator set. Out-shouting either camp adds nothing to it.

4.2.6 Valuation: P/E, and the growth-durability question

Why is P/E the right lens here, and almost nowhere else in this phase? Because an IT major's earnings are nearly cash. Capex is 1–2% of revenue, and leverage is nil because these firms hold net cash. Working capital grows slowly, and payout ratios run 80–100% of FCF. TCS and Infosys have returned essentially all free cash via dividends and buybacks (illustrative, so check the current policy). So P/E ≈ P/FCF, and the valuation machinery you built earlier applies directly. Cross-check with FCF yield, meaning FCF divided by market cap. Tier-1 names at 22–28× earnings offer ~3.5–4.5% FCF yields, and that's your comparison against bond yields and other compounders.

What P/E should an IT services company trade on?

The whole debate about the multiple is growth durability, because return on capital isn't the constraint. Use the key-value-driver formula: fair P/E ≈ (1 − g/ROIC)/(r − g). Take core ROIC ≈ 45%, cash-adjusted, since the operating business earns enormously on tiny capital. Take r = 11.5% in INR terms:

• g = 7%: (1 − 0.156)/0.045 = ~19×
• g = 4.5%: (1 − 0.100)/0.070 = ~13×
• g = 9%: (1 − 0.200)/0.025 = ~32×

This is single-stage teaching math, so read it for direction and magnitude and ignore the decimals. Two lessons fall out. First, because g/ROIC is small, reinvestment barely matters, and the multiple is almost purely a bet on g and its duration. Second, you can extract the market's implied g. A 26× multiple at these inputs implies perpetual growth of ~8.4%.

Set that against nominal GDP (~9–10% India, ~4–5% developed) and against the GenAI indicator set, and you've turned a price into a testable claim. That's what question 5 asks of you. When tier-1 Indian IT trades at 25–30× while CC growth runs 3–5%, the market is asserting that the slowdown is cyclical and the compounding resumes. It may be right. Your job is to know that this is the claim. Then watch net adds, book-to-bill, and rev/employee to see whether it's confirmed or refuted.

The India premium over Accenture-style peers isn't only irrational, either. A higher-margin delivery mix, net-cash balance sheets, near-total FCF payout, and two decades of governance cleanliness earn a durability premium. Whether today's premium is right is a question you now know how to interrogate.

4.2.7 IT-services red flags