पाठशाला Pathshala · हिसाब Hisāb, Unit economics · Lesson 07 · Start
Services and agency economics: utilisation, rate and bench
A services firm sells hours. Three numbers decide whether those hours make money: how many are billed, what each one actually collects and what the person costs. How to compute them, and when to stop selling hours at all.
Pathshala, The Founder Library · 11 October 2026 · 8 min read
A services firm sells time. Every billable person arrives each month with a fixed number of hours and a fixed cost, and the firm’s whole economics is what fraction of those hours gets sold, at what price, and how much of that price is actually collected. Most agency founders know their rate card to the rupee. Far fewer know their realised rate, their utilisation or the utilisation at which the firm stops losing money, which are the three numbers that decide whether the next hire makes the company richer or poorer.
This lesson builds those numbers from a timesheet and a payroll register, works one firm through to its monthly profit, shows the bench for what it is, and ends with the question every good services firm eventually asks: should we keep selling hours at all?
The services equation
Revenue from one billable person in a month is the product of four numbers. Available hours are the hours a person can work after leave, public holidays, training and internal meetings; for a full-time employee in India this is usually 140 to 160 a month, not the 200 that a five-day week suggests. Utilisation is the share of available hours spent on client work that is meant to be billed. Rate card is the hourly price the firm quotes. Realisation is the share of the hours worked on clients that is actually invoiced and collected, after the hours written off because the job overran, the discount that won the renewal and the scope that crept in without a change order.
Multiply the last two and you have the realised rate, the only rate that matters. A firm with a ₹2,500 rate card and 80 per cent realisation is a ₹2,000-an-hour firm, whatever the proposal says. Most fixed-fee projects are priced from an estimate of hours; when the project takes longer than the estimate the overrun is a realisation loss, and it is the most common reason an agency that looks busy is not making money.
Against that revenue sits the loaded cost of the person: salary, the employer’s provident fund contribution, gratuity accrual, health insurance, the laptop and the software seats. Revenue per head less loaded cost is gross margin per head. Everything that does not bill, the founders’ time spent selling, finance, HR, the office, is overhead and has to be paid out of the sum of those margins. The [lesson on COGS](/library/what-belongs-in-cogs) explains why the people who deliver the work sit above the gross-margin line and the people who sell it sit below.
One firm, worked through
A performance-marketing agency in Pune has sixteen billable people at an average loaded cost of ₹1.2 lakh a month and ₹15 lakh a month of overhead: two founders, an account director who does not bill, a finance executive, the office and the tools. Each billable person has 150 available hours. The rate card is ₹2,500 an hour and realisation runs at 80 per cent, so the realised rate is ₹2,000. Utilisation last quarter was 70 per cent.
Each person therefore bills 105 hours at ₹2,000, which is ₹2.1 lakh a month. Gross margin per head is ₹90,000, or 43 per cent. Sixteen people produce ₹14.4 lakh of gross margin against ₹15 lakh of overhead, and the agency loses ₹60,000 a month while every person in it feels fully stretched. The break-even utilisation is 71 per cent. At 80 per cent each person bills ₹2.4 lakh, gross margin per head reaches 50 per cent and the agency makes ₹4.2 lakh a month. Ten points of utilisation are the difference between a firm that is slowly dying and one that can pay a bonus.
Realisation is worth the same attention. Moving it from 80 to 90 per cent at 70 per cent utilisation takes the realised rate to ₹2,250 and monthly profit to ₹3.6 lakh, without a single extra hour of work. Every point of realisation recovered is a point of price increase that no client had to approve.
The figure starts at the Pune agency. Drag utilisation and watch the dot cross the break-even line. Then lower realisation to 70 per cent, which is what a firm that accepts every scope change for free looks like, and watch the break-even line move to the right of anything a team can sustain. Finally add four people at the same utilisation: if the line was below zero, more people only make the loss larger.
What good looks like, from the largest services firms
India’s IT services companies publish the numbers a small agency should be tracking, and they are a useful calibration. Infosys’s fact sheet for the quarter to 31 March 2026 reports utilisation of 83.0 per cent excluding trainees and 79.7 per cent including them, across 3,28,594 employees. Its income statement for the same quarter shows revenue of ₹46,402 crore and cost of sales of ₹32,058 crore, a gross margin of 30.9 per cent. That is a firm with offshore delivery, enormous scale and decades of process discipline, and it lives on low eighties utilisation and a gross margin near thirty.
Two lessons follow. First, utilisation above the mid eighties is not a target a small firm should plan for; leave, training, sales support and the gaps between projects consume the rest. Plan for 70 to 80 and treat anything higher as a warning that the team is being burned. Second, an agency gross margin far above the large firms’ is usually a sign that delivery people are being counted as overhead, not that the agency is better run. Andreessen Horowitz’s 16 Startup Metrics asks for all costs of manufacturing, delivery and support of a service inside gross profit; an investor will put them back.
The bench, honestly counted
The bench is every available hour not spent on billable client work: the designer between projects, the developer waiting for a client’s sign-off, the new hire still learning the tools. Its cost is exactly the unbilled share of the loaded cost, and it is the most expensive thing in a services business because it is invisible on the invoice. At the Pune agency, 30 per cent of 150 hours across sixteen people is 720 hours a month, worth ₹14.4 lakh at the realised rate, which is almost precisely the firm’s overhead.
Count it every week, by person, from the timesheet. Hire against signed work with a start date, not against the pipeline; a hire made for a deal that slips by two months is two months of bench. Keep a short list of internal work worth doing when people are free, the agency’s own case studies, a reusable template, a training module, so that the bench produces an asset rather than nothing. And watch the senior people above all: a principal on the bench costs three juniors.
A busy agency and a profitable agency are different things. The difference is utilisation and realisation, and both are written in the timesheet before they show up in the bank.
When to stop selling hours
Hourly billing has a structural flaw: the better a firm gets at its work, the less it earns for it. A team that has run the same marketing audit forty times does it in half the hours, and if it bills by the hour it has halved its revenue for the same outcome. Paul Graham’s Do Things that Don’t Scale calls consulting the canonical example of work that does not scale, and draws the line at the moment a client starts paying by the hour. A services business is not a failed startup. But it is worth knowing which one you are building, because they are priced and valued differently.
Productising means selling a fixed scope for a fixed price, and keeping the efficiency gain. The audit sold at ₹1.5 lakh that took 90 hours the first time earns ₹1,667 an hour; when the team does it in 45 hours it earns ₹3,333. Run four tests before making the change. Is more than one client buying something close to the same scope? Can the scope be written in a paragraph a client can sign without a call? Is there a step that can be templated or automated, so that the hours fall each time? And will the client pay for the outcome rather than ask for the timesheet? Four yeses mean the deliverable is a product. Two or fewer mean it is still consulting, and should be priced as such.
The end of that road is software, where the margin moves from the 30s and 40s of a services firm to the 65 to 70 per cent that Bessemer’s Scaling to $100 Million reports as the cloud average. Bill Gurley’s All Revenue is Not Created Equal explains why that shift changes what a business is worth: revenue saddled with large variable costs generates less cash per rupee and is priced accordingly. A firm moving from hours to products should report the two kinds of revenue separately, so that the better one is visible.
The monthly services review
On the third working day of each month, once timesheets are locked and invoices raised, compute five numbers by person and for the firm: available hours, utilisation, realisation, realised rate and gross margin per head. Then compute break-even utilisation against the current overhead and put the gap between it and actual utilisation at the top of the page. Mark every person below 60 per cent utilisation and every project below 85 per cent realisation, and ask why for each one.
Decide three things from the answers. Which written-off hours become a change order next month? Which bench hours are planned against signed work and which are hope? And which deliverable sold more than twice this quarter at roughly the same scope, and could be priced as a product from the next proposal? Write the decisions down with an owner. A services firm improves by a point of utilisation and a point of realisation at a time, and only if somebody looks.
Nothing here is legal, tax or investment advice. The Pune agency is illustrative; the Infosys figures are as published for the quarter to 31 March 2026.
Sources
- Infosys Limited, Fact Sheet, Q4 FY2026 (quarter ended 31 March 2026) — Utilisation 83.0 per cent excluding trainees and 79.7 per cent including; revenue ₹46,402 crore, cost of sales ₹32,058 crore, gross profit ₹14,344 crore; 3,28,594 employees.
- Paul Graham, Do Things that Don’t Scale, July 2013 — Consulting as the canonical example of work that does not scale; the line crossed when clients pay by the hour.
- Jeff Jordan, Anu Hariharan, Frank Chen and Preethi Kasireddy, 16 Startup Metrics, Andreessen Horowitz, August 2015 — Gross profit should include all costs of manufacturing, delivery and support of a product or service.
- Mary D’Onofrio, Ethan Ding and Atlas editors, Scaling to $100 Million, Bessemer Venture Partners, September 2021 — Cloud gross margin averages 65–70 per cent; services costs including customer success sit in COGS.
- Bill Gurley, All Revenue is Not Created Equal: The Keys to the 10X Revenue Club, Above the Crowd, May 2011