पाठशाला Pathshala · दल Dal, The team · Lesson 12 · Build

Compensation bands for a startup that cannot pay like Google

A startup that cannot pay top of market can still pay fairly and predictably. How to build salary bands by level from Indian market data, set the cash and equity mix, and keep the bands honest every year.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

Rows of tea bushes run in bands across green hills in Munnar under a clear blue sky. — Photograph: Jebakumar Samuel · Pexels

A startup that cannot pay like Google can still pay fairly, predictably and defensibly. That needs bands: a range of cash for each level, a rule for where in the range each person sits, and an equity grant that closes part of the gap with the market. Without them every salary is a negotiation, and by the tenth hire the company has ten private deals that everyone will eventually compare.

This lesson builds bands from levels, sources Indian market data, sets the midpoint, the spread and the cash-and-equity mix, and closes with the annual review that keeps the bands honest.

Why bands before the fifth hire

The case for bands is not bureaucracy. It is that pay decided one hire at a time, under pressure, drifts. The candidate who negotiated hard earns more than the one who did not, for the same work. The hire made in 2024 earns less than the 2026 hire at the same level. Each gap is small, and each is discovered, usually at the worst moment. Bands replace a negotiation over a number with a conversation about a level, which is a conversation about the work.

Bands also make the equity conversation possible. A company that knows its cash is fifteen per cent below market at a level can say so, show the grant that compensates and be believed. A company with no bands can only say “we pay competitively”, which every candidate has heard and none believes.

Levels first, numbers second

Write five levels before any rupee figure. For a team under thirty, five is enough. Level one, entry: does defined work with supervision. Level two: works independently on defined problems. Level three, senior: owns an area and decides how its problems are solved. Level four, lead: sets direction for an area and makes others better. Level five: heads a function. Write two or three sentences per level on scope, autonomy and impact, the same for every function, so that a level-three engineer and a level-three salesperson are comparable in what the company expects of them.

GitLab, which publishes its compensation calculator and the method behind it, sets pay as a benchmark for the role times a location factor times a level factor: junior at 0.8 of the benchmark, intermediate at 1.0 and senior at 1.2. Buffer publishes every salary and the formula behind it: a market benchmark for each role at a percentile the company chooses, adjusted by a cost-of-living band. Neither is a template for an Indian startup of fifteen people. Both show the structure that works: one benchmark per role, a factor per level, and a formula everyone can read.

Finding Indian market data that is not guesswork

Market data comes from four places, in falling order of reliability. Paid salary surveys from the firms that run them in India, which collect pay by role and level from hundreds of employers: Aon’s 2026 survey, its thirty-second edition, covered more than 1,400 organisations across forty-five industries. Startup-specific compensation reports from advisory firms and investors, some free to portfolio companies. Your own offer log: every offer made, at what number, and whether it was accepted; ten offers are a dataset. And self-reported salary sites, useful for a range and dangerous for a point.

Eggplants and gourds piled on a vegetable stall at a New Delhi market. — Every stall in the mandi knows the going rate by midday because it checks the market every morning. Salary bands need the same habit. Photograph: Switabh Shrivastava · Pexels

Two sets of numbers frame the bands. Aon projected Indian salaries rising 9.1 per cent in 2026 after 8.9 per cent in 2025, with technology platform and product companies at 9.4 per cent, and attrition falling to 16.2 per cent in 2025 from 18.7 per cent in 2023. A band set today is about nine per cent behind the market in a year unless it moves. And xto10x’s report on startup compensation found leadership pay rising twenty to thirty per cent at each stage of a startup’s growth, with variable pay around fifteen per cent of the package in functions such as engineering, finance and HR, and twenty-five to fifty per cent in sales, marketing and growth.

Use the surveys for the market midpoint at each level and city. Use the offer log to check it. When the two disagree, the offer log is telling you what the survey cannot: what people will actually accept from your company, at its stage, with its equity.

Setting the midpoint and the spread

Each band has a midpoint and a spread. The midpoint is what a fully competent person at that level is paid; the spread is the room below for someone new to the level and above for someone outgrowing it. GitLab’s calculator uses a forty per cent spread, twenty per cent either side of the median, and notes that spreads up to fifty per cent are common. The step between levels, the gap from one midpoint to the next, is set so that adjacent bands overlap: a strong level two can earn what a new level three earns, but the level-two band should not reach the level-three midpoint.

The decision particular to a startup is the cash position: what share of the market midpoint you pay in cash. A company that pays eighty-five per cent of market and says so, with a grant to close the gap, is making a clear trade. Set your numbers in the figure: the market midpoint at level three for one function and city, the step between levels, the spread, and the share of market you pay.

Read the overlap first. With a twenty-five per cent step and a twenty per cent spread, half of each band overlaps the next: a strong performer at one level earns what a new joiner at the level above earns, which is what the step is for. Narrow the step to ten per cent and the overlap rises to four-fifths, so a promotion barely changes pay. Widen it past fifty per cent and a gap opens that only a promotion can cross, which turns every raise into an argument about level.

Pay below market if you must. Pay it the same way to everyone at a level, say so, and show the equity that closes the gap.

The cash and equity mix by level

Equity is the part of the package that scales with the outcome, and it should rise as a share of pay with seniority, because senior people influence the outcome more and can usually carry more risk. Index Ventures’ Rewarding Talent handbook gives a Series A framework that sets the grant as a value equal to a share of base salary: 75 per cent for director-level engineering, product and business development roles, 50 per cent for senior and 33 per cent for individual contributors, with sales and customer success at 33, 10 and 5 per cent because more of their pay is variable cash. For a small all-employee grant it suggests something like 5 per cent of base.

Translate that into the bands. A level-three engineer at a ₹24 lakh market midpoint, paid eighty-five per cent of market, earns ₹20.4 lakh in cash and receives a grant worth about ₹10.2 lakh at the last round’s price on the senior line of the framework. Vested over four years that is about ₹2.5 lakh a year at today’s price, a little under the ₹3.6 lakh a year of cash forgone. The honest pitch says the grant beats the cash only if the company is worth more than it is today, and by how much. The [first engineer lesson](/library/hiring-first-engineer-below-market) has the figure that makes that case; the [ESOP lesson](/library/esop-design-pool-grants-strike-price) has the pool the grants must fit inside.

Raises, exceptions and the annual review

Three rules keep bands honest. Every offer lands inside the band for its level; an offer that needs to go above the band says the level or the band is wrong, and the fix is made for everyone rather than for one candidate. Every person’s position in their band, the compa-ratio, is known to them and to their manager and moves with performance: below the midpoint while learning the level, at it when fully competent, above it when ready for the next. And the bands move once a year with the market, before individual raises are decided, so that someone whose pay rises only with the band has kept their place rather than had a raise.

Exceptions will happen: a scarce specialist, a counter-offer worth matching. Record each one in the same sheet as the bands, with the reason and the date it will be reviewed. An exception log with three entries is a company making deliberate choices. One with fifteen is a company without bands.

The compensation calendar

Once a year, in the same month: refresh the market data from at least two sources and the offer log; move the bands by the market change; place each person in their band from their performance review; decide raises and equity refreshes in one sitting, so cash and equity are decided together; and tell each person their level, their band and where they sit in it. Every quarter: review the offer log and the exception log, and check the acceptance rate by level, because a level where most offers are declined has a band below its market. At every offer: the level first, from the scorecard, then the band, then the position in it, then the grant from the table. Once the company passes fifteen people, write all of it on one page the whole team can read.

Survey figures are averages across employers; your market is the people you are trying to hire. Nothing here is legal, tax or investment advice.