Your first salary bands at 40 people
Simple bands now beat perfect bands at 200.
Short answer: Salary bands for startups at around 40 people should be simple: A few levels, one band per level for each group of roles, room for jobs that are still changing and a date to rebuild them. Their job is to make the next twenty offers consistent and explainable. They do not need to be the structure the company will need at 200.
Forty people, forty deals
Picture a 40-person software company in Chicago that has just raised its next round. The plan is to hire about 25 people in the coming year. So far every salary has been set offer by offer, mostly by the two founders. The first engineers joined three years ago for what the company could afford, plus equity. The last three engineers joined this year at noticeably higher salaries, because that is what it took to close them. A customer success lead negotiated hard and is paid more than her manager. Two early employees carry the title "Head of" with no team.
Then the first job ad of the new plan goes to the recruiter, who asks for the pay range. Illinois requires one in the posting at this size. Nobody has a range. The founders have a spreadsheet of current salaries and a strong sense of who is "worth more".
The new head of people does what feels responsible. She downloads a leveling framework from a large tech company, with twelve levels and separate tracks for managers and individual contributors. She starts building bands for every cell. Three weeks later the model has 96 bands, most of them empty. The engineering lead disputes where his team sits. The founders want the bands to fit the people they already have. The job ad is still waiting for its range.
The company did not need a structure for 200 people. It needed a simple set of bands that could price the next twenty offers and stand up when a current employee asks how their pay compares.
This company is a composite of situations common at this size. It is not a client and the details are illustrative.
Why first bands are hard at this size
A salary band is the pay range for a level of work: A minimum, a maximum and usually a midpoint the company aims for. At 40 people the idea is simple. Getting the first set right is not, because two good goals pull against each other.
The first goal is to be ready for growth. A company planning to double wants a structure that will still work when there are managers of managers, specialist roles and staff in several states or countries. Building that once feels efficient.
The second goal is to be useful now. The company has a handful of real roles, most of them held by one person each. Jobs are still being invented. Pay is a record of past deals, each made under different conditions. A structure built for the future has to describe roles that do not exist yet, while the ones that do exist are still moving.
The market makes this harder. Startup salaries have moved in the last two years. They have not moved evenly. Across companies on one equity management platform, the median salary for individual contributors rose 6.4% between January 2024 and February 2026, while the median for AI and machine learning engineers rose 9.1% [1]. A company that hired its first engineers in 2023 and its latest ones this year is paying two different markets for the same work.
Size works against precision too. At 40 people most roles have one holder. A band built around one person tends to become that person's salary with a range drawn around it. And the market data a young company can afford often has thin coverage for hybrid startup roles, such as a "founding product engineer" or a first operations hire who also runs finance.
The law sets the timing. In several US states, a 40-person company is already required to put a pay range in its job ads. Current staff can also ask what the range for their own job is [4] [5] [6] [8]. A company without bands ends up inventing a range for each ad. Those ranges then become the structure, one posting at a time.
Five traps with first salary bands
Most of these traps come from trying to solve the problem of a later stage now, or from putting the problem off until a later stage arrives.
"Copy the structure of the company we want to be"
What it looks like: The company adopts the leveling and band structure of a much larger firm: Ten or twelve levels, separate manager and individual tracks, a band for every family at every level.
Why smart teams fall into it: It looks rigorous. It promises to avoid a rebuild later. Large companies' frameworks are easy to find and many early employees came from those companies and like the familiar shape.
What it costs: Most of the structure is empty, so there is nothing to test it against. Levels end up separated by distinctions nobody at this company can observe, such as the difference between a "senior" and a "staff" engineer when there is one engineering team of six. People are placed on fine gradations that cannot be explained, which invites argument. The pay steps between levels get small, so a promotion means little money and a lot of debate. And the structure needs constant exceptions, because a company of 40 people changes faster than a structure designed for 400. The rebuild the company hoped to avoid happens anyway, usually at about 100 people. By then it has commitments to undo. Our guide on how many job levels you need looks at that choice in more detail.
"Draw the bands around the people we have"
What it looks like: The company lists everyone's salary, groups people by role and sets each band to cover the lowest and highest paid person in it. Nobody falls outside a band, because the bands were drawn to make sure of that.
Why smart teams fall into it: It avoids awkward conversations. It feels like a description of reality rather than a judgment. It costs nothing on day one.
What it costs: It turns every past deal into policy. The customer success lead who negotiated hard sets the top of the band for her role. The engineer who joined for equity in a lean year sets the bottom of his. The bands become very wide and very hard to defend, because their shape reflects bargaining power and timing rather than the work. It also hides the problems the company most needs to see. A first set of bands is the cheapest moment to find the people who are paid well below their peers for the same work. If the bands are drawn to include them, the company loses that moment. In the EU, a worker can ask for average pay levels by sex for people doing the same work or work of equal value, whatever the employer's size [9]. A band drawn around an unexplained gap does not explain it.
"Wait until we are bigger"
What it looks like: Bands are on the list for "after the next round" or "when we hit 100". Meanwhile each offer is set by the hiring manager and a founder, based on what the candidate asks for and what similar offers paid.
Why smart teams fall into it: At 40 people, a founder can still see every salary. Bands feel like bureaucracy that slows hiring. The company is busy.
What it costs: The structure gets built anyway, offer by offer, with no one choosing its shape. Each new hire is priced at this year's market, while earlier hires stay where their own offer put them. Within a year the company has people hired later earning more than people doing the same work who joined earlier. The fix then costs real money, in a lump, at the moment the company least expects it. Meanwhile, the posting laws do not wait for size. New York covers employers with four or more employees [4]. California, Washington and Illinois start at 15 [5] [6] [7]. Massachusetts has required ranges in postings since October 2025 for employers with 25 or more employees in the state [8]. Washington counts employees outside the state once the company has one Washington-based employee [6]. Illinois counts employees inside or outside the state toward its 15 [7]. A company that waits for bands still has to post ranges. It just posts them without a structure behind them.
"Benchmark every role precisely before we start"
What it looks like: The company buys or borrows market data and tries to match every role, including the hybrid ones, before setting any band. Weeks go into debating whether the first operations hire is a "finance manager" or an "operations generalist".
Why smart teams fall into it: Precision feels like defensibility. Nobody wants to set a band that turns out to be wrong.
What it costs: The precision is mostly false. A hybrid role matched to a single survey job is a guess with a decimal point. Small samples for niche roles move a lot from one data cut to the next. The decision that matters most, where the company wants to sit against the market, often gets made by default while the team argues about matches. And the delay has a price: Every offer made while the model is being perfected is still made the old way. A first set of bands needs to be sound for the roles the company actually hires, not exact for every role it has. Our guide on whether market data can set your salary bands covers the choices the data cannot make.
"Set them once and leave them"
What it looks like: The first bands are approved, put in a spreadsheet and used for every offer from then on. Nobody owns them. There is no review date.
Why smart teams fall into it: The first build took effort and settled some arguments. Reopening it feels like inviting the arguments back.
What it costs: Startup bands go stale faster than most. Roles split as the company grows: "Engineer" becomes front end, back end and data. Markets for some skills move faster than others, as the AI and machine learning premium shows [1]. The company may open in a new state or country, with a different pay level and different posting rules. Bands that are not reviewed start to lose offers. Managers then make exceptions, the exceptions become the real bands and the official ones fall out of use without anyone deciding to retire them. Among organizations with pay structures in one 2025 regional survey, 64% adjust them every 12 months [2]. A startup growing fast usually needs to look sooner than that, or at a clear trigger, such as a funding round or crossing a headcount line.
What a sound first set of bands looks like
A sound first set is not a smaller version of a large company's structure. It is a different thing with a different job. In our view it has these qualities.
- It covers everyone on one page. A founder or the head of people can see the whole structure at once. If it needs a manual, it is too big for 40 people.
- It has only the levels the company can tell apart. Each level describes a difference in the work that people here can actually see. If two levels cannot be told apart in this company today, they are one level for now.
- It groups roles that share a market. Roles are grouped by the labor market they compete in, not by the org chart. A band per group, rather than per job title, keeps the structure small while roles keep changing.
- It has room for roles still being shaped. Bands are wide enough to hold someone growing inside a role that is still being defined, but not so wide that the range says nothing on a job ad. Our guide on how wide a posted range can be covers where that line sits.
- It shows the outliers instead of hiding them. People above or below their band are listed, each with a reason and a plan. The list is short and someone owns it.
- It has an anchor the company chose. Where the company aims to sit against the market, by group of roles, is written down. So is the market data used and its date.
- It has an exception rule. When an offer needs to go above the band, who can approve it and what gets recorded is decided before the first such offer.
- It has a rebuild date. The first set is labeled as a first set. The date or the trigger for the next full review is part of the design.
- It can be explained to an employee who asks. A manager can tell someone what their band is, why it sits there and what would move them within it.
None of this is a recipe. Each quality is a test a founder or a board member can put to the bands the company has.
A short test helps. Take the last three offers the company made. Would the new bands have produced roughly the same numbers, with any difference explained? Then take the three longest-serving people in the same roles. Where do they sit? If the bands fit the new hires and leave the early employees near the bottom with no plan, the company has priced the market but not its own history.
The calls only you can make
A first set of bands puts several founder decisions in writing for the first time.
How to treat early deals. Some early employees took less cash for more equity. Others negotiated well. The company has to decide whether bands reset cash pay toward a common standard over a set period or whether early deals are honored as they are and fade over time.
Where to sit against the market. A young company may decide to pay below the market median in cash and make up the difference in equity, or to pay at the market to compete for people who discount equity. Either can be sound. It should be one decision, not twenty separate ones.
Whether scarce skills get their own band. Roles such as AI and machine learning engineering are moving faster than the rest of the market [1]. A separate band or premium keeps the main bands honest. It also creates two classes of engineer, which some teams accept and others will not.
How much to share. Job ads will show ranges in many states. Whether current employees see the whole structure, their own band or only what the law requires is a culture decision. A 2025 study found that 61% of organizations had salary ranges they did not share with employees [3].
How firm the edges are. Some companies treat the band maximum as a hard limit and lose the occasional hire. Others allow exceptions and track them. The choice shapes how much the bands will mean in a year. Our guide on pay above the band maximum covers the honest options when someone is already outside.
The law and the numbers
- Team sizes. Across companies on one equity management platform, the average headcount at Series B fell from 53 in 2023 to 45 in 2025. Companies that raised a seed round in 2025 averaged 5.8 employees [1]. On those figures, a 40-person startup is often around the Series B stage.
- Startup salaries. On the same platform, median salaries for individual contributors rose 6.4% between January 2024 and February 2026 and those for AI and machine learning engineers rose 9.1%. Median salaries for new hires in the first quarter of 2026 were about $177,000 in product and $167,000 in engineering [1].
- Formal structures. In a 2025 survey of 196 Michigan organizations, 41% of them with 100 or fewer employees, 38% had no formal base pay structure. Of those with one, 64% usually adjusted it every 12 months [2].
- Ranges and sharing. A 2025 global study found that 61% of organizations used salary ranges without sharing them with employees and 8% did not use ranges at all [3].
- US posting laws. New York requires a compensation range in postings from employers with four or more employees [4]. California requires a pay scale in postings from employers with 15 or more employees, at least one of them in California. Employees there can ask for the pay scale for their own position [5]. Washington's 15-employee threshold counts employees outside the state once the employer has one Washington-based employee [6]. Illinois counts employees inside or outside the state toward its 15 [7]. Massachusetts has required wage ranges in postings, to applicants and to current employees on request since 29 October 2025, for employers with 25 or more employees in the state [8].
- EU law. Under the EU pay transparency directive, applicants are entitled to information on the starting pay or its range, whatever the employer's size. Workers can ask for their own pay level and average pay levels by sex for the same work or work of equal value. Employers must make the criteria for pay and pay progression available to workers, although member states may exempt employers with fewer than 50 workers from the pay progression part [9].
This is general information, not legal advice. Thresholds and dates change. Check the rules where you hire.
Our view
We think a company of about 40 people should have bands. They should also be simple. A few levels the team can tell apart, one band per level for each group of roles that shares a market, a written anchor and a written exception rule. That is enough to price the next twenty offers consistently and to answer an employee who asks.
The ambition to build the final structure now is understandable. It usually costs more than it saves, because the company will look different at 100 and different again at 200. A first set that is labeled as a first set, with a rebuild date, is easier to change than one that pretends to be permanent.
We would argue hardest against drawing the first bands around current pay. It is the one choice that is hard to undo. Once past deals become policy, every later correction looks like a pay cut or a special favor.
Where we help
Our pay band work builds a first set of bands sized for where the company is now: The levels, one band per group of roles, everyone placed with outliers flagged and the wording for job ad ranges. We use the market data you have or buy. The starting point is the Pay Band Sprint, at a fixed price by number of roles. If a conversation would help first, Book a free call.
Sources
- Carta, State of Startup Compensation H2 2025 (4 May 2026), checked 2026-10-25
- American Society of Employers, 2025 Pay Administration Survey (1 August 2025; 196 Michigan organizations), checked 2026-10-25
- Aon, 2025 Global Pay Transparency Study, checked 2026-10-21
- New York Labor Law § 194-b, subdivisions 1 and 6 (New York State Senate), checked 2026-10-25
- California Labor Commissioner, Equal Pay Act FAQ, checked 2026-10-24
- Washington L&I, Administrative Policy ES.E.2, Job Posting Requirements (issued 3 December 2025), checked 2026-10-24
- Illinois Department of Labor, Equal Pay Act salary transparency FAQ (questions 3 and 4), checked 2026-10-24
- Massachusetts Attorney General's Office, Pay Transparency in Massachusetts, checked 2026-10-25
- Directive (EU) 2023/970 on pay transparency, Articles 5, 6 and 7 (EUR-Lex), checked 2026-10-25