How to Build Salary Bands for the First Time
The hard part is not drawing the ranges. It is placing the people you already have, and deciding what you will do about the ones who fall outside.
Every company reaches a point where salaries can no longer be explained. Each one was defensible when it was set — a strong candidate, a counter-offer, a promotion that could not wait — and together they no longer form a system anybody could describe out loud.
Building bands is how that gets repaired. The design is the easy half. The half that decides whether it works is what you do with the people already in the building.
Start with levels, not with numbers
Bands attach to levels. Without a level structure there is nothing for a range to be a range of, and you end up with per-title bands that multiply every time someone invents a title.
If levels do not exist yet, that is the first project, and it is a different one — see career frameworks for scale-ups. Three or four levels per track is enough to start.
Decide the positioning before you look at data
Market positioning is a commercial decision, and making it before opening a benchmark report stops the data from making it for you.
The decision has three parts: which market you are comparing against (your country, your sector, your funding stage, or the companies you actually lose people to — these are rarely the same set), which percentile you target, and whether that varies by function. Paying at the median for most roles and above it for two or three genuinely scarce ones is a legitimate and common answer, as long as it is written down.
Construct the ranges
| Choice | A workable default | What it controls |
|---|---|---|
| Range width | Around 20% from bottom to top, wider at senior levels | How long someone can grow inside a level before promotion is the only remaining move |
| Overlap between levels | Some, but the midpoints must be clearly apart | Whether a promotion is visible in pay. No overlap makes promotions expensive; heavy overlap makes them meaningless |
| Entry point | Lower half of the range for a new hire meeting the level | Leaves room to reward growth. Hiring people at the top of a band is how bands erode |
| Number of geographies | As few as you can defend | Every additional geographic band multiplies the maintenance and the arguments |
Place people on level before you look at their pay
This sequence matters more than it sounds. Placing people while their current salary is visible produces levels that quietly justify the pay, which is the problem you started with, now formalised.
- Place everyone on a level using the level definitions alone, with salary hidden.
- Have a second person review the placements independently. Disagreements are information about the definitions, not just about the people.
- Only then overlay pay, and produce the list of people above and below their range.
- Decide the approach for each group before communicating anything.
The outliers are the real decision
Two groups will fall outside, and how you handle them is what the company will remember about the whole exercise.
Below range. These need a correction and a date. A framework that identifies underpaid people and does nothing is worse than no framework, because it converts a private inequity into a documented one. If the full correction is not affordable at once, a staged plan with committed dates is defensible. Silence is not.
Above range. Almost always held rather than reduced: cutting pay to fit a new model damages trust far beyond the individual. The usual approach is to hold the salary, apply smaller or no increases until the range catches up, and say so plainly to the person. What breaks this is not the decision but the surprise.
A pay framework is judged by what it does for the people it finds are underpaid. Everything else is administration.
The rules that keep it alive
Bands decay through exceptions, not through design flaws. Three rules prevent most of it:
- One named approver for any offer outside a range, and the reason recorded in writing.
- A review of ranges on a fixed annual cadence, so they move deliberately rather than through hiring pressure.
- A quarterly look at pay dispersion within each level. Widening dispersion is the earliest signal that the framework is being worked around.
What to publish
Levels and promotion criteria should be visible to everyone: they exist to answer a question people currently cannot get answered. Whether to publish the ranges themselves is a separate decision, and both answers are defensible.
The thing that is not defensible is publishing before managers can explain the framework in their own words. The most common failure of a pay project is not the model — it is announcing transparency and leaving managers to field questions nobody briefed them on.
This is the work covered by compensation principles and career frameworks, and it is materially cheaper done as one project than as two a year apart.
Co-founder — People & Management
Ana Reis
Builds the People foundations of growing companies — and stays with them until they work day to day.
More from Ana