How to Build Salary Bands and a Pay Structure, Step by Step

What is a pay structure and why build one?

A pay structure is a set of grades, each with a salary range, that groups jobs of similar size and links them to the market. Each range has a minimum, a midpoint and a maximum. The midpoint is your view of the market rate for someone fully proficient in jobs of that size, at your chosen market position.

Organisations usually build one when pay set offer by offer stops making sense: two people in the same job are far apart, titles mean little, and nobody can explain why one person earns more than another. A structure does not decide pay for you. It makes decisions consistent, shows who sits outside the range, and tells you what it would cost to fix.

The table below sets out the steps, what each produces and the mistake it prevents.

What should a reward philosophy say?

Before any spreadsheet, agree in writing what you pay for (usually job size, growth in skill and contribution, and the market for the job), whether you lead, match or lag the market and at which percentile, which organisations you compare with, how pay is split between fixed and variable, how pay grows, and what people can see.

It helps to describe the whole offer, not just salary. WorldatWork's Total Rewards Model names five elements: compensation, benefits, well-being, careers and recognition. If your benefits, flexibility or development offer is genuinely strong, you may choose to match the market on base pay rather than lead it.

How do you level and evaluate jobs?

Group jobs into families by the kind of work, then define tracks (for example support, professional and management) and levels described on the same dimensions everywhere: scope of impact, complexity, autonomy, influence, and expertise or people leadership. A dual career ladder lets senior experts reach the same grade as managers.

Then size each job with job evaluation. Ranking puts whole jobs in order; classification slots them into written grade descriptions; point-factor scores defined factors such as knowledge, problem solving, impact, communication and responsibility, and adds the points. Point-factor takes longer but gives a reason for every score, which matters for equal pay for work of equal value. The International Labour Organization's step-by-step guide to gender-neutral job evaluation is a useful check on your factors.

• Score the job as it should be done, not the person in it
• Use a panel of three or more, scoring alone before agreeing
• Score eight to ten well-understood benchmark jobs first
• Check that factors credit skills common in jobs mostly done by women

How do you price jobs against the market?

Choose sources that match your jobs, comparators and pay definitions. Paid salary surveys usually give the most reliable picture; free job-board and crowd-sourced data can show direction but often mix job sizes, locations and base pay with bonus or commission. Match on job content and level, not title.

Survey data is a snapshot taken months earlier, so age every figure to the date your ranges start: multiply by (1 + annual market movement) raised to the number of years between the data date and the target date. At 3.5% a year, a figure from 18 months earlier rises by about 5.3%. Then blend the aged figures with weights that reflect quality, and apply your policy line: zero to match the market, a positive figure to lead.

For example, three sources for a principal engineer might give 80,400 (data effective a year before the target date), 79,900 (nine months before) and 70,500 (eighteen months before, from free job-board data). Aged at 3.5% a year they become about 83,200, 82,000 and 74,200, and a 50/30/20 blend gives a market reference of about 81,050. The oldest, lowest-quality figure has the least weight.

When sources disagree sharply, for example by more than 20%, check the matches and pay definitions before trusting any of them. A single free source is not a market price.

How do you calculate salary ranges from a midpoint?

Average the market references for the benchmark jobs in each grade, choose midpoints that follow a smooth line, then choose a range spread: the gap between maximum and minimum, divided by the minimum. Narrower spreads (around 30% to 40%) suit jobs learned quickly; wider spreads (50% to 60% or more) suit senior and specialist jobs where full proficiency takes years.

With a midpoint and a spread, minimum = 2 x midpoint / (2 + spread) and maximum = minimum x (1 + spread). A grade with a midpoint of 52,000 and a 50% spread runs from 41,600 to 62,400. Then check midpoint progression between grades and the overlap with the next grade up.

Should you use narrow grades, broad bands or job families?

The CIPD's factsheet on pay structures and pay progression, updated in November 2025, describes the main types. Narrow-graded structures usually have ten or more grades. Broadbanding uses four or five wide bands, which gives flexibility, although some employers have added structure back inside the bands, partly because of equal pay concerns. Job-family structures group jobs by occupation, usually with six to eight levels and separate ranges, which can help where one family's market runs ahead.

A practical middle path for a small or mid-size organisation is one set of grades for everyone with a few named market-premium jobs, approved and reviewed each year, rather than moving a whole grade because one family's market is hot.

How do you place employees and cost the fix?

For each person, calculate the compa-ratio (salary divided by the range midpoint) and the range penetration ((salary minus minimum) divided by (maximum minus minimum)). People new to a job usually sit low in the range; a long-serving strong performer well below the midpoint deserves a look.

For example, someone paid 47,000 in a grade with a minimum of 41,600, a midpoint of 52,000 and a maximum of 62,400 has a compa-ratio of 0.90 and a range penetration of 26%: low in the range, which is normal for a newcomer and worth a question for someone who has done the job well for years.

Anyone below the minimum is a green circle; anyone above the maximum is a red circle. Cost the fix in two scenarios: bringing everyone to the minimum, and bringing everyone to a floor such as a 0.90 compa-ratio. Add employer on-costs and agree the phasing with finance. Do not cut red circles: keep base pay and pay merit awards as lump sums until the range catches up.

Before you finalise anything, compare average compa-ratios by group, such as gender, job family and location, where your policy and local law allow you to hold the data. Compare like with like by grade, ignore groups too small to read, and treat a gap as a reason to investigate rather than proof.

How do you run the annual pay review on a structure?

Set separate budgets for merit, promotions and market adjustments. Build a merit matrix that gives guideline increases by performance rating and compa-ratio band, so strong performers low in their range move faster towards the midpoint. Test the matrix against the budget before managers see it, ask for reasons when managers depart from the guideline, calibrate spend and differentiation across managers, check outcomes by gender and only then write letters.

If you need to measure your own starting point in pay decisions, an AssessAll assessment such as the Pay and Reward Decision Assessment for Managers and HR Teams gives a baseline, and a Jobulary individual development plan turns the results into actions. The Compensation Structure Builder Kit from Bodhih Training includes the workbook that builds the ranges and the simulator that runs the review.

Where does AI fit in compensation work?

AI assistants are useful for cleaning benchmark data, spotting salaries that sit far from peers, explaining formulas in plain words and drafting FAQs and letters from templates. They should not decide anyone's grade, range placement or increase. A person decides, can explain the decision without the tool, and is accountable for it.

Use only approved tools for employee data, remove names and sensitive details, and never paste individual pay alongside gender or other group data. Rules on AI in employment decisions and on pay transparency are changing in several regions, so check the current position where you employ people.