Salary structure design: build a pay structure with salary grades and ranges

Think of it as three layers stacked on each other. Levels describe scope, so a Senior does broader work than a Mid. Grades group jobs from different departments whose scope is comparable, so a Senior Engineer and a Senior Accountant can sit on the same rung even though their market rates differ. Ranges attach dollars to each rung: a minimum, a midpoint and a maximum. Miss any one layer and you get the classic small-company failure mode, where pay is decided per hire and nobody can explain it a year later.

The vocabulary trips people up constantly. A pay grade is the rung; a salary band is the minimum-to-maximum range hanging on it; a posted salary range is the slice of that band you publish in a job ad. If those three still blur together, our breakdown of pay grades vs salary bands separates them properly. The level definitions themselves come from a job leveling framework, which is genuinely the step to do first.

Nearly every US salary structure is a variation on four models. The choice is mostly a trade between precision and administrative load, and it should follow how your company actually makes pay decisions rather than what a textbook prefers.

The grade counts and spreads above are the ranges most commonly cited in US compensation practice rather than a survey census, so treat them as starting points. Market-based is the default recommendation for a team under 200: it needs the fewest internal judgment calls, it survives contact with a candidate who has three other offers, and it produces posted ranges you can defend to a state labor department because every number traces back to a data source.

Broadband is the model that most often disappoints small companies. Wide bands look flexible on a slide, but with a 100 percent spread and no midpoint discipline, two people doing the same work can legitimately sit $60,000 apart, and that is exactly the gap you will be asked to explain in a pay equity review.

This is the order that works. Steps one and two are internal and slow; steps three through five are data work that software does in an afternoon; steps six and seven are the ones companies skip and later regret.

Three numbers hold a structure together, and all three are checkable in a spreadsheet in ten minutes. Range spread is how wide one grade is: maximum divided by minimum, minus one. Midpoint progression is how far apart two grades sit: the next midpoint divided by this one, minus one. Overlap is how much of the lower grade's range is also inside the higher grade's range.

Read the table as guardrails, not law. The pattern that matters is the direction: spreads widen as you go up, progression widens as you go up, overlap shrinks as you go up. If your executive grades overlap by 50 percent, the ladder has effectively collapsed and a promotion carries no money. If your entry grades barely overlap, you will be forced into a promotion every time someone gets good at their job.

Where an individual sits inside a range is measured with compa ratio (salary divided by midpoint) or range penetration (position between minimum and maximum). Run both across the whole company once the structure exists, because that single view is what surfaces underpaid tenure and overpaid new hires.

Here is what the output looks like for a US software company of about 45 people paying at the market median, with a 35 percent spread on professional grades and a 12 to 18 percent midpoint progression. The midpoints are illustrative round numbers, not benchmarks for your company; yours come from your own roles, markets and seniority mix.

Two things to notice. Grades overlap on purpose: the top of G3 ($109,000) sits above the bottom of G4 ($95,000), so a strong mid-level engineer can out-earn a newly promoted senior, which is correct. And the whole grid is generated from six benchmarked midpoints, not from sixty individual negotiations. That is the practical reason to benchmark first and grade second.

A useful audit habit: once a quarter, list everyone below their range minimum and everyone above their maximum. Those two lists are your entire structural problem, and they are usually short enough to fix in one budget cycle. Everything else is noise.

A dozen US states now require a good-faith pay range in job postings, and the range you publish has to be one you actually intend to pay within. Companies without a structure end up inventing a range per posting, which is both a compliance risk and a retention problem the moment a current employee reads the ad. With a structure, the posted range is just a slice of the grade the job sits in, and it is defensible because it traces back to benchmark data and a documented spread. The mechanics of turning a band into a compliant ad line are on our job posting salary ranges page, and the state-by-state rules live in the pay transparency hub.