Compensation bands that survive scrutiny

60-Second Summary

Band width (max ÷ min) typically lands at 30–50% — wider bands enable longer in-level tenure; narrower bands force promotion velocity and predictability.
Midpoint spacing between levels of 10–18% is the modal range — too tight (<8%) and promotions feel meaningless; too wide (>20%) and lateral promotions become rare exceptions.
Compa-ratio (employee pay ÷ band midpoint) is the single most useful comp KPI — target a population mean of 0.95–1.05 with a healthy spread.
Bands drift annually by 4–6% in tight markets; budget for a re-benchmark every 12 months or watch your bands become decorative.

Most compensation problems trace to one of three failures: bands were never built, bands were built once and never maintained, or bands exist on paper but are routinely violated for 'special cases.' All three are fixable.

What a band actually is

A compensation band is the legal and defensible pay range for a single role at a single level in a single geographic market. Min represents new-in-role pay, midpoint represents fully-proficient pay (the market rate for this role), and max represents elite/long-tenured pay for that level.

Designing the band structure

The four design decisions

1
Band width (max ÷ min)
30% for tightly-scoped levels with fast promotion velocity (e.g., L3 engineer). 50% for broader leadership levels where in-level tenure is long (e.g., Senior Director).
2
Midpoint progression (level N+1 midpoint vs level N midpoint)
10–18% per level. <10% trivializes promotions; >20% makes lateral moves into the next level impossible without a big raise that strains budgets.
3
Overlap between bands
Typically 50–70% of band N+1 overlaps band N. Zero overlap forces every promotion to be a large raise; high overlap weakens the level structure.
4
Number of levels
5–8 levels per track is the modal range. More invites comp politics; fewer creates promotion droughts.

Geographic differentials

Geographic strategy choices

Tiered (most common)

3–5 geographic tiers tied to market data
Tier 1: SF/NYC/Seattle/London = 100%
Tier 2: Boston/Chicago/Berlin = 92–95%
Tier 3: Austin/Denver/Lisbon = 80–85%
Tier 4: rest = 70–75%

Location-agnostic

One band for the role globally
Often the SF rate or close
Simple to administer, expensive at scale
Attractive to top talent in low-cost geos
Politically difficult once the company hits 1,000+ people

GitLab, Buffer, and Doist are the canonical location-agnostic case studies; Stripe, Airbnb, and Brex run sophisticated tiered systems. Most companies above 500 employees end up tiered.

The compa-ratio discipline

Compa-ratio range	Meaning	Action
< 0.85	Significantly underpaid relative to band midpoint	Targeted raise to remediate or document why (new hire, performance issue)
0.85 – 0.95	Developing — common for new-in-role	Track; expect natural progression toward midpoint
0.95 – 1.05	At market for fully proficient	Healthy steady state
1.05 – 1.15	Above-midpoint — typically high performer or long tenure	Watch for compression with new hires
> 1.15	Approaching band max	Promote, expand band, or accept they're topped out

Maintenance

Annual market re-benchmark against 2+ data sources (Radford, Mercer, Pave, Levels.fyi for tech)
Quarterly compa-ratio dashboard reviewed at exec staff — drifting compa-ratios are an early warning system
Mid-year refresh for hot-skill roles (AI engineers, ML platform) where the market moves faster than annual cycles
Pay-equity regression at least annually — bands prevent new gaps but don't fix accumulated ones
Documented exceptions process — when, why, who approves, what the limit is

Frequently asked questions

Should we publish our bands?

Mandatory in California, New York, Colorado, Washington on job postings; mandatory across EU by 2027 under Pay Transparency Directive. The trend is one-directional; companies hiding bands now will be required to publish them within 2–3 years. Most companies that publish proactively find recruiting and retention benefits.

What do we do about new hires above existing-employee pay?

Compression is the most-litigated comp issue. The defensible answer: have a clear policy (e.g., 'we will remediate when new-hire pay exceeds existing-employee pay by >5% in same role/level'), audit annually, fund the remediation. The undefensible answer: hope nobody notices.

How do we set bands when we don't have market data?

Pave, Carta Total Comp, and Option Impact have lowered the floor — at 50+ employees most VC-backed companies can access participant-pooled data. Below 50, talk to your investors' comp leads who have visibility across the portfolio.