Engineering comp benchmarking: how to actually price an engineer

60-Second Summary

Engineering comp has three components — base, bonus, and equity — and the mix varies by company stage and level. At senior+ in tech, equity is often 30–60% of total comp; pricing on base alone underprices the role by 2–3x.
Generic comp surveys (Mercer, generalist Radford) systematically underprice tech roles by 20–40%. Use Radford Tech, Pave (private), Option Impact (early-stage), and Levels.fyi (public, candidate-facing) for triangulation.
Build bands on total comp at the 50th and 75th percentiles for your geography and stage. Document the methodology; the band IS the methodology, not a number pulled from a survey.
The HRBP owns the band, the philosophy, the geographic differentials, and the audit trail. The CFO owns the budget envelope. The CTO/managers own the placement of individuals within bands.

Engineering compensation is where HR earns its credibility with engineering leaders. If you can speak fluently about Levels.fyi data, Radford Tech percentiles, equity refresh strategy, and why your senior band is at the 65th percentile not the 50th — you'll be the HRBP a CTO trusts with comp. If you can't, you'll be cut out of comp decisions within a quarter.

The three components of engineering comp

Component	What it is	% of total comp (typical)	Stage dependence
Base salary	Cash, paid biweekly	40–70%	Higher % at late-stage / public; lower % at seed
Bonus	Annual cash, target-based	5–20%	Often 0% at early-stage; 10–20% at public tech
Equity (stock / options / RSUs)	Ownership grant	20–60%	Highest % at seed/Series A; 20–35% at late-stage; 25–40% at public tech with RSU refresh

At seed-stage, total comp might be $140k base + $0 bonus + equity worth 0.3% of the company (typical seed senior eng grant). At Stripe/Datadog/Snowflake post-IPO, the same level role might be $220k base + $30k bonus + $200k/year RSUs vesting over 4 years. Both are 'senior engineer' offers. Both are competitive in their context. Pricing them with the same methodology fails both.

Why generic comp surveys underprice engineers

Mercer, generalist Radford, Willis Towers Watson, Korn Ferry — these are excellent for finance, ops, sales, marketing, HR. They systematically underprice tech roles by 20–40% because their submitter base skews to industries where 'senior software engineer' means an internal IT role, not a product engineer at a top tech company. If you use the Mercer 50th percentile for a senior backend engineer in a tech company, you'll lose every offer to candidates with multiple options. This isn't a minor issue — it's the single most common comp mistake HRBPs new to tech make.

The 'we benchmark to industry median' trap

When a CFO says 'we pay at the 50th percentile of the market,' the question is WHICH market. The 50th percentile of all US software engineers (BLS data) is roughly half the 50th percentile of senior product engineers at well-funded VC-backed startups. Your competitive set for senior IC tech hires is the latter, not the former. Document this explicitly in your comp philosophy.

The four datasets that matter

Source	What it covers	Strength	Weakness	Cost
Levels.fyi	Self-reported total comp at large tech companies	Free; candidate-facing; granular by company/level	Self-reported bias; skews to top-tier comp	Free / $100s for API
Radford Technology Survey	Submitted by 800+ tech companies	Industry standard for tech HR; defensible	Expensive; lags 6 months	$15–40k/year
Pave / Compa	Real-time private comp data via integrations	Live data; granular cuts	Smaller dataset at staff+	$10–30k/year
Option Impact (Aon/Carta)	Early-stage venture-backed startups	The only good source for seed–Series B comp + equity	Sparse at senior levels	$5–15k/year

Triangulate. Don't pick one. A defensible band cites at least two of these sources, plus your own offer-acceptance data (the offers you've made in the last 6 months, accepted vs declined, and the comp delta when declined).

Building a band from raw data

The 6-step band-build

1
1. Define the role + level
'Senior backend engineer, L5, primary location SF Bay Area, fully remote acceptable.' Vague role definitions produce vague bands.
2
2. Pull comp data for matched roles
From each source, pull 25th / 50th / 75th / 90th percentile total comp for the matched role/level/geo.
3
3. Choose your target percentile
Most VC-backed tech companies target the 60th–75th percentile of their competitive set. 'Median' is a losing strategy at senior+ unless you have a strong non-comp value prop.
4
4. Set the band width
Typically the band min is 80% of midpoint, max is 120% of midpoint. So a $200k midpoint band runs $160k–$240k.
5
5. Split into base / bonus / equity
Use your company's mix philosophy. E.g., 'senior band: 70% base, 0% bonus, 30% equity at grant date FMV.'
6
6. Document everything
The band document should be 1–2 pages: methodology, sources, target percentile, mix philosophy, geographic differentials. This document is your defense in any pay equity audit or DOL inquiry.

Geographic differentials

Remote work made geo-differentials one of the highest-stakes policy decisions in tech HR. The major models:

Model	How it works	Pro	Con	Used by
No differential	Same comp regardless of location	Simple; helps low-cost-of-living hires	Expensive in low-cost markets; not market-competitive in NY/SF	37signals, some early-stage
Tiered by metro	3–5 tiers (e.g., Tier 1 = SF/NY, Tier 2 = Seattle/Boston, Tier 3 = Austin/Denver, Tier 4 = remote-US)	Simple, defensible, market-aligned	Cliff effects at tier boundaries	Most VC-backed tech
Cost-of-labor-indexed	Each metro's comp is a % of SF (e.g., Denver = 85%)	Most market-accurate	Operationally complex; demotivating when read as 'punishment for moving'	Google, Meta
Global tiers (for global workforce)	US, Europe, India, LatAm tiers with separate bands	Necessary for international scale	Equity perceptions are hard	Most companies hiring globally

Equity math the HRBP must understand

When you make an offer that includes equity, you must be able to explain what the candidate is being offered in plain English. The three forms:

ISOs (Incentive Stock Options) — early-stage, taxed favorably if held 1+ year after exercise. Candidate must pay the strike price to convert to shares. Risky if the company doesn't IPO.
NSOs (Non-qualified Stock Options) — similar mechanics, less favorable tax. Common at later stage when ISO limits are hit.
RSUs (Restricted Stock Units) — late-stage/public companies. Candidate receives shares (not options) on vest; taxed as ordinary income at vest. No strike price.

The single math you must do

For every offer, compute: 'If the company is worth $X in 4 years, this grant is worth $Y to the candidate at vest.' For early-stage offers, use the last round's preferred valuation as the floor and 3–5x as an optimistic ceiling. Show this math in the offer letter or a separate explainer. Candidates compare offers; you want them comparing yours accurately.

Defending the band

When a manager pushes for an off-band offer ('this candidate is special'), or when a candidate negotiates ('Google offered me $40k more'), or when the CFO asks why you're at the 70th percentile, the HRBP needs a defensible answer in 60 seconds.

The 60-second defense

1
Methodology
'We benchmark to the 65th percentile of the tech-startup competitive set using Radford Tech + Pave + Option Impact, refreshed annually.'
2
Mix philosophy
'Our mix is 70% base / 0% bonus / 30% equity at the senior level, calibrated for late-Series-B stage.'
3
Off-band approval
'Off-band offers require my approval and document the rationale (counter-offer, unique skill, market shift). We track off-band rate as a metric.'
4
Audit trail
'We re-band annually, run pay-equity analysis quarterly, and document every off-band exception. The band is the system, not the number.'

FAQ

Frequently asked questions

Do we have to disclose comp bands?

In several US states (CA, CO, NY, WA, IL effective 2025) yes, by law for job postings. The EU Pay Transparency Directive requires it across the EU by June 2026. Build for transparency now.

How often should we re-band?

Annually for the full band; quarterly mini-refreshes if the market is volatile (as in 2021–2022). Big-bang re-bands every few years are destructive — engineers hear about it through the grapevine and lose trust.

What about counter-offers from current employees?

Match only if you would have proactively given them the raise. 'Pay raises only when threatened' poisons your retention culture. Better to do proactive market adjustments.

Should our band be public internally?

Yes, with caveats. Buffer, Whereby, GitLab publish full bands. Most VC-backed companies publish bands to employees (not external) — this is the modern norm. Hiding bands is a 2010-era practice that doesn't survive Glassdoor + Levels.fyi.