Am I underpaid in 2026? A step-by-step check that beats Glassdoor

Here's the awkward truth about salary benchmarking in 2026: every company you'll interview with has a private, current, role-specific compensation database — Radford, Willis Towers Watson, Mercer — that costs $50K-250K/year per subscription. You have Glassdoor, which is an unverified self-reported database with a 3-year-old median.

This guide shows how to close that information gap without paying for a paid database, using a step-by-step method that works for any role in any city.

Why Glassdoor is systematically low (or high)

Glassdoor's salary data has three structural biases:

1. Self-report skew. Self-reported numbers over-index on base salary and under-index on equity, bonus, and location premiums. For a role where 40% of comp is equity (tech startups, financial services), Glassdoor will show comp 30-50% below reality.
2. Stale data. A "$95,000" median for a product manager in Seattle reflects submissions averaged over 24+ months. Wages moved 11-18% in some roles between 2022 and 2024. Glassdoor's median is the 2022-2023 number.
3. Survivorship bias. People whose salary matches their expectations submit less often than people who are surprised (high or low). The sample is non-random.

Levels.fyi is better for tech roles at FAANG-sized companies (verified offers, tighter cohorts) but essentially useless for roles outside that ecosystem — regional marketing managers, Midwest physical therapists, any non-tech role.

Better free sources in 2026:

• BLS Occupational Employment and Wage Statistics — bls.gov/oes. Government data, refreshed annually, covers 800+ occupations with geographic detail. Trailing by ~18 months but rigorous.
• Pay transparency laws. As of 2025, 16 US states + NYC require salary ranges in job postings. You can collect real ranges by searching for your exact role on LinkedIn/Indeed in a transparency state.
• ERI Salary Expert — free tier shows median + range for role+city combos; premium is expensive but the free tier is enough for most.
• Blind — anonymous forums by company. Data is noisy but current.

The 5-variable benchmark method

Your actual market value depends on 5 variables, in decreasing importance order:

Variable 1: Role + Level (30-50% of variance)

Level mapping is where most self-benchmarking goes wrong. A "Senior Engineer" at a 50-person startup is typically a Staff or Lead Engineer at a 5,000-person company. Same person, different title. The comp gap between those two titles is $40-80K.

Use the Radford global leveling guide (free summary on radford.aon.com) to map your responsibility level (P2/P3/P4/P5/P6) to titles at different company sizes.

Example:

• "Senior PM" at a 150-person Series B = P4 level
• "Senior PM" at Google = P5 level
• $40K comp delta for the same title.

Variable 2: Geography + Cost of labor (20-30% of variance)

Not cost of living — cost of labor. These are different.

Cost of living in San Francisco is ~2.3x Des Moines. But cost of labor for a senior software engineer is only ~1.6x. Companies pay a labor premium but not dollar-for-dollar with cost of living, because remote work has compressed the differential.

Use BLS MSA (Metropolitan Statistical Area) wage data, not state averages. Wage variance within a state can be 40%+ between major metros and rural areas.

Variable 3: Company stage (15-25% of variance)

Same role, different stage:

• Pre-seed to Seed: cash comp 60-75% of market, heavy equity
• Series A-B: cash comp 80-95% of market, moderate equity
• Series C+: cash comp 95-105% of market, lower equity %
• Late-stage / pre-IPO: cash comp 100-115% of market, RSUs
• Public: cash comp 100-125% for top-25% companies, RSUs + refreshers

If you're at a Series A and your cash comp is at 110% of market, you're either at a top-25% percentile company OR overpaid relative to your stage norms.

Variable 4: Skill stack premium (10-20% of variance)

Specific skills compound:

2026 skill premiums (observed from job posting text analysis, sources: LinkedIn Salary Insights, Hired.com 2024 state of tech salaries):

• AI/ML engineering: +15-25%
• Kubernetes / cloud infrastructure: +10-18%
• Mobile native (iOS/Android): +8-15%
• Security engineering: +12-20%
• Data engineering: +10-18%
• Typescript/React (baseline): 0% (saturated)
• Python: +0-5% (role-dependent)

These stack. An AI/ML engineer with Kubernetes + security clearance commands 40-60% premium over a generalist.

Variable 5: Years of experience, bucketed (10-15% of variance)

Experience is less important than the above 4, but still matters:

| YoE | Typical Level | Compensation Band | |---|---|---| | 0-2 | Junior / Associate | 70% of market | | 2-5 | Mid / IC2 | 85-95% of market | | 5-8 | Senior / IC3 | 100-110% of market | | 8-12 | Staff / IC4 | 115-135% of market | | 12+ | Principal / IC5 | 135-180% of market |

The senior-to-staff jump (years 5-8) is the single largest comp discontinuity in most career paths. The 20-30% jump happens all at once, typically at a role change rather than an internal promotion.

The 15-minute self-benchmark protocol

1. Open BLS OES for your occupation + MSA. (bls.gov/oes, 3 minutes)
• Get the 50th percentile, 75th percentile, and 90th percentile raw wage.
• Adjust for company stage. (2 minutes)
• Multiply BLS median by your company-stage factor (above).
• Adjust for skill stack. (3 minutes)
• Add skill premiums for skills you actually use on the job (not ones you have experience with but don't use currently).
• Collect 3-5 real salary-range data points. (5 minutes)
• Search LinkedIn/Indeed/Built In for your exact role in a pay-transparency state (NY, CA, WA, CO).
• Note the min and max of the posted range.
• Compute your expected range. (2 minutes)
• Low: BLS 50th × company factor × 1.0
• Mid: BLS 75th × company factor × (1 + skill premium)
• High: top of LinkedIn real postings range

If your current salary is below the Low, you're underpaid. If between Low and Mid, you're in the normal range but could push. If above Mid, you're well-paid for your role/level.

When you're underpaid — the negotiation script

The script that works in 2026 has a specific structure:

1. Open with a concrete number, not a request. Not "I'd like a raise," but "Based on market data, I believe my compensation should be in the $X–Y range."
2. Justify with sources. Cite BLS, LinkedIn postings (with screenshots), and your specific skill premiums. This forces the counter to engage with data, not feelings.
3. Attach to value delivered. List 3-5 specific wins in the past 6-12 months with dollar or percentage outcomes.
4. Name the alternative. Not a threat, a reality. "I want to stay — compensation is the only variable that's out of alignment."
5. Ask for a decision date. "Can we come back to this by [specific date, 2-3 weeks out]?" prevents indefinite deferral.

What NOT to do:

• Don't mention competitor offers you don't have.
• Don't threaten to leave unless you're prepared to follow through in 30 days.
• Don't negotiate over email if you can negotiate in person or on video.
• Don't negotiate against yourself. When they say "let me check," don't lower your ask.

The raise math: why you should ask every 18 months

A 4% annual raise compounds to 48% over 10 years. A 10% raise every 18 months compounds to 122% over 10 years.

The difference between "market raise" and "negotiated raise" compounds brutally. A $100K salary with 4% annual becomes $148K after 10 years. The same salary with a 10%-every-18-months pattern becomes $222K. Over a 30-year career, that's $1.2M–$2.5M in compounded differential.

The expected value of a 30-minute raise conversation is enormous. Even if the answer is no most of the time, the upside dominates.

Offer negotiation: the 3 numbers that matter most

When you have an offer, don't optimize base salary. Optimize total comp at year 2, which depends on 3 numbers:

1. Base salary. Largest weight.
2. Equity grant vesting schedule. 4-year standard, typically 25% after year 1 cliff, then monthly. Startup equity is often worth nothing; late-stage equity is often worth 30-60% of base. Ask for the current 409A valuation or last round price.
3. Signing bonus. Often negotiable within 20% without anyone checking. $10K ask, $5K counter is typical.

Target: 10-25% above the initial offer, not 5%. The company's opening offer is usually ~10% below their max. Ask above that, get countered back toward their max.

When the data says you're paid fairly

Sometimes the math says you're in the right range. That's a good outcome — you're not leaving money on the table, and you can focus on things that actually compound (skills, networks, equity). Paid-fairly is underrated relative to paid-maximally; the negotiation costs of constantly re-opening comp are real.

But "paid fairly for my level" is different from "paid fairly for a higher level." Sometimes the right move isn't a raise — it's a promotion to the next level, which unlocks the +20% band. That's a different negotiation (scope, ownership, visibility) with a longer timeline.

Sources and tools

• BLS OES: bls.gov/oes (free, most rigorous)
• Levels.fyi: levels.fyi (tech-heavy, verified offers)
• Built In salary: builtin.com/salaries (startup ecosystem)
• ERI Salary Expert free tier: erieri.com
• Radford leveling guide: radford.aon.com (free summaries)
• LinkedIn Salary Insights: linkedin.com/salary (medians, aggregated)

The one-hour investment of doing this math on yourself, once every 12-18 months, is the highest-ROI time an employed person can spend. Most underpayment is invisible because people don't do the math. The math is doable in an hour.