Salary Research: How to Know What You're Worth in 2026
Most candidates undershoot. Not because the market is brutal, though it is. Because they walk in without doing the math. The number you say first is the ceiling unless you negotiate hard. Research is the leverage that moves it.
Four sources, a three-number framework, and the script for the moment a recruiter asks for your range.
Key Takeaways
- The first number you say is the ceiling unless you negotiate hard. Research before any conversation about pay.
- Use four sources, not one. Each is good at something specific and gets something else wrong.
- Build three numbers: a floor (the bottom you'd take), a target (the realistic ask), a stretch (the number you say first).
- Adjust for geography and normalize equity. A $200k offer in San Francisco isn't a $200k offer in Atlanta, and a $200k base with no equity isn't a $200k base with $400k of RSUs over four years.
- When asked "what's your salary range?" pivot back to research. Your number comes after the role and total comp are clear.
The reality: your first number is the ceiling
Salary expectation questions are filters. Recruiters ask early because the answer disqualifies candidates before anyone wastes a calendar block. A low answer caps the offer. A high answer with no research behind it gets you screened out. A researched answer holds.
Most articles tell you to "deflect." That's half right. You deflect long enough to give a researched range, not forever. The leverage is in the research, not the dodge.
If you're doing this mid-search after a cut, the rest of the playbook is in Tuesday's full layoff checklist. Comp is one of the bigger items on it.
The four sources to use (and what each gets wrong)
No single source is right. Use the strengths, discount the weaknesses.
levels.fyi
Good for: tech and engineering. Real numbers from real offers, geo-adjusted. Strong on base, equity grants, signing bonus, and total comp at FAANG-tier and Series-B-or-later companies. If you're an engineer, PM, designer, or data scientist, levels.fyi is the most accurate public source you have.
Gets wrong: outside tech, the data is sparse. Don't try to research a Senior Marketing Manager at a manufacturing company on levels.fyi. The sample size lies.
Glassdoor
Good for: breadth. Almost every role at almost every company has a few entries. Need to ballpark a Senior HR Business Partner at a regional health system? Glassdoor has data where most aggregators don't.
Gets wrong: self-reported and noisy. The people who post are either underpaid and angry or overpaid and bragging. Use it for direction, not precision. Median is roughly trustworthy. Individual data points aren't.
Bureau of Labor Statistics — Occupational Employment data
Good for: official, free, every occupation, every metro. The BLS Occupational Employment and Wage Statistics program (bls.gov/oes) publishes wage data for hundreds of occupations across hundreds of metros. Methodologically transparent.
Gets wrong: lags 12 to 18 months. No equity, no signing bonus, no most variable comp. Categories are broad ("Marketing Managers" covers very different jobs at very different price points). Use it as a floor, not a target.
Personal network
Good for: the most accurate signal you'll get. Current, role-specific, company-specific. A peer at a competitor who says "they offered me $X for the same role last quarter" beats every aggregator combined.
Gets wrong: socially expensive. Asking what someone makes is a real ask, even from a friend. Don't blast a Google Form to twenty contacts. Pick two or three people who know your work and the target market, ask once, in private, with context for why.
The three-number framework: floor, target, stretch
You don't walk into a salary conversation with a single number. You walk in with three.
Floor: the bottom you'd take
Derive from BLS occupational median for your role, adjusted up for cost of living. Not the number you say. The number you walk away from.
Math: BLS reports a national median for your occupation. Pull the metro 75th percentile (BLS publishes the percentile breakdowns) — that's roughly the bottom of where the market actually lives in 2026. Below it, you're losing money to take the role. Round to a clean number. Write it down.
Target: the realistic ask
Derive from Glassdoor median for your role at the target company tier and level, calibrated against the levels.fyi 75th percentile in your geo if you're in tech.
Math: Glassdoor median for Senior Product Manager at a company your target's size is $Y. Levels.fyi 75th percentile for the same role at the same tier is $Z. Target is the average, biased toward levels.fyi if you're in tech, biased toward Glassdoor if you're not.
This is the number the offer should land at if the conversation goes well. Not the first number out of your mouth — the number you're aiming the negotiation toward.
Stretch: the number you say first
10 to 15 percent above target. Target is $200,000, stretch is $220,000 to $230,000.
The reason is anchoring. The first number on the table sets the ceiling. Recruiters move down from your number more easily than they move up from theirs. You can come down. You can't go back up. Asking for the stretch and landing at target is the standard outcome of a researched negotiation.
The stretch isn't a fantasy — it has to be defensible within the realistic top of the market for your role, level, and geo. If the recruiter pushes back with "that's above our band," you haven't blown it. You've learned the band.
Geographic adjustment: remote, in-office, hybrid
Most companies pay differently across geos. The framework has to adjust.
Remote roles increasingly use geo-tiered bands. Tier 1 (SF, NYC, Seattle) gets the top. Tier 2 (Austin, Boston, Chicago, LA, DC) gets the middle. Tier 3 (most other US metros) gets the bottom. Tiers vary by company; some publish them, most don't.
In-office roles in expensive metros usually pay more in absolute dollars, but cost of living eats most of the spread. A $180k offer in San Francisco and a $130k offer in Atlanta produce roughly the same take-home after rent and taxes.
Hybrid roles usually use the geo of the office, not your home address. If the office is in SF and you live in Sacramento, your band is the SF band.
Rough COL multipliers: SF/NYC roughly 1.4x national median, Tier 2 metros roughly 1.15x, Tier 3 roughly 1.0x. Adjust floor and target by the multiplier for the geo you'd actually be earning in.
Equity normalization: comparing offers with different stock structures
Comparing a $180k base plus $400k of RSUs against a $210k base with no equity isn't a math problem most people are taught. Walk through it the same way every time.
RSUs (public companies): vest typically over 4 years, 1-year cliff, quarterly after. Annualize: total grant ÷ vest years. A $400k grant ÷ 4 = $100k of equity comp per year. Add to base.
ISOs / options (private companies): harder. Strike price matters and the equity is theoretical until a liquidity event. Discount aggressively. A rough rule in 2026: value private-company option grants at 25 to 50 percent of paper value against public-company RSUs. Below Series B, discount harder.
No equity (most non-tech): equity is zero. Comparison is base plus bonus, period.
Build a "total annual comp" line for each offer: base + target bonus + (equity ÷ vest years, discounted). Compare the lines, not the headlines. The bigger headline isn't always the bigger offer.
The script: "what's your salary range?"
Three sentences. No apology, no number first.
"Based on my research for similar roles in this geo, the market is roughly $X to $Y. Before I lock a number, I'd want to understand the role, scope, and total comp structure. What range do you have budgeted for this position?"
That's the pivot.
If they push for a number first, give the stretch. Frame it as the top of your researched range: "Based on what I'm seeing for senior roles in this metro, I'd be targeting around $Z, depending on equity and total comp." You said a number. You attributed it to research. You opened the rest of the conversation.
Never give a single number with no range and no caveat. That's the move that caps the offer in one sentence.
The bottom line
Research before any number leaves your mouth. The four sources aren't equivalent — use each for what it's good at. Three numbers, not one: a floor you walk away from, a target you aim toward, a stretch you say first. Adjust for geo. Normalize equity. Pivot the question back to research, every time.
The interview itself is its own job. The first phone screen will ask why you're leaving before it asks for a number — the framework that lands is the next read. If you want the resume rewrite and a 12-question interview script tied to the role you're actually interviewing for, three resumes plus interview prep for $4.99 handles it in one session. No subscription. The files are yours.
Do the math. Then say the number.