Salary survey vs. salary websites: which should you trust?

The P50 team

Published October 2, 2026

Last updated October 5, 2026

Trust a salary survey when a real decision rides on the number, and use a salary website for a quick first look. The difference is where the data comes from. A survey collects pay from employers, usually straight from payroll records. Most salary websites collect it from workers who type in their own pay, or from the ranges listed in job ads, and nobody checks those numbers against a paycheck. Below we explain the two kinds of data, the problems with self-reported numbers, when each source is the right tool, and how to use both together.

The two kinds of pay data

Employer-reported data comes from the company that writes the checks. A salary survey asks HR or payroll to submit what each person in a matched job is paid, then pools that across many employers (see what is a salary survey). The biggest example is free: the Bureau of Labor Statistics' Occupational Employment and Wage Statistics program, or OEWS. BLS describes it as "a semi-annual mail survey of non-farm establishments," and the states that run it "make follow-up calls to request data from nonrespondents or to clarify data" (BLS OEWS FAQ). BLS calls it "a survey of business establishments (employers)" and samples roughly 186,000 to 189,000 of them in each of two panels a year (BLS OEWS survey methods). The full sample covers about 1.1 million establishments over three years, roughly 57 percent of U.S. employment (BLS OEWS FAQ). For the May 2024 estimates, 65.7 percent of sampled establishments responded (BLS OEWS technical notes).

Self-reported data comes from individuals or from job ads. There are two flavors:

Worker-submitted sites. A person fills out a form with their title, employer, location, and pay. One well-known site's methodology page explains that people fill out its pay questionnaire and get back reports comparing their pay to others (site methodology page). The person is the source.
Job-posting aggregators. The site reads the pay ranges printed in job ads and averages them. One job board's registered nurse page states its figure is built from salaries "taken from job postings on Indeed in the past 36 months" (Indeed registered nurse salaries). That is what employers advertised, not what anyone was hired at.

Both flavors can be useful. Neither is the same thing as a survey.

The known problems with self-reported data

Who reports is not a planned sample. A survey picks the employers it asks and chases the ones that do not answer. A website gets whoever shows up. People job hunting, feeling underpaid, or hoping for a raise are more likely to type in their pay than people who are content. Nobody designs that sample, so nobody knows which way it leans.

People pick their own titles. On a website, the person decides whether they are a "coordinator," a "specialist," or a "manager." Titles drift upward when nobody checks. In OEWS, by contrast, "all occupational classification is done by the expert OEWS staff instead of burdening employers with the task" (BLS OEWS FAQ). A survey matches jobs by what the person does, not by what they call themselves.

Base and total pay get mixed. One person enters base salary. The next enters base plus bonus plus overtime. A third enters what the offer letter said. The site shows one number. A survey defines the number first. OEWS wages, for example, are "straight-time, gross pay, exclusive of premium pay," which includes base pay and production bonuses but excludes overtime, shift differentials, and nonproduction bonuses (BLS OEWS technical notes). You know what you are reading.

Nothing is verified. A survey can tie each number to payroll records and ask follow-up questions. A website usually takes the number as typed. Honest mistakes, rounding up, and made-up entries all get the same weight.

Local samples are small. A national figure may rest on thousands of entries. The figure for your metro, your industry, and your job may rest on a dozen, and the page may not tell you. Job-ad numbers add a second problem: posted ranges are often wide, and the midpoint of a wide advertised range says little about what the hire actually earns.

When a salary website is still useful

None of this means you should never open one. A website is the right tool when:

You need a first look. You have a new role and no idea whether it pays $50,000 or $90,000. A website gets you to the right neighborhood in two minutes.
You want a sanity check. Your survey says $72,000 and the website says $70,000. Good. If the website says $110,000, look again (often a title mismatch).
No survey covers the job. Niche or brand-new roles may not have a survey match. See how to price a job without a salary survey for how to build a defensible estimate from what is available.
You want to know what candidates expect. Candidates read these sites. If the number they see is far above your range, you will hear about it in interviews. Knowing it ahead of time helps you explain your offer.

When you need a survey

Use employer-reported data whenever the number has to hold up to a second person:

Setting pay ranges. A range built on unverified entries is a range you cannot explain when a manager pushes back.
Budgeting. Finance is going to ask where the market rate came from. "A survey of 40 employers in our industry" is an answer. "A website" is not.
Defending a decision. If an employee, an auditor, or a board member asks why a job is paid what it is, you want a source with a method you can point to. Ours is on the how we calculate page.
Pay equity work. Comparing pay across people in similar jobs requires clean, defined numbers (base vs. total, hourly vs. annual). Self-reported data does not give you that.

Side by side

	Employer-reported survey	Worker-submitted website	Job-posting aggregator
Source of data	Employer HR or payroll records	Individuals typing in their own pay	Pay ranges printed in job ads
Verification	Defined pay fields and follow-up with the employer	Usually none beyond basic form checks	None; the ad is the data
Sample size shown	Companies and employees behind each number (BLS shows employment estimates)	Often a count of entries, rarely a count of employers	Count of postings, not hires
Cost	Free (BLS, P50) to thousands of dollars (one big survey lists modules at $2,600 to $21,300 as of October 2026, per its product page)	Free to view; paid tiers for employers	Free to view
Best use	Ranges, budgets, pay equity, anything you must defend	First look, sanity check, candidate expectations	Seeing what competitors advertise right now

How to use both together

The sources are not rivals. They answer different questions, so use them in order.

Start with a website to learn the lay of the land for a job you have never priced. Note the number and the title it used.
Price the job with a survey. Match on duties, not title, and read the P25, P50, and P75 with the counts beside them (how to read P25, P50, and P75).
Compare the two. A small gap is normal. A large gap usually means a title mismatch, a base-vs-total mixup, or a thin local sample on the website side.
Fall back to the website only for jobs the survey lacks, and write down that you did, so the next person knows that number is softer.
Check candidate expectations on the website before you post or make an offer, so you are ready for the conversation.

Keep a note for each job: which source, which date, which number.

Where P50 fits

P50 is employer-reported data, the same kind as the big surveys and BLS, built from pay that HR teams submit from their own records. We publish the number of companies and employees behind every figure, and the minimum-company rules that keep any one employer from being picked out, on the how we calculate page. Our sample report shows what that looks like for one job.

P50 is a free salary survey for employers. Registration for the 2027 survey is open now, and data collection runs March 1 to May 3, 2027. Register for free.