Indeed & Glassdoor vs BLS OEWS: Why Self-Reported Trade Pay Data Misleads

A $62,000 Posting That Didn't Match Reality

A shop owner in the Midwest posted a journeyman electrician opening last spring. Before setting the number, she did what most owners do on a Friday afternoon between service calls: she pulled up Glassdoor, typed in "electrician," and skimmed the range the site showed for her metro. It looked reasonable, so she used the midpoint as her offer.

Two weeks later, a candidate turned her down. He'd pulled his own number from a different site — a few thousand dollars higher — and assumed her offer was lowballing him. Neither number was obviously wrong. Neither was obviously right either. Both were estimates built from whatever a handful of self-selected users had typed into a form at some point in the past, for a job title that may or may not match what she was actually hiring for.

This is the quiet failure mode of crowd-sourced salary data in the trades: it feels authoritative because it's specific, instant, and free. But "specific" isn't the same as "verified." Before the next posting goes up, it's worth understanding exactly what Indeed and Glassdoor numbers are built from, what BLS OEWS data is built from, and why those two things are not interchangeable — especially for a signature that commits real payroll dollars.

How Indeed and Glassdoor Build Their Salary Estimates

Indeed and Glassdoor salary figures are self-reported and crowd-sourced. A number on either platform typically reflects whatever current or former employees, job seekers, or employers have voluntarily submitted for a given title — not a structured survey of actual payroll records. That means the underlying sample for any one job title in any one metro can be thin, skewed toward people motivated enough to submit a number (often because they feel under- or overpaid), and inconsistently matched to the actual duties behind the title. A "lead electrician" entry on one of these sites might include anything from a licensed journeyman running his own crew to an apprentice with a fancier business card.

None of this makes the sites useless — they're free, fast, and genuinely helpful for a gut-check or a conversation starter. But they are not drawn from employer payroll records, and they don't publish percentile breakdowns tied to a federal occupational classification. For a dollar figure that's about to become a line in a signed offer letter, that distinction matters.

How BLS OEWS Builds Its Wage Percentiles

The Bureau of Labor Statistics' Occupational Employment and Wage Statistics (OEWS) program works from the other direction entirely. It's built from actual employer payroll survey responses — roughly 1.1 million establishments sampled across six semiannual survey panels over three years, representing about 55% of national employment, according to the OEWS Technical Note for the May 2025 release. Employers report actual wage data for their workforce, sorted into standardized occupational codes, and BLS aggregates that into statistics for over 800 occupations across roughly 530 metropolitan and nonmetropolitan areas.

Instead of a single average, OEWS reports five wage percentiles per occupation — the 10th, 25th, 50th (median), 75th, and 90th — along with the mean. Annual wages are calculated as the hourly mean multiplied by 2,080 hours, the standard full-time work-year assumption. That percentile structure is the whole point: it tells you not just "what's typical" but where a specific candidate's experience, certifications, or responsibility level should land them inside a real distribution. If the mechanics of mean versus median affect how you read that distribution, it's worth walking through — see our breakdown on mean vs. median wage data in the trades.

As one BLS technical note puts it plainly:

"OEWS estimates are constructed from a survey of employers reporting actual wage and employment data for their establishments" — a structural difference from estimates built on voluntary, unverified submissions.

Indeed and Glassdoor Salary Estimates vs. BLS Data: What's Actually Different

Lay the two approaches side by side and four gaps show up consistently.

Sample composition. OEWS draws from employer payroll records tied to a standardized Standard Occupational Classification (SOC) code — the same code regardless of how a given shop titles the job on its org chart. Crowd-sourced sites draw from whoever submits a number under whatever job title they typed in, with no guarantee the title maps cleanly to the actual duties.

Transparency of methodology. BLS publishes its full survey methodology, response-rate data, and sample sizes in a public technical note for every release. Indeed and Glassdoor's self-reported model is publicly described at a high level, but the underlying sample size and composition for any single title-and-metro combination isn't something you can audit the way you can audit a federal statistical release.

Granularity you can defend. OEWS gives you five percentiles, not one number — a structure built specifically so you can say "this candidate's 12 years and master's license put them at the 75th percentile for this SOC code in this metro," and point to the public source behind that statement. A single crowd-sourced range doesn't carry that same defensibility if a candidate — or a pay-transparency auditor — asks where the number came from. For a primer on reading those percentile columns correctly, see how to read BLS OEWS data for the trades.

Update cadence and vintage. BLS OEWS is released on a known annual cadence with a stated reference period — the current release covers May 2025 data, published May 15, 2026. Every figure carries a date you can cite. Crowd-sourced estimates blend submissions across an unstated time window, so a number might reflect a mix of wages reported two years ago and two months ago, with no way to separate the two.

None of this means Indeed and Glassdoor are "wrong" in some absolute sense — self-reported data can be directionally useful, especially in markets or titles where OEWS data is suppressed or unavailable. It means the two sources are answering different questions: one asks "what did people who filled out a form say," the other asks "what did employers actually report paying."

Where Crowd-Sourced Numbers Still Help

It would overstate the case to say Indeed and Glassdoor have no place in a hiring process. They're useful for a fast directional check — "is my instinct in the right neighborhood" — and for watching how candidates themselves are framing their expectations, since job seekers often anchor on these same numbers. If a candidate quotes you a Glassdoor range as their justification for a counteroffer, understanding how that range was built lets you have a more productive conversation about it, rather than treating it as an immovable fact.

The practical move is to use crowd-sourced numbers as a sanity check and OEWS percentiles as the number you're willing to defend — in writing, to a candidate, or to anyone asking how you arrived at a pay band. If you want the free federal source alongside a plain explanation of how a wage-tool platform might save the hours that go into pulling it, our piece on free BLS data versus a paid wage tool walks through that tradeoff directly.

Turning a Percentile Into an Actual Offer

Knowing which source is more defensible is only half the job. The harder part is turning a percentile — say, the 50th percentile wage for electricians, $62,350/year nationally per the BLS Occupational Outlook Handbook's May 2024 data — into an actual min/midpoint/max band for a specific candidate in a specific metro. That requires picking an anchor percentile based on the role's seniority, then applying a spread buffer around it so the band has room for negotiation without drifting into territory you can't justify.

Worked simply: if you anchor a senior electrician role at the 75th percentile for your metro and apply a ±10% spread, your band runs from 90% of that percentile wage to 110% of it, with the percentile figure itself sitting at the midpoint. The arithmetic is simple; the part worth getting right is choosing a defensible, correctly dated, correctly sourced anchor in the first place — not splitting the difference between two unverifiable crowd-sourced ranges a candidate and an owner each found on different sites. For a deeper walk-through of anchor selection and spread logic, see our guide to wage percentiles explained for the trades, and for a quick way to sanity-check a specific offer against a percentile band without building a spreadsheet from scratch, the Trade Offer Check Calculator does the percentile-to-band math directly.

Whichever source you start from, the underlying standard is the same: a wage figure you can trace back to where it came from holds up better — to a candidate, to a bank, to your own second-guessing six months later — than one you can't.

If you'd rather have pieces like this land in your inbox as they publish, rather than tracking them down later, you can subscribe to our newsletter below.