Where the salary numbers come from

Most salary sites give you one number per title and hope you never ask where it came from. The number is usually an average of self-reports, scraped ads, and modeled guesses, all melted together. Our salaries page does the opposite. It shows three kinds of evidence on one scale and never averages across them.

How to read a figure

Each row on the page is one role at one level, in one country and one currency. Three marks can appear on it.

The yellow band is what members report anonymously: the middle half of reports, with a line at the median. A band only appears once a cut has five reports, so a lone salary can never be reverse-engineered.

The diamonds are exact salaries employers filed with the US government. When a company sponsors a work visa, it must file the base salary on a public Labor Condition Application. One diamond is one person's actual pay, on the record.

The dashed outline is the range printed on a job ad. These are compliance bands, drawn wide on purpose, so the page treats them as context. When a range runs off the scale, it clamps at the edge with an arrow rather than flattening every other mark on the row.

Every mark carries its sample size, source, and date. If two marks disagree, you see the disagreement. That is the point.

Why the sources stay separate

Each source is wrong in its own way. Self-reports skew toward people motivated to share. Government filings skew toward visa-sponsored roles and prevailing-wage compliance. Advertised ranges are legal cover, not market midpoints; one verified IQVIA posting for a principal biostatistician ran from $96,400 to $294,800, a three-fold spread that tells you almost nothing about what the job pays. Average these three and you manufacture precision that none of them has. Keep them apart and each one checks the others.

What the public record already shows

The filings hide a trap we had to learn the hard way: entity names. IQVIA's clinical research roles file under "IQVIA RDS Inc", 567 records with a median base of $125,000 across 2019 to 2025. Search "IQVIA Inc" instead and you get the consulting and tech entity, 160 records, not one clinical title among them. Anyone benchmarking against H-1B data by brand name is reading the wrong company.

Query the right entity and a ladder appears. Senior statistical programmers at IQVIA filed at a median of $122,400 across 89 records. Principal level ran $150,565. CRA 2 filings from 2023 and 2024 cluster between $95,790 and $105,000 depending on location, and a 2025 senior CRA filing came in at $120,660. These are not estimates. Each row was re-checked against the government dataset one filing at a time, because our first research pass caught automated extraction inventing plausible table rows. Three claims died in verification; what survived is on the page.

Parexel, for contrast, posts no pay ranges at all, even on US-remote roles open to states with transparency laws. We checked in July 2026. The silence is its own datapoint.

The part only you can add

Government filings cover visa-sponsored roles. Job ads cover whatever employers must disclose. The middle of the market, the ordinary raise, the sponsor-side counteroffer: that only shows up when people report it. The survey takes about a minute, asks for one number and a few taps, and shows you where you sit in your cut the moment you submit. The figures on the salaries page sharpen one anonymous report at a time.