How to read the wage percentiles on this site
Why the median is the headline number, what the 25th and 75th percentiles tell you that an average cannot, and how to spot a figure built on too few filings to trust.
Last updated
Every entity page here leads with five numbers: a median, three other percentiles, and a filing count. That last one decides whether the other four are worth anything.
Why the median and not the average
Wage data has a long right tail. A single physician or executive filing at $900,000 sitting among engineers at $150,000 will drag an average upward by thousands and describe nobody in the group.
The median — the 50th percentile — is the middle filing: half are above it, half below. One extreme value moves it barely at all. That robustness is the entire reason it is the headline number on this site.
What the other percentiles add
- 25th percentile. A quarter of filings are at or below this. Read it as the lower end of normal, not the floor.
- 75th percentile. A quarter are at or above. Read it as a strong offer for the group.
- 90th percentile. The top tenth begins here. On a large employer this is roughly where the senior band starts.
The distance between the 25th and 75th — the middle half — is the useful shape. Narrow means a consistent band. Wide means the group mixes very different jobs, usually different wage levels.
How many filings is enough
This is the question most worth asking before trusting a number.
| Filings | How to read it |
|---|---|
| 1–4 | Anecdote. The median is just one of a handful of salaries. Percentiles are close to meaningless. |
| 5–29 | Indicative. The median is worth something; the 90th percentile is still one or two people. |
| 30–199 | Solid. The whole distribution is meaningful. |
| 200+ | Strong, but now check what is being blended — many roles and levels under one label. |
The filing count sits beside the wage figures on every page precisely so this check is possible. Where a company-and-title combination had too few filings to say anything, this site does not publish a page for it at all.
Comparing fairly
Three things distort a comparison more than anything else:
- Occupation mix. A company median blends every role it sponsors. Compare within one occupation instead.
- Location. The same job pays differently in different markets. The state and city pages exist for this.
- Seniority mix. An employer hiring mostly juniors shows a lower median without paying worse for equivalent work.
The bar chart
The band under the headline numbers shows the middle half of filings with the median marked. Its axis is deliberately scaled to that band rather than to the true minimum and maximum — with one outlier at ten times the median, a full-range axis would compress the interesting part into a sliver. The endpoints shown are real filed wages, just not always the most extreme ones.
For the limits of the underlying data, see what this data does not say.