Salary percentiles
What Salary percentiles means
A percentile is the value below which a given share of observations fall. For salaries, the 10th percentile is the pay that only one in ten reports is below, the 50th is the median, and the 90th is what only one in ten exceed. Compensation teams set bands from them, usually around the 25th to 75th.
Percentiles are more useful than an average for pay, because a few very high salaries pull the average up while the median stays put. They are only as good as the reports behind them: a title with a handful of reports has percentiles that move with every new one.
Clair's Glassdoor salaries endpoint returns base_pay and total_pay at the 10th, 25th, 50th, 75th, and 90th percentiles for each job title, with salaries_count, the number of reports behind them.
Reading a pay range: an example
Glassdoor's figures for Software Engineer III at Google in the United States, from 950 salary reports, show how to read percentiles. The middle half of reported base pay runs from $176,568 to $221,906: a band about $45,000 wide, or 23% of the median. Someone paid $230,000 base is above the 75th percentile for the title, but under the 90th.
| Percentile | Base pay | Total pay | Read it as |
|---|---|---|---|
| P10 | $159,325 | $271,280 | One in ten reports is lower |
| P25 | $176,568 | $321,338 | Bottom of the middle half |
| P50 | $197,943 | $390,969 | The median: half earn less |
| P75 | $221,906 | $492,142 | Top of the middle half |
| P90 | $245,922 | $599,160 | One in ten reports is higher |
Base pay and total pay
Base pay is salary alone. Total pay adds what else the reports include: bonuses, stock, commission, and profit sharing. The gap depends on the employer and the level. In the example the median total pay is about twice the median base, which is typical of large technology companies that pay much of their compensation in stock, and rare elsewhere.
The two spread differently. Base pay is held in bands, so its percentiles stay close together; total pay follows bonuses and stock prices, so its percentiles spread out more and move year to year. Compare base with base and total with total, and say which one a figure is.
When percentiles are reliable
- Count: percentiles from a handful of reports shift with every new one. Hundreds, as in the example, give stable figures; under about 20, treat the range as rough.
- Freshness: pay moves with the market. Check when the latest report came in; Clair returns it as last_reported_at.
- Location: the same title pays differently by city and country. Glassdoor's figures are for the employer's home country unless you ask for a location, and Clair's endpoint lists the states and cities Glassdoor has figures for.
- Self-reporting: the reports come from people who chose to share their pay, so they are an estimate of the market, not a payroll record.
Common questions
What does P50 mean in salary?
The 50th percentile, or median: half of the reported salaries are below it and half above. It is the usual single figure for what a job pays.
Is the 75th percentile good pay?
It means pay higher than three in four reports for the same title and place. Employers that aim to pay above the market often target it.
Why use percentiles instead of the average salary?
A few very high salaries pull the average up, while the median stays put. Percentiles also show the spread, which an average hides.
What is a compa-ratio?
Salary divided by the midpoint of its pay band, often the market median. A compa-ratio of 1.0 is paid at the midpoint; 0.9 is 10% below it.

Glassdoor API
Crunchbase API
Instagram API
Bloomberg API
Associated Press API