How We Calculate Creator Salaries

Every figure comes from a job posting that actually states a salary, either the employer's structured pay field or a clearly-stated salary range in the description. Postings without pay are excluded from the statistics, never filled in as zero, and we never infer, estimate, or scrape a number that isn't there. If a posting doesn't disclose pay, it simply isn't in the sample.

A salary statistic is only ever computed within a single currency. We never pool different currencies into one average and never apply FX conversion (rates go stale and fabricate precision). That single rule produces three distinct surfaces:

• On a job: the pay is shown exactly as the employer posted it, in its native currency (rupees stay rupees, euros stay euros).
• Global view: averages across USD-disclosed postings only, labeled "USD, all locations". Honest because it is one currency by construction.
• Location view: you pick a country (default United States) and every figure is computed only from that country's postings, in that country's currency.

The board-wide pay filter and sort use USD only, kept deliberately simple; non-USD jobs still show their native pay everywhere they appear.

The benchmarks compare like with like: they use full-time, annual-basis postings. Contract, freelance, temporary, and internship roles, and anything paid per hour or per month, are left out of the averages (those jobs still appear on the board with their real pay). Monthly salaries are converted to a yearly figure; hourly pay is not extrapolated into a yearly number. Implausible parses (a maximum below the minimum, a range that is too wide, or a figure below a sane floor) are dropped.

Every figure carries its sample size (n) so you can judge it. We show a median only once a cohort has at least 5 disclosed postings, and the full minimum-to-maximum spread only at 8 or more, because extremes from a handful of postings are noise. Below those thresholds we say "limited data" rather than show a shaky number. Empty beats wrong.

Salary disclosure is most common in US and Canadian postings, so the global USD view skews toward those markets; the location views correct for that by staying within one country. Statistics are recomputed from the entire dataset on each ingestion run (roughly hourly) and stamped with an "as of" time, so they are current to the last run, not real-time. Where a page shows a live count next to a cached statistic, the count is read live.