Benchmarks are only useful if you can trust how they were made. This page is the complete method — including what this dataset can’t tell you.
The comparison dataset contains only data StrongApply may legitimately hold (Phase MI-1): rates and public signals the operator manually observed and recorded, data freelancers submitted themselves, and licensed/approved datasets. No marketplace scraping, no session reuse, no private endpoints. Every profile stores its source basis (manually verified / user submitted / approved dataset) and an observation note.
Coverage is small and non-random. It skews toward the niches the operator tracks (frontend, ecommerce, product design). Benchmarks describe this dataset, not the entire freelance market. Treat every output as directional.
Where shown, this is: publicly observed or self-reported earnings over a trailing period ÷ number of months. It is an estimate, not verified income and not SaaS-style MRR. It is always labeled as an estimate.
Medians and percentiles are used instead of means because freelance rates and earnings are highly skewed. Percentiles are computed over profiles with a non-missing value for that metric only.
Public benchmark outputs require at least 8 comparable profiles; internal benchmarks require at least 5. Below the threshold, the result is suppressed and shown as “not enough data” — never a percentile from a thin sample. Samples under 30 are additionally labeled small.
No opaque composite “best freelancer” score exists. Any comparative output states exactly which metric it orders by.
Each profile stores when it was observed. Benchmarks are computed live from the current dataset; there is no stale page cache in this release.
If a data point describes you and is wrong — or you want it removed — it is deleted on request, without requiring an account or sign-up. The dataset never contains personal contact details (no emails, no phone numbers): professional market signals only.