Who runs this

What this site is for

Most self-tests on the web give you a number and stop there. A number on its own is close to meaningless: 240 milliseconds is either unremarkable or genuinely quick depending on whether you were on a mouse or a phone, and no page that hands you the bare figure tells you which. The reason this site exists is the second half of that sentence.

So every ranked test here ends the same way: a percentile, the device class it was computed within, and a visible statement of what the comparison population actually is. When that population is still a fitted reference curve rather than real people, the page says so in plain words rather than hiding it in a footnote. When it stops being a reference curve, the page says that too.

How it is paid for

Right now, it is not. There is no advertising on this site, no affiliate link, no sponsored test, and nothing to buy. It runs on a small edge deployment that costs very little to keep up.

If that changes it will change visibly. The commitment we are willing to put in writing is narrower than "no ads forever" and more useful: the results database is not for sale, and no third-party script will ever be loaded into a test page. A test that pauses to fetch an ad is a test that measures the ad network's latency along with your reflexes, which would quietly corrupt the one thing this site claims to do well.

What happens to your results

One row per finished test: which test, the score, the device class, and a random token your browser generated. No account, no address, no fingerprint. The token exists to count your first attempt and ignore the rest, because people who practise get faster and counting every retry would drag the whole distribution away from the truth.

Those rows are the population. Once a device class on a test reaches 5,000 real first attempts, its reference curve is dropped and the comparison becomes entirely measured. The full detail is on the methodology page, and the storage side is on the privacy page.

What we cannot tell you

These are screening tools running on consumer hardware, not clinical instruments. A hearing test through laptop speakers is bounded by the speakers. A colour test is bounded by your panel and the light in the room. A reaction time is bounded by your display's refresh rate and your browser's scheduling, which together can move a result by tens of milliseconds regardless of you.

Each page states its own limits where the result appears. No page here diagnoses anything, and a surprising result on any of them is a reason to see a professional rather than a reason to believe the number.

Corrections

If a figure on this site is wrong, we would rather hear it than not. That includes the reference curves: several are our own estimates, they are labelled as estimates, and a better published source would replace them.

Contact

support@benchself.com