Skip to content
speedtest.doctor
E-E-A-T

Our Measurement Method

A test is only as trustworthy as its method. Here is exactly what we measure on api.speedtest.doctor, how we keep the numbers honest, and where the limits are.

Every figure on speedtest.doctor comes from measurements against our self-hosted endpoints at api.speedtest.doctor — download, upload and latency probes on infrastructure we control. We report stability rather than a lucky peak, and we're upfront about what a browser test can and can't tell you.

What each metric means

Download / Upload

Parallel streams to multiple high-bandwidth endpoints; we report a stable aggregate, not a single lucky burst. Throughput ramps up before we start counting, so slow-start doesn't drag the number down.

Ping (latency)

Round-trip time sampled repeatedly against several servers. We surface the median and the spread, not just the best result — a single low ping can hide a shaky line.

Jitter

Variation between consecutive latency samples. Low jitter means a steady line; high jitter is what makes calls choppy even when the average looks fine.

Packet loss

The share of probes that never come back, measured over a sustained window rather than a quick burst, so transient blips don't masquerade as a healthy line.

Bufferbloat

Added latency while the link is saturated. We compare idle ping to ping-under-load — the gap is the metric that ruins gaming and video calls.

DNS

Resolution time across common public resolvers, so you can see whether name lookups (not bandwidth) are what makes pages feel slow to start.

Principles

Honest about the limits

No browser-based test is perfect. We'd rather tell you why than pretend otherwise.

Your device matters

An old phone, a busy CPU or a weak Wi-Fi radio can cap results below what your line can do. Test wired when you want the line's true ceiling.

Wi-Fi is a variable

Distance, walls and interference change results room to room. We measure your connection as it reaches this device — not the theoretical plan speed.

Time of day counts

Evening congestion is real. One test is a snapshot; re-running at peak hours reveals patterns a single run can miss.

Detailed, versioned methodology is published alongside the live tester. This page documents the approach today and will be expanded with exact server locations and sampling windows as the measurement engine ships.