Measurement Methodology
How we measure your connection. Open, documented, and reproducible.
Download speed
We open multiple parallel HTTP/HTTPS streams to a test server and transfer a large payload. The number of streams scales with your measured bandwidth so that we genuinely saturate your connection rather than under-utilise it. We sample throughput throughout the transfer window and report the median value — not the peak — making the result resistant to single-second spikes that would otherwise inflate it. See accuracy tips for how to get the best result.
Upload speed
Upload is measured symmetrically: parallel streams send data to the test server. We use the same multi-threaded saturation approach and report the median throughput over the transfer window. Upload testing follows the download phase so it does not interfere with latency measurements.
Ping / latency
Latency is measured by sending a series of lightweight HTTP requests to the test server and recording each full round-trip time. We take the median of these samples. Median latency is more representative than minimum latency for real-world usage, because it reflects the connection state under normal conditions rather than the single best case.
Jitter
Jitter is the standard deviation of the individual ping samples collected during the latency phase. A connection with consistent 20 ms latency has near-zero jitter; a connection that swings between 10 ms and 80 ms has high jitter. Jitter above 30 ms is generally perceptible in voice and video calls. Learn more about jitter →
Packet loss
We send a fixed number of packets to the test server and count how many acknowledgements we receive. The difference, expressed as a percentage, is your packet loss rate. We filter transient single-packet losses that occur at test startup, which can skew the result during connection establishment.
ISP and country detection
We identify your ISP using a combination of reverse-DNS lookup and ASN (Autonomous System Number) database lookups. The ASN is the most reliable signal for ISP identity; reverse-DNS is used as a fallback and for enriching the display name. Your full IP address is never stored — we retain only a masked version (first two octets) for geographic ranking purposes.
Country detection uses the same ASN database combined with GeoIP data. In the small number of cases where these disagree, ASN data takes precedence.
Rankings methodology
Country and ISP rankings are computed from a rolling 7-day window of verified test results. Each day, we recalculate the median download speed, upload speed, and ping for every country and ISP that has accumulated at least 5 samples in the window.
We use medians rather than averages throughout because medians are robust to outlier results — a single user on a 10 Gbps research network should not distort a country's ranking. The 5-sample minimum prevents rankings being dominated by a single user's results on low-volume days.
Fraud detection and result quality
All results pass through an automated quality pipeline before entering the ranking dataset. The pipeline checks for:
- Implausibly high speeds inconsistent with known network infrastructure
- Results that are identical to past results from the same origin (scripted or replayed tests)
- Tests that complete abnormally fast, suggesting the payload was not fully transferred
- Repeated testing from the same source within a short window
Results that fail these checks are excluded from rankings and the weekly competition. Accounts with repeated violations are flagged for manual review.
Test servers
We operate test servers in multiple regions. When you start a test, we select the server with the lowest measured latency to your location. All servers run identical software to ensure consistent measurement conditions. View all active servers →
Data retention
Individual test results are retained for 12 months. Aggregated ranking data (medians per country/ISP per day) is retained indefinitely. You can request deletion of your personal test history at any time via the contact page.