# Methodology

A speed test you cannot verify is a speed test you cannot trust. Below is exactly how PulseSpeed measures your connection — without marketing jargon.

## Principles

PulseSpeed runs on a globally distributed network and measures your connection against servers near you. We are transparent about how we measure so you can trust the results — below you can read exactly what happens for each metric and why we make the choices we make.

## Ping & Jitter

We send 12 lightweight measurement packets sequentially and measure round-trip time. For the average ping we drop the two highest outliers — this eliminates transient hiccups from your operating system or WiFi. For jitter we calculate the mean absolute difference between consecutive measurements before sorting, so we measure actual variation and not artifacts of a sort algorithm.

## Download

We open multiple parallel data streams to fully saturate your connection. The test runs long enough to bypass warmup effects (like TCP slow-start) and produces a fair average. We report mean throughput over the full measurement in megabits per second (1 Mbps = 1,000,000 bits/sec, not 1,048,576). The configuration is tuned for both older and modern web protocols (HTTP/1.1, HTTP/2 and HTTP/3) so the measurement is representative across all browsers.

## Upload

For the upload measurement we similarly use multiple parallel streams. This is the only way to realistically measure how fast you can actually send large files. Server-side, incoming data is streamed without buffering, so the measurement exactly reflects what the network can handle and is not skewed by a fast queue.

## Bufferbloat

After upload we measure ping while multiple downloads run in the background. The difference between your idle ping and loaded ping is bufferbloat: the extra latency caused by network equipment (often your own router) buffering packets instead of processing them immediately. Grade A = less than 5ms increase, F = more than 200ms increase. This is the single most defining factor for whether your connection "feels fast" during video calls or gaming.

## Stability Test

Our 5-minute stability test measures every second — 300 samples total. This reveals patterns a short test misses: dropouts, periodic slowdowns, throttling by your provider. We use a custom aggregate score based on range (max−min ping), jitter trend and packet loss. A connection with great peak speed but high variation still gets a lower stability grade because that feels worse in practice.

## Privacy

PulseSpeed does not store test results on our servers. Your history stays exclusively in your browser. We send no third-party analytics, use no tracking cookies and do no browser fingerprinting. We also do not log IP addresses or measurement results on the server. A rate limiter temporarily counts requests per minute, but these counts automatically disappear after a few minutes.

## Endpoint

All tests run against a globally distributed network so you always test against a nearby server. After each test run you see which node (e.g. AMS, FRA, LHR) you hit. This means the measurement reflects your actual performance on the modern web rather than the distance to a single central server.