How Pulse measures your speed

Pulse on UKSpeedTest.co.uk is a browser-based UK speed test. It measures download speed, latency, and jitter. It does not currently measure upload speed. Results are shown in your browser; this tool does not store your speed results on our servers. One run is a snapshot - useful when repeated fairly, not a single verdict on your line or package.

What we measure

Download speed (Mbps)

What it is: How fast data is delivered to your device during the test, in megabits per second.

How Pulse measures it: The test downloads a fixed payload from a content delivery network (CDN) and times how long it takes using high-resolution timers. The headline figure uses multiple samples and a median (see below) - not a single lucky spike.

Why it matters: Higher download capacity often means more headroom for streams, updates, and several users - but it is only one part of how “fast” the internet feels.

How to interpret it: Compare similar setups (same device, room, time of day). If results swing wildly, look at Wi-Fi, background apps, and repeat testing before drawing conclusions.

Latency (ms)

What it is: Round-trip time for small requests, in milliseconds. Lower usually feels more responsive for games and real-time calls.

How Pulse measures it: Several small GET requests to the same CDN host and path used for the download test, so the measurement reflects an internet path - not a lab-only check inside your home.

Why it matters: Interactive applications care about delay and consistency, not only Mbps.

How to interpret it: A single latency figure can hide variation; read it alongside jitter and repeat tests.

Jitter (ms)

What it is: Variation between consecutive round-trip samples - how “steady” the delay is.

How Pulse measures it: Derived from the same small-request sequence used for latency on that CDN path.

Why it matters: High jitter often shows up as uneven audio/video or inconsistent feel in real-time use, even when average Mbps looks fine.

How to interpret it: Lower and steadier is generally easier on calls and streaming; combine with latency and download for a full picture.

Upload: Not measured in the current version. Anything that depends heavily on upstream (some backups, large video uploads) needs a different check.

Test flow (start to finish)

Load the page: Pulse may measure latency and jitter with a short burst of tiny requests to the CDN so the analytics panel can show current round-trip behaviour.
Start the test: When you press Start speed test, Pulse runs multiple download passes. Each pass downloads a fresh payload (see Cache-busting).
Multiple samples: The app runs 3 runs × 3 samples per run (9 download measurements in total by default), updating the on-screen speed as it goes.
Headline speed: When sampling completes, Pulse computes statistics (including highest, lowest, mean, mode) but uses the median download speed as the main headline.
Confidence label: Pulse adds High, Medium, or Low confidence based on your environment (for example if the tab was hidden) - a nudge to interpret fairly, not a separate “quality score.”
Results in the browser: Final numbers and interpretation stay on the page; nothing is sent to our servers as a stored speed result for this tool.

Sample logic: why the fastest spike is not the headline

Any single download sample can catch a brief moment of ideal conditions - or noise. Pulse therefore records several samples and shows the median as the headline. The extended results still show your highest sample for transparency, but we do not promote that peak as the main number, because it is the least representative for “typical” performance during the test window.

Why we use the median

The median (middle value when samples are sorted) is less swayed by one extreme spike or dip than a single run or a simple average in many real-world cases. It is a practical compromise for a short browser test - not a statistical guarantee for every scenario. If you need provider-grade evidence, combine repeated tests, fair setup (often Ethernet), and notes of date and time.

Confidence (High, Medium, Low)

Confidence is not a measurement of your line quality. It reflects how trustworthy this browser session is likely to be for the headline number:

High: Typical conditions - tab visible, no strong signal that the browser is constraining the test.
Medium: For example, reduced-motion preference is enabled, or the Network Information API suggests a very constrained connection type - worth another run under normal use.
Low: For example, the tab was in the background during the run, which can throttle timers and networking - rerun with the tab in the foreground.

Use confidence as a prompt to repeat the test fairly, not as a label on your ISP.

Cache-busting

Pulse adds unique query parameters to each request (including a timestamp) so browsers and intermediaries are less likely to reuse a cached payload and inflate the measured speed. Fetches also use no-store cache behaviour where supported. Together, this reduces “phantom Mbps” from stale content.

Endpoint and CDN scope

Download and ping traffic use Cloudflare’s public speed endpoint: https://speed.cloudflare.com/__down. Downloads request a fixed byte size (currently 20 MB per sample); latency and jitter use minimal-byte requests on the same host so path and timing stay comparable. This measures performance to that CDN path; speeds to other sites, apps, or regions may differ.

What can affect your result

Browser and device - CPU load, power saving, and other apps.
Other tabs and apps - downloads, sync, streaming elsewhere in the home.
Wi-Fi quality - distance, walls, band, interference.
Router performance - age, firmware, and how many devices are active.
Background traffic - on this device and others on the same connection.
Time of day - peak-time congestion outside and inside the home.
Local contention - shared medium and neighbourhood demand.
Single endpoint - we use one CDN path; other services may differ.
VPNs or proxies - can change path and timing.
Conditions between runs - moving rooms or switching from Wi-Fi to Ethernet will change numbers.

How to use Pulse well

Run more than once, including at the time of day when problems show up.
Keep the tab in the foreground for the run when possible.
Compare Wi-Fi and Ethernet if you can, and optionally different rooms.
Note date and time if you are building a picture for your provider.
Read how to run an accurate broadband speed test for a fuller checklist.

What Pulse can support

Practical home diagnosis - is this device and path performing as expected right now?
Evidence gathering - dated, repeated tests with fair setup.
Spotting repeat patterns - evening vs daytime, Wi-Fi vs wired.
Understanding whether speed, responsiveness, or stability looks like the limiting factor on this path - for this test.

What Pulse cannot prove

Upload performance.
A contract breach or guaranteed fault from one test.
A full provider network diagnosis (routing, backhaul, contention beyond this path).
Performance of every internet service or app you use.
A complete map of your whole-home Wi-Fi environment (this is a browser test along one path).

Privacy and data handling

No account is required. Speed results for this tool are displayed in your browser and are not stored on our servers as part of the Pulse tool. Approximate location or network details shown in the analytics panel come from third-party or browser signals as described on the site; we do not use them to build a personal profile of your tests for this tool. Full detail: privacy policy.

UK context and consumer reality

UK providers advertise broadband using rules that include peak-time averages for many products. Regulators encourage consumers to understand what they buy and to contact providers when speeds are persistently below what they reasonably expect - after fair testing. Household needs vary; always check your contract, minimum guaranteed terms (where applicable), and your provider’s own speed information. This page is educational, not legal advice.

Versioning and review

Method version: Pulse method v3 - April 2026

Reviewed: 11 April 2026

What changed in v3:

Restructured methodology with explicit sections for test flow, sample logic, median, confidence, cache-busting, and CDN endpoint scope.
Documented default sampling (9 download samples), 20 MB payload size, and Cloudflare __down path.
Clarified why the median is the headline (not the fastest spike) and what confidence does and does not mean.
Aligned privacy and UK consumer framing with the live site; expanded related reading to priority answer pages.