Methodology · Measurement Practice
How We Test and Report
What every speed test on this site measures, how long each measurement runs, and which conditions can push a reading away from the number your connection actually delivers.
A speed test is only worth trusting when you know what it measured and what it ignored. This page lays out both: the instrument behind every reading we publish, and the editorial rules that keep our guidance aligned with how U.S. broadband actually behaves.
A U.S.-focused internet speed test resource covering download speed, upload speed, latency, jitter, and packet loss across fiber, cable, Wi-Fi, and cellular connections.
What we measure
Five numbers, five different jobs
Each figure in a result panel answers a separate question about your connection. Reading them together tells you whether a problem lives in the line, the router, or the device you are holding.
Download speed
Megabits per second
The rate at which data reaches you. It governs streaming stability, large file transfers, and how many devices can pull at once before something buffers.
Upload speed
Megabits per second
The rate at which your data leaves the house. Video calls, cloud backups, and shared screens depend on it far more than most people expect.
Latency
Milliseconds
How long a single round trip takes. Low latency keeps calls natural and games responsive even when headline speeds look generous.
Jitter
Milliseconds of variation
How much that round trip wanders between samples. Steady beats fast: a stable 30 ms feels better than a jumpy 12 ms.
Packet loss
Percentage of packets
Data that never arrives. Even a small amount shows up as frozen video, dropped audio, and retries that slow everything down.
How a run works
One measurement, step by step
A reading is not a single ping. The tool opens several parallel streams, warms them up, discards the ramp, and only then records what the link held under load. That sequence is why a proper run takes roughly thirty seconds rather than two.
How to read your resultsConnect & warm up
The client picks a nearby server, opens connections, and pushes traffic for a few seconds purely to stabilize the route. Nothing measured here counts.
Download load
Multiple streams pull at once until throughput plateaus. We record sustained rate, not the first burst that a cache or a modem buffer can fake.
Upload load
The same approach in reverse. Upload tests matter more than download figures for anyone working from home, backing up, or streaming out.
Latency sampling
While the link is busy, small probe packets time their round trips. Jitter comes from how much those samples disagree with each other.
Loss & report
The client counts probes that never returned, then averages the run and flags variance high enough to make the result worth repeating.
Between the line and the reading
Your connection is not the only thing on the route
Between the fiber in the street and the screen in your hand sit a modem, a router, a wireless link, and whatever else is streaming in the house. A single number cannot describe all four.
What moves the number
Conditions that shift accuracy
A speed test is a snapshot taken under whatever conditions happen to exist at that moment. Two runs ten minutes apart can differ without anything being wrong. These factors move a reading the most, and each one has a way to check it.
Wi-Fi
Wireless distance and interference
Walls, floors, and neighboring networks cut wireless throughput before the data ever reaches the line. A wired test at the router tells you whether the shortfall is in the air or in the service.
Busy
Other devices sharing the link
Streaming, backups, and game downloads compete for the same bandwidth. If a run is lower than usual, check what else is awake before blaming the provider.
Peak
Time of day on the network
Shared neighborhood capacity dips during evening hours on many cable routes. Testing at midday and again at night shows whether the pattern is yours or the area's.
Server
Which endpoint answers
A distant test server adds travel time that is not part of your connection. Comparing nearby and distant servers separates real slowness from routing detours.
Device
Hardware on your end
An older phone, a saturated laptop, or a router working hard on other traffic can cap a result below what the line actually delivers. Repeating on a second device is the fastest check.
Before you trust a reading
- Run the test twice and compare. Agreement between runs matters more than the higher figure.
- Pause big downloads and streaming on other devices for the length of the run.
- Test from the router with a cable, then from your usual seat, and note the gap.
- Keep both figures for a few days. A pattern explains more than any single result.
Line-state reference
Three connection types, read the same way
We draw the same four line patterns on every page so a reading can be compared across chapters. Fiber, cable, and wireless routes differ in where the loss tends to appear, not in how it is recorded.
Solid · steady line
Symmetric fiber is the usual source of a flat, repeatable reading. Little variation between runs.
Dashed · shared line
Cable routes share capacity. Expect more movement in the evening than a fiber link shows.
Dotted · wireless link
Throughput depends on distance, walls, and channel crowding. The final hop matters as much as the provider.
Crossed · loss present
Packets disappearing before arrival. Small percentages cause freezes and retries that a speed figure hides.
How guidance stays current
We revisit pages when the rules change
Speed tiers, wireless standards, and the way providers shape traffic all move over time. When a published definition or a measurement convention shifts, the pages that depend on it get rewritten rather than patched. Definitions in the glossary stay short and plain so a reader can check one term without reading a whole article.
Where a result points toward a fault, we send readers to the troubleshooting pages instead of repeating general advice here. Each guide covers one problem: slow wireless, inconsistent evenings, or a link that drops entirely.
Reporting rules
What we publish, and what we leave out
Editorial choices made before any number reaches a page. They exist so a visitor can tell the difference between a measurement and an opinion.
Every figure on the site comes from a live measurement
We do not display sample results, editorial placeholders, or rounded-up numbers dressed as readings. If a page shows a value, a test produced it.
Explanations name their limits
When guidance only applies to one kind of connection, we say so on the page. A wireless workaround is not offered as a fix for a faulty line.
Provider claims are treated as claims
Advertised tiers describe what a plan is sold as, not what a device receives. We keep the two separate so a disappointing reading does not read like a broken promise.
Updates replace pages instead of stacking on them
Corrected guidance replaces the original wording. Readers should not have to compare dates to find out which paragraph is still true.
Questions we get
Answers about accuracy
The questions readers send most often about how a speed test is run and what its results mean.
ContactHow long should a speed test take?
A full run with warm-up, download, upload, and latency sampling usually takes around thirty seconds. Shorter tests exist, but they skip the settling period and tend to report bursts rather than sustained throughput.
Why does my result change between runs?
Because conditions change. Another device may have started streaming, a wireless channel may have gotten busier, or the route to the test server may have shifted. Two runs compared side by side tell you more than either one alone.
Should I test over Wi-Fi or with a cable?
Both, and compare them. A wired run at the router shows what the service delivers; a wireless run from where you actually sit shows what you experience. The difference between them is usually the wireless link, and that is the part you can often fix.
Do you publish results from your own testing?
No. This site runs measurements on request for the visitor sitting at the connection, then explains what those readings mean. We do not publish ranking lists or averages for providers, because a single connection is not a fair sample of a whole network.
I see packet loss but normal speed. Which matters more?
Both describe different problems. High throughput with packet loss usually means the link is fast but delivering data unreliably, which shows up as stuttering media and failed downloads. Repeat the run first; if loss persists, work through the slow-internet guide before assuming the hardware is at fault.
Where to go next
Run the test, then read what it told you
Start a speed test from the home page, then use the results page to match each figure against a plain explanation. If a reading still looks wrong after repeating it, send us the details and we will point you at the right guide.