Start here
A result is only useful once you know what you asked for
A speed test runs a file transfer between your device and a nearby server and reports how much data moved per second in each direction, how long a single round trip took, and whether any of those round trips failed to arrive. That is the whole measurement. The difficult part is not running it — it is deciding whether the numbers you got describe a healthy connection or a home network with a problem in it.
This guide walks through the five figures that appear on almost every result screen, what each one actually measures, and the range that holds up for streaming, video calls, gaming, and ordinary browsing on a U.S. home connection. Nothing here replaces a test on your own line, but it gives you a way to read one without guessing.
The five figures
What each number on the result screen measures
Read them in this order. Download and upload describe capacity; latency, jitter, and packet loss describe how steady that capacity feels while you are using it.
Download
Download speed in Mbps
How much data your connection pulls toward you each second. It sets the ceiling for anything you watch or receive: streaming in 4K, a large game update, a cloud photo library syncing down. Test sites report it in megabits per second, written Mbps.
Upload
Upload speed in Mbps
How much data leaves your device each second. Your camera feed on a video call, a file you send to work, a backup leaving the house, and a livestream all depend on it. Cable plans usually give you far less upload than download, which is normal and not a fault.
Latency
Latency in milliseconds
The time a small packet takes to travel to the server and back, measured in ms. Also called ping. It matters most in conversations and games, where a slow reply feels worse than a narrow pipe. Any figure under about 30 ms is a fast response on a home connection.
Jitter
Jitter in milliseconds
How much latency bounces around during the test. A steady 40 ms ping is workable; the same 40 ms average with swings of 30 ms in each direction is what makes a call sound clipped and a game feel unresponsive. Look for jitter under about 10 ms on a wired connection.
Packet loss
Packet loss as a percentage
The share of test packets that never arrived. Even a fraction of a percent shows up as a frozen frame, a gap in audio, or a dropped round in a game, and unlike a slow speed it will not improve just because the line is otherwise fast. Any reading above zero is worth investigating.
What good looks like
Ranges that hold up for everyday use
These are working ranges for a single typical household device, not a promise from any provider. Your plan, the time of day, and how many people are online at once all move the line. Use them as a sanity check when a result looks wrong.
A speed test that reports comfortably above the range for what you actually do means the connection is doing its job, even if the headline number is lower than the one on your bill.
Comfortable ranges by activity, per device
Everyday browsing
Download 10 Mbps or more
Latency under 100 ms is fine
Streaming in 1080p
Download 8 to 15 Mbps
Jitter under 20 ms
Streaming in 4K
Download 25 Mbps or more per stream
Packet loss at or near zero
One-on-one video calls
Upload 3 to 5 Mbps
Latency under 50 ms, jitter under 15 ms
Online games
Download 15 to 25 Mbps to install and patch
Latency under 40 ms, packet loss at zero
Working from home with a shared screen
Upload 10 Mbps or more, download 50 Mbps or more
Latency under 60 ms on both directions
Add ranges together when several devices stream at once
Solid — steady line
Speed held a flat rate for the whole test. This is the reading you want on download and upload.
Dashed — variable rate
Throughput rose and fell during the run. Common on Wi-Fi, on a busy evening, or when another device is saturating the line.
Crossed — unsteady timing
Latency swung while the test ran. Expect clipped audio on calls and rubber-banding in games even when the speed looks fine.
Dotted — broken delivery
Packets went missing. Treat this as a fault to chase down rather than a slow day on the network.
Before you trust it
Run it twice
A single reading is a sample, not a verdict.
Same device, both times
Swapping hardware mid-comparison hides the real change.
Quiet the house
Pause backups, updates, and streams while you measure.
Run a speed test you can actually compare
A speed test measures the path it happens to take, so the way you run it changes the answer. A laptop on the far side of an apartment, on Wi-Fi, while a console downloads an update, will report a much smaller number than the same line measured on a cable near the router with the house quiet. Both results are real. Only one tells you whether your plan is delivering.
For the cleanest reading, plug into the router with an ethernet cable if you can, close streaming apps and pause cloud backups, then run the test two or three times and look at the spread rather than a single pass. If those repeats agree with each other, you are reading your line. If they disagree by a wide margin, you are reading something in between your device and the router.
Repeat the same run later in the evening. Home networks are busiest between roughly 7 and 11 p.m., when neighbors on the same street are streaming at once. A cable or fiber line that reads several hundred Mbps at noon and a fraction of that at night is telling you about congestion on the shared portion of the network, not about the hardware in your living room.
Keep a short note of the readings you collect: date, time, whether you were wired or on Wi-Fi, and which figures moved. Three consistent results are far more useful when you talk to your provider than one dramatic screenshot.
When a number looks wrong
Match the symptom to the figure that explains it
Buffering while streaming
Download is short or jitter is high
If download sits below the range for the quality you are watching, the picture will stall while it refills. If download looks generous but the stream still stutters, check jitter and packet loss instead — the pipe is wide but the road is lumpy.
Frozen video calls
Upload or latency is the limit
Calls send as much as they receive. A thin upload makes your camera feed break up for everyone else, while high latency makes you talk over each other. Small improvements in either figure show up quickly here.
Lag in games
Latency, and packet loss above zero
Games move very little data but need replies fast and in order. Latency sets how quickly your actions register; packet loss is what makes a character snap backward. Bandwidth rarely fixes either.
Everything feels slow
All figures moved together
Download, upload, and latency dropping together points at the connection itself or at how your device reaches the router, rather than one service misbehaving. Re-run wired near the router to separate the two.
What the reading describes
One figure describes one path at one moment
A result describes the route between your device and the server that responded, at the time you pressed start. Change any part of that and the numbers move: a different test server, a different room, a different hour. That is why a single reading on its own should not convince you the plan is broken, or that it is perfect.
Compare like with like. Wired to wired, Wi-Fi to Wi-Fi, same device, similar time of day. The gap between those two pairs is what usually explains a reading that looks smaller than the plan you pay for.
Wired at the router
This reading isolates the line itself. If it meets or beats your plan, the connection arriving at your home is doing its job.
Wi-Fi across the house
This reading describes what your devices actually experience. Walls, distance, and channel crowding show up here, not at the line.
Peak evening
Repeating the test between 7 and 11 p.m. shows how the shared part of the network behaves when the whole street is online.
Common questions
Reading results in practice
The questions that come up most often once someone has a result in front of them and is not sure which part to trust.
Why is my speed test lower than the plan I pay for?
Plan figures describe the line into your home, measured under favorable conditions. A test on Wi-Fi, several rooms away, with other devices busy, will land well below that. To compare fairly, run the test wired to the router with the house quiet. If the wired reading still falls far short of the plan, that is the result worth taking to your provider.
Is a high download speed enough on its own?
No. Download tells you how much can arrive each second, but not how evenly it arrives or how quickly the network answers. A connection with a wide download and poor latency still stutters on calls and games. Read the five figures together and let the weakest one set your expectations.
How many times should I run a speed test before I believe the result?
Two or three runs on the same device, at the same time of day, is enough to see whether a reading is representative or a one-off. If the repeats sit close together, treat that as your baseline. If they swing widely, note both ends of the range — the spread itself is useful information about your connection.
Does the server I test against change the numbers?
Yes, and it should. The distance and route to the server add to your latency and can cap throughput. Test tools usually pick a nearby server automatically, which is the right default for judging your connection. A far-away server is useful when you want to know how a service hosted in another region will feel.
What should I do if packet loss shows anything above zero?
Repeat the test wired to the router to see whether the loss survives a cleaner path. If it does, note the time and the readings and take them to your provider, who can look at the line from their side. If it disappears when wired, the loss is happening over Wi-Fi or inside your own equipment, and that is where to work next.
Do I need the same numbers on every device in the house?
No. Older phones and laptops may negotiate a slower Wi-Fi standard than newer ones, and a device in a back bedroom will always lose more than one next to the router. Judge each device against what it needs to do, and use a wired reading near the router when you want to know what the line itself provides.
Keep reading
Put the numbers to work on your own connection
Run a speed test on a wired device, then on Wi-Fi from where you actually sit. Compare the two and read the gap. If something in the results still does not add up, we are in Brooklyn and reachable during business hours.
Monday to Friday, 9:00 AM – 6:00 PM · +17185243042
Prentiss Scholey Speed Test · 2724 Clarendon Rd, Brooklyn, NY 11226, USA