A speed test can show 300 Mbps down while your face freezes mid-sentence. The result isn’t necessarily wrong. It’s answering a different question: how quickly your connection moved test data during a short window. A video call needs something harder—steady delivery of your voice and picture, in both directions, without long delays or missing pieces.

That’s why the big download number is often the least useful figure on the screen. To understand a bad call, look at upload speed, latency, jitter, packet loss, and what happens to latency when the connection is busy. Then check whether the trouble is on your network, someone else’s, or the device running the call.

Illustrative image: internet speed test results

A speed test measures capacity, not conversation

Download speed is how fast data reaches you; upload speed is how fast you send it. Both matter on a call. Your connection receives other people’s audio and video while sending your microphone, camera, and screen share. Speed is measured in megabits per second (Mbps).

A typical speed test moves data between your device and a test server. It may show what the connection can deliver briefly, but the result depends in part on the server and network path used. Your meeting may take a different path, and a clean test at noon won’t capture a Wi-Fi problem or an upload that becomes crowded at 3 p.m. Cloudflare’s explanation of how its speed test works makes this distinction clear: a high throughput result doesn’t, by itself, establish that a connection will handle real-time communication well.

Downloading a file can tolerate uneven delivery. If a piece arrives late, the transfer can keep going. A conversation has little room to wait. A late syllable is still a late syllable, even on a connection capable of moving a large file quickly.

The four numbers behind a smooth call

Upload speed: the number your coworkers experience

A plan with ample download speed can still have limited upload capacity. If the upload side is full—perhaps because a cloud backup starts or someone sends a large file—your camera and microphone must compete for space. The people on the call may see you freeze or hear your words break up while you can see and hear them clearly.

For a sense of scale, Zoom recommends 1.2 Mbps in each direction for a one-to-one 720p HD call. Its recommendations rise for some group and higher-resolution calls. Those figures describe the call’s bandwidth needs, not a promise that an internet plan with exactly that much upload will perform well. Other people and apps share the connection, and capacity can vary during the day.

Read the upload result separately from download. If it’s only a few Mbps, a competing upload deserves attention. If it’s comfortably above the call’s needs, don’t stop investigating: plentiful bandwidth won’t repair unstable delivery.

Latency: how long a response takes

Latency is delay, usually shown in milliseconds (ms). Speed tests commonly display a round-trip measurement: how long a small exchange takes to reach the test destination and return. On a call, too much delay leaves awkward pauses and causes people to talk over one another.

Check whether the test shows both idle and loaded latency. Idle latency is measured when the connection is quiet. Loaded latency is measured while data is being transferred. If idle latency is low but loaded latency climbs sharply, traffic may be waiting in a queue when the connection gets busy. That can happen even while the test reports a high download speed. Cloudflare’s test measures latency during transfers as well as at idle, which makes the difference easier to spot.

For example, a result of 20 ms idle and 180 ms during upload tells a more useful story than “300 Mbps down.” It suggests the call may feel responsive until another device starts sending data. It doesn’t, on its own, identify whether your router, Wi-Fi, or internet provider is responsible.

Jitter: how uneven that delay is

Jitter measures variation in packet timing. A connection can have a reasonable average latency but deliver some packets promptly and others much later. The call software can absorb some variation, but irregular arrivals can make audio sound choppy or video stall.

Treat jitter as a pattern, not a verdict from one reading. A short test might miss a burst of instability that occurs every few minutes. As a reference for its own meeting statistics, Zoom recommends jitter of 40 ms or less. That isn’t a universal cutoff for every calling app or network test; it’s most useful alongside what you actually experience during the call.

Packet loss: the pieces that never arrive

Packet loss is the share of data packets that fail to reach their destination. A test that shows 1% loss means roughly one in every hundred measured packets was lost during that test—not that every call will lose packets at the same rate. Missing packets can show up as clipped words, blocky video, or brief freezes.

A zero-loss result is reassuring but limited. The test may have been too short to catch an intermittent problem, or it may have tested a different path from your meeting. Conversely, loss reported inside your calling app while the call is breaking up is strong evidence that data delivery, rather than raw download speed, needs attention.

How to read a “fast” result

Imagine a household speed test shows 300 Mbps down, 8 Mbps up, and 18 ms idle latency. At first glance, nothing looks wrong. But loaded latency rises to 200 ms during the upload test, and jitter jumps when a laptop begins syncing files. The call becomes unreliable at the same time.

The useful reading is not “8 Mbps is too slow for video.” It may be enough for the call on its own. The stronger clue is that other upload traffic makes delivery uneven and delayed. Pause the sync and repeat the test: if loaded latency falls and calls improve, you’ve found a practical cause without buying a faster plan.

Use this order when scanning your results:

Result Question to ask
Download speed Is it sufficient for the video and other activity you’re receiving?
Upload speed Is there enough capacity left while the household is using the connection?
Idle versus loaded latency Does delay rise sharply when uploads or downloads run?
Jitter Does packet timing become uneven, especially during a bad call?
Packet loss Are packets going missing, even if the speed figures look healthy?

No single row proves the cause. The aim is to match a measurement to the symptom and the time it happens.

Test the problem where it happens

Start on the device and connection you use for calls. A speed test run on a phone next to the router says little about a laptop using Wi-Fi in a back bedroom. Test at your usual workspace, then repeat at the time calls tend to freeze. Let each test finish, and avoid running a bandwidth-heavy test in the middle of an important meeting—it can create the very congestion you’re trying to measure.

Next, compare a quiet connection with a busy one. Run a test with backups, downloads, and other heavy traffic paused; then observe what changes when ordinary household activity resumes. Note upload speed, idle and loaded latency, jitter, and packet loss, not just the headline download figure. A result that deteriorates only when the connection is busy points toward competition for capacity or traffic queueing.

Check the calling app during the problem, too. Zoom’s desktop app has a Statistics view for audio and video that shows measures including latency, jitter, packet loss, and CPU use. In Google Meet, Troubleshooting & help shows a network stability graph. These views are especially useful because they reflect the call you’re having, rather than a separate test to another server.

Finally, try the same laptop on Ethernet if you can. If the call becomes stable while the rest of the setup stays the same, focus on Wi-Fi coverage, interference, or the wireless equipment before blaming your internet plan. If the problem remains on Ethernet and appears across devices, record the times and test results for your provider or workplace IT team. A single bad result is less persuasive than a repeatable pattern.

Match the fix to the failure

When calls fail only as someone uploads files, pause cloud sync and backups during meetings. If that consistently helps, examine your router’s traffic-management options or the upload capacity of your service. Don’t assume that paying for a bigger download number will change the upload side.

When Ethernet fixes the call, work on the wireless link: move closer to the router, try a less congested Wi-Fi band if your equipment supports it, or use a wired connection for important calls. Google Meet recommends Ethernet as a way to test and improve meeting quality. Replacing equipment may help in some homes, but make the wired comparison first so you know what problem you’re trying to solve.

When the network measurements look sound but video still freezes, check the computer. A busy processor, too many open apps, or demanding video effects can disrupt a call without producing a bad internet speed result. Try closing heavy applications or lowering video quality and watch whether the symptom changes.

There’s also a crucial direction check: if only one person’s picture freezes for everyone, the trouble may be with that person’s outgoing connection or device. If you alone can’t see or hear several participants, your receiving side is a better place to start. A speed test describes your connection to its test server; it can’t certify everyone else’s connection.

A fast result is useful evidence that your connection can move data. For video calls, the better question is whether it can deliver small pieces of audio and video steadily, in both directions, while the rest of your network is in use.