Why a clean speed test can hide a bad connection
A standard speed test mostly reports throughput, meaning how much data moves per second. Many also show a ping figure. In Gogia's account, that ping usually reflects idle latency, measured while the connection isn't busy. (That's her description of the typical test.) The catch is that the figure says nothing about how the line behaves when something else is using it.
That's where bufferbloat comes in. Waveform, the company behind the test Gogia used, describes bufferbloat as high latency caused by your router buffering too much data. The test itself measures how much latency rises under load, which causes the lag spikes behind stuttering video calls and unplayable games.
Picture a single-lane road with a very long on-ramp. Throughput is the road's capacity. Bufferbloat is the line of cars waiting at the ramp. Your packets get there eventually, but a video call is a live conversation and can't wait that long.
Upload is usually the first thing to saturate. Home connections typically give you far less upload than download, so a photo backup or cloud sync can fill the queue. Video calls depend on upload, so they suffer first. This matches Gogia's experience. She noticed trouble when uploading to Google Photos, and a call with her team lead fell apart.
What the Waveform test does
Gogia describes a test that needs no app or account and runs in any browser in about a minute. Independent descriptions agree on the method. It measures baseline latency, then runs a download saturation phase and an upload saturation phase, measuring the latency change in each. Whichever phase causes the larger increase determines the final grade. A third-party summary also notes the test relies on Cloudflare's measurement infrastructure.
How to run it properly
- Pause streaming, game downloads, backups and cloud syncs on every device in the house.
- Don't run another speed test at the same time. The bufferbloat test generates its own heavy traffic.
- Press Start Test and wait for the idle, download and upload phases to finish.
- Note the grade and the latency figures for each phase.
- If possible, repeat the test over Ethernet, then over Wi-Fi, and compare.
For the most useful result, run it when your calls are actually misbehaving. One troubleshooting guide, SpeedtestHQ, makes this point: bufferbloat only appears under load, and if calls degrade when someone else is downloading or streaming, it is the most probable cause.
Reading the grade
The letter reflects added latency, not how fast your plan is. Sources differ slightly on the exact cut-offs, so treat the numbers below as approximate:
| Grade | Added latency under load (approximate) |
|---|---|
| A+ / A | Small increase. One guide puts A at under about 30 ms, while another says an A means under 5 ms |
| B | Roughly 30 to 60 ms added |
| C | Occasional spikes during heavy use |
| D / F | Several hundred milliseconds or more added |
One explainer notes the grade is based on how much latency rises under load. A guide from the Australian provider Capti groups results as A+ to B as low bufferbloat, C as moderate, and D to F as high. I couldn't confirm Waveform's exact published thresholds directly, so the table above is only a rough guide.
Gogia's own result was dramatic. Her latency passed 800 ms under load and the grade was an F. That is her firsthand figure and I haven't independently verified it. It does match the pattern of a freezing call followed by a robotic voice.
A note on her "rule of thumb" table
Gogia includes a table of healthy versus worrying values for idle latency, packet loss and jitter, and she says thresholds vary by app. Treat those as informal guidance, not standards. The Waveform grade is based on added latency under load. It isn't a measure of idle ping, download speed or upload speed.
What the test can't tell you
A bad grade is a strong clue, but it is not a full diagnosis.
- It doesn't pinpoint the culprit. The queue could be in your router, your modem or your ISP's equipment. Wi-Fi itself can also be the bottleneck.
- It isn't a packet-loss test. Waveform's own guidance, as reflected in its page text, says that if you see problems over Ethernet with nothing else running, you should run a proper packet loss test with ping.
- It won't catch long-term dropouts. A one-minute test can't show whether a line stays stable for hours.
- Browsers vary. Waveform says it used workarounds to make the test work across browsers, but the grade and the increase in latency should be fairly consistent. Expect small differences between runs and between devices.
If you fix the loaded latency and calls still freeze, look elsewhere. Check Wi-Fi signal and interference, the conferencing app, your PC's CPU and driver health, and your ISP line.
How to fix a poor result
Gogia recommends trying free settings changes before paying for a faster plan. That is sensible, because extra bandwidth doesn't remove a queue that sits in your router.
- Test over Ethernet. This rules out Wi-Fi interference and weak signal. If the grade improves a lot on a cable, your wireless link is the main suspect.
- Pause big transfers during calls. If calls improve, competing traffic is involved.
- Look for Smart Queue Management (SQM). Gogia says this reduces delays when the network is busy, and she lists eero, Asus, TP-Link and OpenWrt setups as offering it. Menu names and availability vary by router and firmware, so check your own router's documentation. One troubleshooting site advises that a D grade calls for SQM on the router, with testing on Ethernet to rule out Wi-Fi.
- Try QoS if SQM isn't offered. QoS is a broad label, and it isn't necessarily equivalent to SQM. It worked for Gogia, but results will vary by router.
- Update the firmware. Gogia suggests this if results stay poor after you configure traffic management.
- Re-test under the same conditions. Compare like with like, with the same device, the same connection type and the same quiet household.
- Escalate. If nothing helps, contact your ISP or consider a router with better queue management.
SQM does have a cost. It runs on the router's processor, so a weak router may struggle to shape traffic at full gigabit speeds. That is general networking knowledge, not something Gogia reported.
Her result
After enabling QoS, Gogia says the added upload latency was only about 30 ms and her grade rose to a B. She says her video calls stopped being choppy. That is one person's before-and-after on one router, not a guarantee. A B is a good improvement from an F, but it still has some added delay.
Why this matters on a Windows PC
Windows laptops and desktops running Teams, Zoom or Meet are often blamed for call problems that start in the home network. Before you reinstall drivers or blame the app, a one-minute test can show whether the queue is the problem.
The same advice applies if you manage remote staff. A "my connection is fast" report doesn't rule out bufferbloat. Asking a user to run the test and send a screenshot of the grade is a quick triage step.
Bottom line
- A fast speed test measures capacity, not responsiveness under load.
- Waveform's free browser test grades how much latency rises when your line is busy.
- Test when the household is quiet, then compare Ethernet with Wi-Fi.
- Try SQM or QoS and a firmware update before buying a faster plan.
- If the loaded latency is fixed and calls still fail, look at Wi-Fi, packet loss, the app and the ISP.
Gogia's claim that the fix takes ten minutes and one setting won't hold for every router. It is still a cheap test to run before you blame the plan or the app.