Every speed test shows a big download number. That’s roughly where the similarities end. Under the surface, these tools make very different choices — about what to measure, where to measure it from, and how much to explain — and those choices change what your result actually means.
We just shipped a major SpeedPulse update built around jitter, bufferbloat grading and explained verdicts, so it felt like the right time to lay out the landscape honestly: what each tool does well, where each falls short, including our own. Test every claim yourself — that’s rather the point of this article.
The comparison matrix
And the same data in crawlable form, with nuance the matrix can’t hold:
| Criterion | SpeedPulse | Ookla Speedtest | Fast.com | Waveform |
|---|---|---|---|---|
| Download / upload | ✅ | ✅ | ✅ | ✅ |
| Idle ping | ✅ | ✅ | ✅ | ✅ |
| Jitter | ✅ 8-sample measurement | ✅ shown in results | ❌ | ❌ not reported |
| Latency under load | ✅ separate download & upload phases | 🟡 recent apps/CLI | 🟡 single loaded figure | ✅ |
| Bufferbloat letter grade | ✅ A+ to F | ❌ | ❌ | ✅ their core feature |
| Explained verdicts | ✅ every score says why | ❌ | ❌ | ❌ |
| Mbps ⇄ MB/s toggle | ✅ | ❌ | ❌ | ❌ |
| History export, no account | ✅ CSV, one click | ❌ account required | ❌ no history | 🟡 shareable page only |
| Multi-CDN cross-check | ✅ 8 CDN edges after every test | ❌ | ❌ | ❌ |
| Runs in browser, free | ✅ | ✅ | ✅ | ✅ |
What the matrix doesn’t show: the methodology gap
The nearest-server problem
A speed test measures the weakest link between you and its server. Most tests pick the nearest, least-congested server they can — which is precisely not where your Netflix packets, game servers or work VPN live. Gigabit to a server 10 km away says little about your evening experience with a service on another continent.
Fast.com’s answer is elegant: it tests against Netflix’s own CDN, so the number approximates your actual streaming experience. Waveform chose interpretability: rather than more raw numbers, they popularized the bufferbloat grade. Ookla’s strength is scale and history — their server network is unmatched, which makes their relative data (your line vs your neighbors’) genuinely valuable.
SpeedPulse’s answer is the multi-node panel: after every test, we ping 8 independent CDN edges — Cloudflare, Fastly, jsDelivr, Bunny, UNPKG, GitHub, Google, AWS — and show the latencies side by side. If your speed-test CDN is dramatically faster than everything else on the chart, that’s a signal worth knowing about: some networks treat speed-test traffic more gently than real traffic, whether by caching, peering or traffic shaping. A single-CDN test structurally cannot show you that.
The “flattery” problem
A speed test that consistently reports the highest number wins the popularity contest — this is a known failure mode of the industry. Reporting a 95th-percentile burst instead of a sustained rate, testing over a warm connection, or ignoring the upload entirely all make results look better. None of these are lies, exactly. They’re choices. Our stance: publish the methodology, show the sample counts, and let readers argue. (Ours is documented in how SpeedPulse measures jitter and bufferbloat — including its limitations, of which there are several.)
The explanation problem
Ookla’s mobile apps show glossy rating dots; Fast.com shows one number and hides everything else behind a “Show more info” link. Both stop short of answering the only question a normal person has: “is my internet good enough for what I do?” That’s why we ship explained verdicts — every score for streaming, gaming, calls and browsing names the metric that limited it. We’d rather be argued with than merely trusted.
Which test should you use?
- Diagnosing lag on a fast line → a test with latency-under-load and a bufferbloat grade (SpeedPulse, Waveform). Start here.
- Quick sanity check while troubleshooting streaming → Fast.com, since it measures against Netflix’s CDN.
- Comparing ISPs or proving a fault to support → SpeedPulse’s CSV history (timestamped jitter + bufferbloat trends are exactly what support desks can’t wave away) plus Ookla for its brand-recognition screenshots.
- Checking whether your speed test is being gamed → any test with a multi-CDN cross-check. (Ours. But even running two different speed tests back to back — and comparing — beats trusting one.)
Discussion
- Have you ever gotten wildly different numbers from two tests seconds apart? Which two, and which direction?
- Do you care about jitter and bufferbloat, or is bandwidth still the only number that matters to you? Genuinely curious where the crowd lands.
- Should speed tests publish their measurement methodology the way we’re trying to? What would you want disclosed first?
If this comparison helped you, the share bar has the usual suspects — and a Hacker News link, where methodology debates like this tend to get properly adversarial.