Why peak speed is misleading
Most speed tests run for 10–30 seconds and intentionally use multiple parallel streams to hit your line's ceiling. That number tells you what's possible — not what's typical. Real workloads (video calls, gaming, streaming, video conferencing) live on the floor of your connection, not the ceiling. Stability testing measures that floor.
The three numbers that define stability
- Latency 95th percentile — 95% of probes were faster than this. If average is 30 ms but P95 is 200 ms, calls feel laggy even though the average looks fine.
- Jitter — standard deviation of latency. Under 10 ms = smooth. Over 30 ms = stuttery.
- Sustained packet loss — single-digit dropped packets are normal. Sustained loss above 0.5% will break voice and gaming.
How to run a meaningful test
- Test wired first. If wired is stable, your line is fine and Wi-Fi is the issue.
- Run for at least 5 minutes. 30 minutes is better. Cover your worst hours (typically 7–10 pm).
- Don't run anything else heavy during the test — you want baseline, not load.
- Repeat at different times of day to find ISP congestion patterns.
Common patterns and what they mean
- Spikes every 15 minutes — usually a background app or scheduled scan on your device.
- Slow degradation across the evening — ISP peak-hour congestion. Document it for support.
- Loss only on Wi-Fi — interference. Try 5 GHz, change channel, or move the router.
- Random total dropouts — modem thermal cycling or a failing line. Replace the modem, then escalate.
Make it operational, not occasional
One stability test gives you a snapshot. A scheduled recurring test gives you patterns — the kind of evidence that gets ISP issues fixed and IT tickets approved. Pro accounts run continuous stability checks and surface multi-day trends in the dashboard.