What are jitter and packet loss, and why do they matter more than speed?
Because a video call does not need much bandwidth — it needs packets arriving consistently, and these two measures describe whether they do.
Latency is how long a packet takes to travel, measured as round-trip time in milliseconds. Lower is better, and it is governed largely by distance and by the number of hops.
Jitter is the variation in latency. If packets take 30ms, then 31ms, then 29ms, jitter is tiny. If they take 30ms, then 120ms, then 25ms, jitter is severe — even though the average is acceptable.
Why jitter breaks real-time media. Audio and video must be played out at a constant rate. If packets arrive irregularly, the player has nothing to render at the moment it needs it. Applications compensate with a jitter buffer, deliberately holding packets briefly to smooth out the variation — which costs added delay. Larger jitter needs a larger buffer, which makes the conversation feel less responsive. Beyond what the buffer can absorb, audio breaks up.
Packet loss is the proportion of packets that never arrive. Causes include congestion — where a queue overflows and packets are discarded, the most common cause — Wi-Fi interference, faulty cabling, and overloaded equipment.
Its effects differ by protocol. TCP retransmits, so loss appears as slowness rather than corruption. UDP does not, so loss appears directly as glitches, dropouts and frozen video.
Why loss hurts TCP disproportionately. Congestion control interprets loss as a signal to slow down, so even 1–2% loss can reduce throughput dramatically — far more than the raw percentage suggests.
This is why a speed test can look fine while calls are terrible. A speed test measures sustained throughput over seconds and is largely indifferent to brief variations. Jitter and loss are precisely those variations.
What causes it at home: Wi-Fi rather than the internet connection, in the large majority of cases; bufferbloat, where oversized queues in routers add enormous latency under load; another device saturating the upload; and interference on 2.4GHz.
What helps: wired Ethernet where possible; a router with smart queue management to control bufferbloat; and testing latency under load, not idle.