Latency, Jitter and Packet Loss in Plain Language

One connection reports five hundred megabits and still makes calls unusable. Another reports eighty and feels perfectly responsive in every application. The difference lives in three figures that almost no advertising mentions.

Latency, jitter and packet loss describe how a connection behaves in time rather than how much it can carry, and for interactive use they matter more than bandwidth.

1. Latency, defined without jargon

Latency is the time a packet takes to travel to a destination and back, measured in milliseconds. A wired connection in a well-connected region typically sits between eight and twenty milliseconds.

It is not the same as speed. A connection can move enormous amounts of data and still take a long time to respond to the first request, which is exactly what an interactive application notices.

2. Where latency comes from

Physics sets a floor through the distance the signal must travel, and processing adds to it at each device along the route. Queuing contributes the largest variable component, and it is the one that worsens under load.

This is why latency is usually acceptable at noon and poor at nine in the evening, even when the same route and the same equipment are in use.

3. Jitter as inconsistency rather than delay

Jitter is the variation in latency between successive packets. A steady forty milliseconds is workable for most applications; alternating between twenty and ninety is not.

Voice and video are built to absorb a small amount of delay and very little variation. This is why jitter, rather than raw latency, is the figure that explains a call that sounds broken rather than slightly delayed.

4. Packet loss and why it is the worst of the three

Packet loss is data that never arrives. Transport protocols retransmit it, which converts loss into delay, and real-time protocols cannot wait, so they simply drop the missing audio.

Even one percent loss is audible on a call. Sustained loss above that level makes interactive applications effectively unusable regardless of how much bandwidth is available.

5. Bufferbloat, the cause most households never hear about

Routers queue packets when the connection is busy, and large queues add delay long before the connection is actually full. The result is a line that tests well and feels sluggish the moment anything else is happening.

Some routers expose queue management settings that control this, and enabling them can transform perceived performance without changing the package at all. It is the most underused setting in home networking.

6. Measuring the three figures

A wired test reporting latency under load is the useful measurement, since idle latency is almost always good. Tools that report latency during an active transfer reveal the queueing behaviour that matters.

Recording latency, jitter and loss together across several evenings gives a profile. The profile identifies which of the three is responsible, and they have different remedies.

7. Wireless contributions to all three

Wi-Fi adds its own latency, its own jitter and its own losses through retransmission and interference. A wired connection to the device removes these contributions entirely and often resolves the complaint.

Where wiring is impossible, reducing the distance, moving to a clearer channel or adding an access point reduces all three figures at once.

8. Technologies and their inherent latency

Fibre and cable sit at the low end, copper slightly above, fixed wireless higher still because of the shared radio link, and satellite dramatically higher because of the distance to orbit.

That hierarchy is fixed by physics and cannot be improved by equipment. Any application requiring quick responses over satellite will struggle no matter what hardware is installed.

9. When latency matters less than bandwidth

Bulk file transfers, streaming playback and software updates care about throughput and tolerate delay. Households whose use is dominated by those are well served by judging on bandwidth alone.

Interactive gaming, video calling, remote desktop work and voice over IP sit in the opposite category. For those, the three timing figures decide whether the connection is usable.

10. What can realistically be improved

Wiring devices, enabling queue management on the router, and reducing the number of simultaneous heavy transfers are all within a household’s control and all reduce latency under load.

The provider’s contribution is upstream capacity and routing, which improve when congestion is reported with evidence and rarely otherwise. Separating the two categories prevents wasted effort.

11. Choosing a package on the right figures

For interactive use, ask about latency and packet loss rather than headline speed, and prefer a technology with a low floor. For bulk use, the bandwidth tier is the figure worth paying for.

Matching the measurement to the actual workload produces a connection that feels right in daily use, which is a different objective from winning a speed test.

15. The three figures and the applications they govern

Latency governs how quickly anything responds, jitter governs how consistently it does so, and packet loss governs whether the data arrives at all. Between them they determine the experience of every interactive application.

Throughput governs how much can be moved and has almost no effect on any of the three.

16. What a normal profile looks like

A healthy wired connection shows low single-digit jitter and no packet loss at idle, with latency rising modestly under load. Under load, latency rising by more than a hundred milliseconds indicates a queueing problem.

Recording the idle and loaded figures side by side produces that profile and identifies the specific weakness.

17. Remedies in order of cost

Queue management on the router costs nothing, wiring devices costs a cable, and changing technology costs a contract. Working through them in that order resolves most problems before the expensive option is reached.

Most households find that the free option addresses the majority of the issue, which is why the measurement matters more than the purchase.

18. Why the numbers are worth knowing at all

Households that know their latency and jitter can judge whether a new package will help, whether a router setting is worth changing, and whether a complaint has merit.

Without them, every decision about an interactive connection is made on throughput alone, which is the one figure that does not describe the problem.

19. Tools that report the three figures together

Some speed tests report latency and jitter alongside throughput, and some report latency under load specifically. Choosing one that reports all of them removes the need for several separate tools.

Consistency matters more than the specific tool, since results from different tools are not directly comparable.

20. Interpreting a good result that feels bad

Where throughput, latency and loss all look acceptable and the experience is still poor, the application or the device is the likely cause. Testing the same application on another device isolates it.

Reproducing the problem on a second device that behaves well is evidence that the original device is the subject rather than the network.

21. When to stop measuring and change something

Measurement identifies the cause; it does not remove it. Once the pattern is established, the appropriate change can be made and the measurement repeated afterwards.

Households that measure indefinitely without changing anything eventually stop learning anything new from the numbers.

Bandwidth describes capacity and timing describes experience. Households that measure both stop buying connections that are fast on paper and slow to respond.