Network performance metrics are the measurements that show how well a network moves data: its speed, its reliability, and how efficiently it carries traffic. Track the right ones and you can catch a problem before users feel it. Track none, and you'll find out when the complaints arrive. This guide covers what the metrics are, what affects them, how to measure them, and the rough targets that separate a healthy network from a struggling one.
Table of contents
- What are network performance metrics?
- What affects network performance?
- Key network performance metrics
- How to measure network performance
- What good looks like: benchmarks and targets
- Network metrics and video call quality
- Frequently asked questions
What are network performance metrics?
Network performance metrics are quantitative measures of how well a network carries data, covering speed, reliability and efficiency. They turn a vague sense that the network 'feels slow' into specific numbers you can monitor, compare against targets, and act on. Teams track them to protect the user experience, keep services available, meet service-level agreements, and catch faults early, before they turn into outages.
What affects network performance?
Before the metrics, it helps to know what moves them. The main factors are:
- Congestion. Too much traffic for the available capacity queues packets and adds delay.
- Bandwidth limits. A link can only carry so much; as traffic approaches that ceiling, performance drops.
- Device health. Overloaded routers, switches and firewalls (high CPU or memory) slow or drop traffic, and are often the root cause behind the symptoms the other metrics show.
- Distance and routing. More physical distance and more hops between source and destination add delay.
- Faulty hardware, cabling and interference. These introduce errors and lost packets.
- Configuration. Weak quality-of-service (QoS) rules, MTU (maximum transmission unit) mismatches and misconfigured devices quietly degrade performance.
Key network performance metrics
1. Bandwidth
Bandwidth is the maximum amount of data a link can carry, usually measured in megabits or gigabits per second (Mbps or Gbps). It is capacity rather than actual speed: the size of the pipe, not how much is flowing through it. As utilisation climbs toward the limit, congestion and delay follow, which is why teams try to keep average use well below full capacity.
2. Throughput
Throughput is the data actually delivered over a period, also in Mbps or Gbps. Where bandwidth is the ceiling, throughput is the real figure, and it is usually lower, because of protocol overhead, congestion, loss and retransmissions. Comparing the two tells you how much of your capacity you are really getting.
3. Latency and round-trip time (RTT)
Latency is the one-way delay for data to travel from source to destination, measured in milliseconds (ms). Round-trip time (RTT) is the there-and-back version, which is what a ping measures. Low latency matters most for real-time traffic: video calls, VoIP and online gaming all feel laggy when it climbs. High latency usually points to congestion, a long route, or an overloaded device along the path.
4. Packet loss
Packet loss is the percentage of packets that never reach their destination. It comes from congestion, faulty hardware or interference, and it hits real-time media hardest, where there is no time to resend. Even a small percentage is enough to cause stutter in a call or a video.
5. Jitter
Jitter is the variation in packet arrival times, also called packet delay variation. A steady delay a network can absorb; an uneven one it cannot, which is why jitter is so disruptive to voice and video. It is measured in milliseconds, and lower is better.
6. TCP retransmission
TCP retransmission is how often lost or corrupted packets have to be sent again. It is the metric that explains why throughput often falls short of bandwidth: every retransmission is wasted capacity and added delay. A rising retransmission rate is an early sign of congestion or a failing link, and because TCP throughput is very sensitive to loss, even a low rate can cut real throughput sharply.
7. Network availability and uptime
Availability is the percentage of time the network is up and reachable. Teams usually express it in 'nines': 99.9% uptime ('three nines') allows about 8.8 hours of downtime a year, while 99.99% allows under an hour. For anything business-critical, availability is often the headline figure.
8. Error rate
The error rate is the proportion of transmitted data that arrives corrupted, usually from faulty cabling, hardware or interference. A low, stable error rate is normal; a rising one points to a physical problem worth chasing down before it degrades everything above it.
9. Device health (CPU and memory)
The network depends on the devices running it. CPU and memory use on routers, switches and firewalls show whether the hardware itself is the bottleneck. Watching device health lets you spot resource exhaustion and rebalance or replace before a device starts dropping traffic, which makes it a root-cause metric for many of the symptoms above.
Beyond the hard numbers, quality of experience (QoE) captures how users actually rate a service. It is subjective, gathered from feedback and observed behaviour, but it is the final check that the technical metrics are adding up to a good experience.
How to measure network performance
Measuring these metrics means collecting data from across the network and comparing it against your targets. The common approaches are:
- Network performance monitoring (NPM) tools poll devices and build dashboards continuously, and alert you when a metric crosses a threshold.
- SNMP (Simple Network Management Protocol) pulls counters such as interface traffic, errors and device health straight from network hardware.
- Flow data (NetFlow, sFlow, IPFIX) shows who is talking to whom and where traffic is going, which is how you find what is eating bandwidth.
- Synthetic (active) tests such as ping, traceroute and iPerf generate traffic on purpose to measure round-trip time, loss and throughput on demand.
- Passive monitoring watches real traffic as it flows, without adding any of its own.
Active monitoring is good for controlled, repeatable checks; passive monitoring reflects what users actually experience. Most teams run both and feed one NPM platform, so the metrics sit together.
What good looks like: benchmarks and targets
Targets depend on the application, and real-time media is the strictest. As a rule of thumb for voice and video:
| Metric |
Rule-of-thumb target (real-time voice and video) |
| Latency (one-way) |
Under about 150 ms (ITU-T G.114 guidance); past roughly 300 ms it becomes hard work |
| Jitter |
Under about 30 ms |
| Packet loss |
Under 1%, and ideally well below |
| Availability |
99.9% as a baseline, 99.99% for critical services |
Treat these as starting points, not absolutes: a bulk file transfer tolerates far more delay and loss than a live call does. The value of a benchmark is that it gives each metric a line, so an alert actually means something.
Network metrics and video call quality
If you run video or voice inside your product, these metrics are not abstract. Latency, jitter and packet loss are exactly what decide whether a call feels natural or falls apart, so it's worth knowing which ones your platform shows you.
Digital Samba is a GDPR-compliant video conferencing API, and it surfaces call-quality metrics directly: MOS (mean opinion score) scoring, connection diagnostics and bandwidth management, so you can see the quality of your own calls instead of guessing. Alongside features like cloud recordings, high-quality video, real-time invitations, screen sharing, a whiteboard, AI captions and customisable branding, that gives you the experience and the data behind it. For what to watch inside an embedded call specifically, see our guide to the key metrics behind video call quality.
To add video on European infrastructure with the quality data built in, explore Digital Samba or start for free.
Frequently asked questions
What are network performance metrics?
Network performance metrics are quantitative measures of how well a network moves data, covering speed, reliability and efficiency. Common ones include bandwidth, throughput, latency, packet loss, jitter and availability.
How do you measure network performance?
By collecting data with monitoring tools and protocols: SNMP for device counters, flow data such as NetFlow for traffic patterns, and active tests like ping and iPerf for round-trip time, loss and throughput, then comparing the results against your targets.
What are the most important network performance metrics?
For most networks: bandwidth and throughput (capacity and actual data moved), latency and jitter (delay and its variation), packet loss, and availability. Real-time traffic like video and voice is most sensitive to latency, jitter and packet loss.
What affects network performance?
Congestion, limited bandwidth, overloaded devices (high CPU or memory), distance and routing, faulty hardware or cabling, interference, and misconfiguration.
