Measure the symptom
Bandwidth is capacity; throughput is achieved transfer rate. Latency is delay. Jitter is variation in delay, and packet loss is the fraction of packets that do not arrive.
- Use a wired baseline to separate Wi-Fi issues from upstream issues.
- Measure both directions and record the time, path and test conditions.
- Compare against application requirements, not a universal “good” number.
- A speed test measures a particular path at a particular time.
Find the bottleneck
A fast link cannot remove a slow server, congestion or a high round-trip time. TCP throughput can depend on window sizes, loss and latency as well as raw capacity.
- Check interface errors and negotiated link speed.
- Compare single-stream and parallel-stream tests.
- Check MTU and path MTU discovery when small requests work but large transfers stall.
- Queue management can reduce delay under load; QoS does not create capacity.
Keep a repeatable baseline
Use the same endpoints and settings before and after a change. Keep raw results and a short explanation of what changed.
- iperf3 measures between endpoints you control.
- Ping samples ICMP reachability and round-trip time, not application health.
- A packet capture can show retransmissions and handshake delays.
- Document percentiles and variation, not just a single average.