A network condition is the combination of bandwidth, latency, reliability, and connection stability that shapes how quickly software communicates. Mobile networks can add 50–300 ms of latency on good 4G, vary sharply in bandwidth, and drop or reconnect during requests. An API that takes 50 ms on development WiFi may take 500 ms or more over cellular, making application code feel slow even when local execution is fast. section 3.2.3 section 7.3.4
Account for network conditions by:
- Segmenting performance targets by user context instead of optimizing only for fast desktop or fiber connections. section 3.4.2
- Showing useful intermediate states while responses load, and defining behavior for routine request failures. section 7.3.4
- Reducing request volume by combining many small sequential requests when a few larger requests can carry the same data. section 7.3.4
- Testing on constrained, delayed, unreliable, and physically changing connections before release. section 7.3.4
Context-specific targets
A web application can target LCP below 1.5 seconds and INP below 100 ms on desktop, below 2.5 seconds and 200 ms on mobile with a good connection, and below 4 seconds and 300 ms on a slow connection. section 3.4.2
Profile a real mobile connection
Use Network Link Conditioner on iOS to simulate slow cellular, dropped packets, or high delay. Use Android Studio Network Profiler to inspect request timing, response size, and status codes; Charles Proxy can throttle physical devices on either platform. section 7.3.4
Table 3.6 Web performance targets by user context
|
Context
|
LCP
|
INP
|
|
Desktop
|
< 1.5s
|
< 100ms
|
|
Mobile (Good Connection)
|
< 2.5s
|
< 200ms
|
|
Mobile (Slow Connection)
|
< 4s
|
< 300ms
|
Compare web performance targets across desktop, good mobile connections, and slow mobile connections. section 3.4.2
