Troubleshooting
Method first
Use a structured approach - bottom-up (cable, PoE, switch port, AP), top-down (application, authentication, IP), or divide-and-conquer - and document what changed. Define the symptom precisely: no association, association but no DHCP, address but no throughput, or intermittent drops.
Layer 1 and RF
Check PoE class and switch budget, cable damage, AP LED state and regulatory domain. Then RF: noise floor, SNR (25 dB or better is healthy; under 20 dB is marginal), channel utilization, co-channel and adjacent-channel overlap, retries, CRC errors and data-rate distribution. High retransmission with a clean noise floor suggests coverage holes, hidden nodes or an overloaded cell; a rising noise floor suggests microwaves, cordless phones, cameras or a neighboring network. A spectrum analyzer finds non-802.11 energy; a protocol analyzer explains what the frames say.
Client and infrastructure
Sticky clients, roaming aggressiveness, driver and supplicant versions, power-save settings and band steering cause many tickets. Verify VLAN assignment, DHCP scope, gateway, captive portal, ACLs, RADIUS reachability and time skew (certificate failures). Check DFS radar events if a 5 GHz AP unexpectedly changes channel.
Traps
Treating low data rate as a signal-strength problem only; blaming the AP for a client driver issue; and expecting throughput to equal the negotiated PHY rate.
Rule of thumb
TCP throughput is typically about half the PHY rate; a 300 Mbps link yielding ~150 Mbps is normal, not a fault.