This is what “you design the network, clients decide how it works” looks like in practice. Inside an RF chamber, three virtual APs were spun up on a Candela LANforge — identical configs, identical settings, only the RF channel changed. A Pixel 7 Pro and an iPhone SE 2nd Gen walked the same simulated hallway through the same RF conditions and made completely different roaming decisions. Then we added traffic and both devices changed their behavior again. A controlled environment removes every excuse — same gear, same config, different clients, wildly different outcomes. That’s exactly why client benchmarking matters.

📡 WiFi Roaming Simulation – 3-AP Hallway

iPhone SE 2nd Gen & Google Pixel 7 Pro  ·  5 GHz Band  ·  Attenuator-Simulated Walk

▶ Play ↺ Reset Speed: 4× t = 0s

AP1 ch40 (left) AP2 ch36 (right) AP3 ch44 (left) Google Pixel 7 Pro iPhone SE 2nd Attenuator pattern

RSSI vs time with attenuation overlay & roaming event markers. Left axis: RSSI (dBm). Right axis: Attenuation (dB, inverted). Purple dashed lines = roaming events.