A freight pallet sits on the warehouse floor while the handheld scanner waits for the handheld terminal to finish its handshake with the site Wi-Fi. The same unit that completed every test cycle inside the vendor lab now drops packets once the signal shares the band with twenty other devices.
Most procurement teams still sign off after a controlled demo on the supplier's isolated network. That setup never reproduces the interference, retries and power-save modes that appear the moment the hardware moves to a real floor with legacy access points.
Build-versus-buy decisions hinge on whether the product survives those conditions without custom firmware or extra access-point spend. When the contract treats the lab run as sufficient proof, the buyer absorbs every subsequent optimisation hour.
Ask for the test harness that includes measured contention, not just peak throughput. Require the vendor to run the full device fleet on the actual spectrum allocation for at least one shift before any milestone payment clears.
The gap shows up fastest in regional sites where backhaul is thinner and spectrum reuse is higher. Melbourne distribution centres have already seen three vendors this quarter quietly add external antennas after the first week of live operation.
Track the exact retry and timeout counters from the first production shift, not the demo log. Those numbers reveal whether the product was engineered for shared infrastructure or only for the cleanroom.
Insist the acceptance criteria reference real packet-loss thresholds observed on the buyer's own access points. Anything less keeps the real integration cost hidden until after the hardware is bolted down.