Metro is the most extreme stress test for access tech. Behind 25-city scale deployment is a reusable engineering methodology.
On-device compute first
On-device 1:N matching under 300ms, no cloud round trip. At high concurrency the cloud is the bottleneck; local compute is what guarantees pass-as-you-approach without queues.
Data compliance before features
Localized face data, authorized retention and minimal collection are hard prerequisites at project start. Private deployment and tiered permission audit are built in from design, not patched later.
O&M closure
Bulk permission push, real-time access logs and alarms, remote control — these decide long-term operability of a city-scale system. Without closure, the larger the scale, the less control.
Integration openness
Connecting existing ticketing, security and emergency systems relies on open API and standard interfaces (Wiegand / RS485 / RJ45 / Relay). Closed systems go nowhere in large projects.
Insights for selection
Selection is not a spec-sheet contest; it is whether "peak stable passage + data compliance + O&M closure + integration openness" all hold at once. CoolCode’s 25-city practice serves as a baseline for such capability.