Once stations are in the field, the work stops being installation and becomes operations. A door that will not latch, a station offline since the small hours, a compartment holding a parcel nobody has collected for a week: each one is small on its own, and each one quietly removes capacity from the network.
State, not screenshots
The first requirement is honest live state. An operator opening the platform should see which stations are reachable, which doors are occupied, which compartments are out of service and when each station last reported in. Anything older than the last heartbeat is a guess, and guesses turn into truck rolls.
- Location and device inventory with connectivity and last-seen time.
- Door-level occupancy, so free capacity is a number rather than an estimate.
- Event history per compartment: deposited, collected, reopened, expired.
- Exceptions surfaced as a queue, not buried inside a dashboard.
Workflows start where something is wrong
Monitoring that only reports is a passive tool. The useful pattern is that every exception carries its action: an overdue parcel can be extended or flagged for return, an unreachable station can be raised as a field task, a jammed door can be taken out of allocation so no courier is sent to fill it. The operator moves through a list of decisions rather than assembling one from separate views.
Integrations are the other half
A locker network is never the whole chain. Orders originate in a commerce platform, parcels are tracked in a courier system, notifications reach recipients over channels the operator does not own. The platform therefore has to expose the events other systems care about, accept the events they produce, and keep the compartment as the single point of truth for custody.
Availability is a product feature. A compartment out of service is inventory the network cannot sell.
Measured that way, operations stops being maintenance overhead and becomes the thing that determines how much the same physical estate can carry.