Fleet Management
Operating a large population of physical devices — knowing what is deployed, where, on which version, in what health, and acting on that at scale.
The discipline that separates an IoT product from an IoT prototype, and the one usually built after the first fleet-wide incident rather than before.
The questions it must answer continuously: what devices exist and which are currently reachable; what firmware and configuration each is running; which are unhealthy by battery, connectivity, error rate or sensor drift; and which have not been heard from at all — the last being the hardest, because absence is invisible to any monitoring that only processes messages that arrive.
The capabilities that follow: grouping devices by attributes so operations target a cohort rather than a list; staged operations with health gating; remote diagnostics rich enough to avoid a site visit, since truck rolls dominate the operating cost of any deployed fleet; and secure decommissioning that revokes identity.
The reporting that management actually asks for is version distribution and unreachable count, and both are usually harder to produce than expected because they require an authoritative registry of what was manufactured and shipped, which lives in a different system owned by a different function.
The failure to anticipate: devices deployed for a decade will outlive the cloud services, protocols, certificate authorities and cipher suites they were built against. Planning for credential rotation and protocol migration over the air is the difference between a fleet that ages and one that expires.