An install is done when the client can live without you
Working sensors are half the job. The other half is making the system legible to the client and to whoever services it next year — possibly you, having forgotten everything.
The handoff pack
1. The placement sheet (from your survey) — every device: name, MAC, photo, socket, zone, network. Now updated with as-built reality.
2. The one-page "what it does" — in the client's language, non-technical: what is monitored, what alerts mean, who receives them, what to do (and not do — "do not unplug the white boxes" has saved more deployments than any watchdog).
3. Access & escalation — where the dashboard lives, who has accounts, what to check first ("is the device ALIVE on the dashboard?"), and when to call you.
4. The acceptance walk — with the client watching: walk the monitored zone → alert fires → point at it. Unplug a device → dead-man notices → point at it. Plug it back → self-heal recovers → point at it. Five minutes; converts skepticism into trust.
Support that does not eat you
Set the rhythm explicitly: the fleet dashboard is checked (by you or them), monthly summaries if contracted, firmware updates batched and announced. Put your device registry to work — when they call about "the sensor in the corner", you look up the site in Latent, see the exact device, config and flash history, and answer from your desk.
The quiet upsell that is not one
Leave the door open: "If you ever need this across more rooms/sites, the same fleet scales." Nothing more. A system that visibly runs itself is the only pitch a good client needs.