1. Site inventory
Devices you have, protocols, network, and who handles an exception. Whether the vendor is still reachable: The old vendor went dark.
Treat a hardware-software project as “buy devices plus a wall” and it dies in color tweaks and extra camera feeds. Run it as engineering: each stage has a sign-off—point ledger, exception loop, live weak-network logs, business reconcile, then drop paper after false alerts are acceptable. Cost and cycle depend on which phase you stop at: What hardware-software projects cost, and how long they take.
2. Closed-loop design
Write alerts, tickets, rights and privacy into the plan before you buy or change hardware. Protocol: If the device protocol is closed, can we still integrate.
3. Access and integration
Prove live devices, weak network and power-loss catch-up—not only a lab green light. You can name the hardware brand: Must we supply the hardware, or can you name the brand.
4. Business hookup
Which events enter HR, warehouse and MES, and how failures reconcile. With vs. without: what exception handling looks like.
5. Pilot cutover
Drop parallel paper after the false-alert rate is acceptable. Diagnosing noisy alerts: The system never alerts, or alerts too much.
Who can sign off decides whether the project idles on a video wall
The usual stall is not soldering. Nobody inside can sign “this exception should open a ticket.” Facilities want more points, production fears a false stop, HR fears faces, security only wants a wall. At kickoff name who owns the loop, who changes thresholds, who reviews privacy—two rounds max. More than two usually means the goal is unset, not that delivery is sloppy. Rebuild vs. extend should be decided after inventory: Rebuild the old IoT platform, or extend it.
Hardware lead times often beat software. If gates, meters and explosion-proof probes are ordered after loop design, the site waits. Mark long-lead items at inventory and what existing points can cover. Freeze point names, location codes and employee IDs then; rename after integration and tickets will not match MES. Acceptance is not “pretty.” Walk at least four paths: events queue after unplug; over-threshold creates a ticket number; exit kills access; the crew can close a ticket on a phone. Test notify channels on go-live so “online” is not a message into an empty WeChat. Same bar as fewer screens to watch: Fewer screens to watch: how alerts become assignable tickets.
DaXi Technology delivers hardware-software in that order and can plan it with business systems, HR and mobile. Go to the Hardware-software service page and list devices on site and the exception you want closed, or call +86 139-2520-6166.
Write “what we will not do” at kickoff
At kickoff write what you will not do: no plant-wide 3D in phase one, no ten parks, no swap of every old sensor. Shrink scope so integration can finish. Change control in writing, or a verbal extra gate restarts construction.
After the pilot, visit once: whether crews actually claim tickets beats wall-satisfaction scores. Vendor authorization, a downtime window and a night contact belong on the integration list. Miss those three and a lab green light becomes waiting on site.