An ugly old platform does not have to be ripped out. An old platform that “still opens” is not always worth patching. Weigh three things: whether point assets still exist, whether data can leave into your own system, and whether an exception opens a ticket. You can reskin. You cannot swap a closed cloud.
Lean extend
- Protocol is open, or a gateway already emits frames
- The point ledger is clear; historic curves can export
- The main pain is an ugly board and alerts that never dispatch
- The vendor is still there; you can add an interface layer
Then adding tickets, rights and a business hookup on the collect layer is lower risk, and the floor need not stop to swap hardware.
Weigh the loop
Lean rebuild
- The vendor went dark; cloud accounts are not the company’s
- The box is locked; tickets and MES get no events
- Point names are a mess and will not match the line
- Keep patching costs more than a rebuild
Rebuild is not ripping every sensor. Move readable points to the new collect; run the old wall in parallel until you can drop it.
A closed box that will not release data has little room to extend: When a closed IoT box will not release your data and Closed box vs. open access. If the vendor already went dark, inventory devices before you decide the platform’s fate: The old vendor went dark. Can the devices still join a new system. Cost and cycle depend on whether you stop at extend or rebuild: What hardware-software projects cost, and how long they take.
On rebuild, keep the points. Drop a cloud you cannot see into
The rebuild fear is “new platform live, old points blind.” List points already in safety ledgers, maintenance contracts and energy reports; keep what you can. Migrate definitions, thresholds and owners—not only the wall skin. Archive curves you can export; admit a gap if you cannot. Do not pretend ten years of data still exist. Do not drop paper patrols in the parallel window; cut after false alerts are acceptable: From brief to launch: how a hardware-software project runs.
In parallel, keep the old wall for visits and safety checks until the new loop is stable. Changing only the wall integrator, leaving collect alone, is the dearest fake extend—next year you still have no tickets. Ask whether events can leave an API before you talk skin.
When DaXi Technology assesses an old platform we ask for accounts (if any), the point list, and one real alert’s handling path—then extend or rebuild, not a default rip. Send the live URL and the exception that hurts most to Hardware-Software Integration.
An extend contract should name the point list and an event API
Write: old point list, event API, how long the parallel window lasts. “Upgrade the platform” defaults to a reskin. On rebuild, move accounts, filings, certs and SIM ownership to the company so the new platform is not bound to one phone. If the old platform is only a monitor wall, the decision is closer to “do we want a work system”: Treating the system as a monitor wall vs. a work entry.