Industry 4.0 presentations tend to open with a dashboard. It is the wrong end of the story. The dashboard is trivial once the data underneath it is trustworthy, and impossible until then.
Start with what the machines can already tell you
Most plants are further along than they think. PLCs are logging, drives are reporting, the newer machines speak OPC UA. The problem is that the data is stranded — accessible on the machine’s own panel and nowhere else, with tag names inherited from whoever commissioned the line.
A tag dictionary is unfashionable work and it is the single highest-leverage artefact in the programme. Until Speed_2 on line three and SPD_ACT on line four are known to mean the same thing, nothing aggregates.
Agree definitions before you agree targets
OEE is the usual first metric, and the usual first argument. Planned downtime, minor stoppages under five minutes, quality losses caught at inspection versus at the customer — each of those definitions moves the number by several points.
Settle the definitions with production, quality and maintenance in the same room, write them down, and accept that the first month’s figures will look worse than the spreadsheet everyone trusted. That drop is the measurement getting honest, not the plant getting worse.
The network is a real project
Connecting the shop floor raises questions the IT team will rightly insist on answering: segmentation between OT and IT, what may initiate a connection in which direction, how a vendor supports a machine remotely without a permanent tunnel into production. These are solvable, but they are not solved by buying a gateway.
Edge or cloud
Both, usually. Anything that must act within a machine cycle stays at the edge — a cloud round trip is not a control loop. Anything comparing lines, sites or months belongs centrally, where storage is cheap and analysis is easy. The mistake is picking one and forcing the other into it.
And what about Industry 5.0
The 5.0 framing puts the operator back at the centre: automation that assists rather than replaces, with sustainability and resilience treated as design goals rather than reporting obligations. In practice, on the floor, it looks like better interfaces, cobots at ergonomically poor stations and energy data on the same dashboard as output. It is a shift in emphasis, not a new technology stack, and it is worth adopting the emphasis early.
A sequence that works
Pick one line. Build the tag dictionary. Get accurate downtime and output for a month. Publish it where the line can see it. Then, and only then, discuss analytics. Plants that follow that order tend to be running the next line within a quarter; plants that start with a platform purchase tend to still be discussing it.
