Smart factory: what the term denotes and what must already work before it means anything

What the phrase denotes once the marketing is stripped out

Industry 4.0 is the umbrella under which most of this is sold, and beneath it sits a short list of concrete capabilities: equipment that reports its own state, a network carrying that state off the machine, storage that keeps it long enough to be useful, and applications turning it into an action. A plant is connected once a stoppage, a process value or a completed part registers itself without anybody writing on a whiteboard. Everything else described under the heading is built on that base and cannot exist without it, however capable the software above turns out to be.

The prerequisites nobody includes in the quotation

What stops these programmes is unglamorous. Equipment needs unique physical identification matching whatever the system calls it. Item numbers, routings and work centres have to describe how the plant runs now, not how it ran when the data was first loaded. The network must reach the machines, be documented, and be secured, which immediately raises the question of who owns it. And the site needs one agreed definition of a stoppage, a good part and a shift boundary, because two departments counting differently will produce two irreconcilable answers and the programme gets blamed for both.

The sequencing error: buying a platform before the data exists

The common failure is procurement-led. A platform is selected, budget approved, and the implementation team then finds that many machines cannot report anything, that routings are wrong, and that nobody agrees which stoppages count. The platform is not at fault; it was asked to consume data that does not exist. Reverse the order. Establish what each target decision requires, find out what the equipment can genuinely provide, correct the master data, prove the loop on a handful of machines with real users, and only then choose the software that has to live alongside all of it for years.

Telling a serious programme from a supplier narrative

A serious proposal asks about your data before describing its own features. It names the decision that will change and the person who currently makes it. It states what happens to machines that cannot be connected. It prices the network, the integration and the sustaining effort rather than the licence alone. It identifies an owner inside operations who will still be there after the supplier leaves. Proposals leading with breadth of capability, showing a reference screen built on somebody else's data, or deferring integration to a later phase are describing a product, and the gap appears during implementation.

What it costs to keep running once it works

A connected plant is a permanent operating commitment. Licences renew. Connections break whenever a machine is replaced or a controller upgraded, and somebody has to reinstate them. Master data decays as products and routings change. Network equipment reaches end of support and needs refreshing under the same constraints as any other plant asset. Security controls require maintaining. Fund a standing engineering allocation instead of treating the build as a project with a closing date, because sites rarely lose the benefit dramatically. They lose it by letting connections lapse one machine at a time until nobody trusts the numbers.

Frequently asked questions

Is a smart factory a project or a capability?

A capability, which is precisely why it gets mishandled. Projects have an end, a budget and a closure report; capabilities need permanent ownership, a maintenance allocation and somebody accountable for whether the data still describes reality. Running it as a project produces a successful launch followed by gradual decay, as connections break and nobody is funded to restore them. Set it up with an owner and a recurring budget from the outset, and judge it on whether decisions actually changed.

We already have machines connected. Does that make us a smart factory?

It makes you connected, which is the necessary first part. The real test is whether anything acts on what arrives. If output appears on a screen nobody uses to decide anything, and the shift meeting still runs off a handwritten sheet, the connection is a cost carrying no return. Ask which decision now happens differently, who makes it, and what would break if the feed stopped. If nothing would break, nothing depends on it yet.

Where should a plant with no connected equipment begin?

Begin with one decision currently made on poor information that matters commercially, then connect only what that decision needs. A stoppage reason nobody records reliably, or a process value people argue about, both work well as a starting point. Correct the equipment identification and master data around those machines, prove the loop with the people who will use it, and expand from a working example. Broad connection programmes launched without a first user tend to stop at the trial.