Industrial digitalisation · Method
Where to start when digitalising a production line
Almost every industrial digitalisation project that goes wrong goes wrong on the order, not the technology. It starts with what can be seen — sensors, screens, a dashboard — and leaves for later the thing that makes any of it useful: being able to attribute a reading to a specific batch. This is the order we recommend, and why the reverse gets expensive.
The problem is not measuring, it is attributing
In a continuous or batch process, by the time something looks wrong it has been wrong for a while. Between a parameter drifting and the defect showing up in the product there is a delay, and during that delay somebody has carried on adjusting the machine. With no record, the only way to reconstruct what happened is for somebody to remember.
That produces three questions which sound basic and which many factories cannot answer with confidence:
- Which specific material did this unit come from? Usually the answer lives in the head of whoever was on the line that day, or on a paper sheet.
- What settings was it made with? They get adjusted as you go and almost nobody writes down each adjustment.
- Why did this run scrap more than yesterday’s on the same material? Without the first two answers, this one cannot be answered.
The third is the one that costs money. The first two are what you must solve to answer it, and neither needs a sensor.
Batches first, before any sensor
The boring thing goes before the eye-catching one. Every run opens a batch, the batch records what material went in and who was there, and every unit leaves with that batch on its label. With that, and with no measurement automated yet, the first question is already answered.
A sensor reading with no batch number is a nice chart. With a batch it is an answer.
It is also the cheapest part and the hardest to sell, because there is nothing to look at. An integrator who opens by showing you screens is showing you the end of the project.
Then only what changes a decision
The rule we use for deciding what to instrument is short: measure what changes a decision. Data nobody will ever look at is not information, it is maintenance cost dressed up as modernity — a sensor to calibrate, a history to store and an integration that breaks when somebody updates something.
In practice that usually halves the initial list. What survives is captured against whichever batch was open at the time, which is what makes it answerable later.
Dashboards last
A panel built on data that can be attributed to a batch answers the only question that matters when something goes wrong: what was different about this run. The same panel over unattributed data answers nothing, though it looks just as good in a meeting.
This is the expensive mistake and it is surprisingly common, because the dashboard is what gets shown in the sales proposal and the batch identifier does not photograph well.
Closing the loop from the production line to the order
When the batch survives the warehouse and reaches the ERP, the part that actually pays for the project appears: from a customer’s order number you can walk back to the conditions of that production run. That turns a complaint into a query of minutes rather than an investigation.
It is also where it breaks most, because these are separate systems with separate owners: production, warehouse, ERP, shipping. Connecting them is the work, and the state they are in determines the scope.
What to ask whoever quotes for it
- That the batch identifier is in phase one, ahead of any sensor.
- That every data point proposed comes with the decision it will change. If there is not one, drop it.
- That the data ends up in a format you can export and read without their software.
- That they tell you what happens the day a sensor stops responding, because it will.
- That the ongoing maintenance is priced from the start rather than appearing at the end.
Connecting industrial systems starts at €9,000, and the operation afterwards from €150 a month. We write this from a company that manufactures: Eolas Prints extrudes filament in Reocín, and the systems behind that operation are ours to maintain.
Frequently asked questions
Where do you start when digitalising a production line?
With the batch identifier, not the sensors. If every run opens a batch that records what material went in and travels printed on the product label, you can already answer which material each unit came from without having automated any measurement. Sensor data hangs off that identifier; without it, it is a chart that answers nothing.
What is worth measuring on a production line?
Whatever changes a decision. Data nobody will ever look at is not information: it is a sensor to calibrate, a history to store and an integration that will break. In practice that rule usually halves the initial list of things to instrument.
Why not start with the dashboard?
Because a panel over data that cannot be attributed to a batch does not answer the question that matters when something goes wrong, which is what was different about that run. It looks just as good in a meeting and is no use for investigating anything. It is the most expensive mistake and the most common, because the panel is what gets shown in the sales proposal.
What does it cost to connect a factory’s systems?
Industrial integrations start at €9,000, and the operation afterwards from €150 a month. The scope depends on how many systems have to talk to each other and what state they are in, so the quote comes out of a discovery rather than a phone call.