What industrial digital transformation is (and what it isn't)
Industry 4.0 has become a label for sensors, cloud, artificial intelligence and digital twins. All of that exists and has its place. But in a small or midsize manufacturer, the bottleneck is almost never a lack of sensors. It's information that starts on a clipboard, passes through a spreadsheet and reaches the manager three days later.
That's why we treat the digitization of industrial processes as a data question: who records it, where, when, and who sees it. The choice of technology comes after those answers, not before.
If the data is born wrong or late, no good-looking dashboard can fix it.
Step 1: Map what currently lives on paper and in spreadsheets
Before any system, take inventory. You don't need a consultancy: you need a few days walking the plant with a notebook and a willingness to listen to the people who do the work.
What to write down in each department:
Then rank the list by two criteria: pain (how much time or error it costs today) and risk of change (how much production depends on it to keep running). The ideal candidates have a lot of pain and little risk.
The most common mistake at this stage is mapping the process as it appears in the written procedure, not as it is actually carried out. Ask the operator. The gap between the two is usually exactly where the problem lives.
- What paper forms exist, who fills them out and what happens to them afterward (are they filed or typed in?).
- What spreadsheets feed decisions, who owns each one and how many copies circulate by email.
- What information customers, audits or regulations require that currently takes hours to put together.
- Where the same information is entered twice.
- Which processes depend on one specific person who is the only one who knows how they work.
Where to start: inspections, checklists, maintenance and traceability
These four processes have something in common: they are repetitive, have a well-defined form, generate evidence and sit at the edge of production. If the new system fails in one of them, the line keeps running. It's the best place to learn without taking risks.
Each one brings a different benefit:
In every case, the app needs to work well on a phone or tablet and tolerate a poor network. If operators have to go back to a desk to record something, they'll go back to paper.
- Inspections: equipment and safety rounds produce a searchable history and follow-up items with an owner and a deadline, instead of a folder of forms.
- Checklists: shift start, setup and line release get required fields, photos and a record of who did it. No more boxes ticked without looking.
- Maintenance: work orders, preventive maintenance and a history for each piece of equipment. Knowing that a machine had many interventions during the year changes the decision between refurbishing and replacing it.
- Traceability: linking batch, raw material, machine, operator and date. Start with one line or one product, not the entire plant.
Shop floor software: what changes in the design
Shop floor software is not an office system on a smaller screen. The operator is standing, maybe wearing gloves, short on time and sometimes without Wi-Fi. The design has to start from that reality.
The criteria we use when designing:
Test with the people who will use it, at the real workstation, before scaling. A clickable prototype reveals in one or two weeks what a month of meetings won't. That's why we offer a free app prototype before any commitment.
Buying an off-the-shelf product is also a valid option in some cases. The criteria for deciding are in custom software: when it makes sense (and when it doesn't).
- Few taps per record, large buttons and pick lists instead of free-text typing.
- Offline operation, with syncing when the network comes back.
- QR code or barcode identification of equipment, batch and workstation.
- Simple roles: the operator records, the supervisor approves, the manager views.
- Audit trail: who did what and when, with no silent edits.
Integrating with ERP and legacy systems without breaking production
Digitizing only the edge, without integrating, creates yet another data island. The opposite temptation, rewriting the ERP or the legacy system all at once, is usually worse. Almost always the answer lies somewhere in between.
The design that works: the new app captures data on the shop floor and an integration layer talks to the existing system, through an API, file exchange or, when there's no alternative, direct database reads. Start by only reading (equipment records, products, orders) and only then begin writing back.
Old systems in production call for extra care. Without documentation, nobody knows for sure what happens when you write to a table, and a forgotten trigger can take down invoicing. Our rules: a test environment with a copy of the data, starting in read-only mode, a clear way to undo the change, and change windows outside peak hours.
When the legacy system itself is the bottleneck, incremental modernization, module by module and with the system running, tends to be safer than a full rewrite. We explain the reasoning in the legacy system modernization guide and show how we work in AI-accelerated legacy system modernization. AI helps read old code, document it and generate tests faster. The decision about what to change and when remains a human one.
Metrics and ROI: how to measure without making up numbers
The most common mistake here is promising a percentage gain before having a baseline. A gain of how much, relative to what? Without a prior measurement, any number presented afterward is a guess.
Measure for two or three weeks before digitizing, even by hand. Metrics that are usually worth tracking:
Repeat the measurement after a few months of use. Part of the return is avoided cost: the audit that caught no one by surprise, the batch located in minutes. If you can't honestly convert something into money, present it in hours and occurrences. It's less flashy and far more defensible.
- Time between an occurrence (a failure, a deviation) and the manager finding out.
- Time to put together a report or respond to an audit or a customer.
- Open and overdue inspection and maintenance items.
- Rework from re-keying data and discrepancies between spreadsheets.
- Time to locate a batch when a quality problem arises.
- Unplanned stops per piece of equipment and time until it's back in operation.
Common mistakes and a simple roadmap for the first few months
What we see go wrong most often: starting with the biggest, most visible project; turning a bad form into a screen without questioning it (the paper becomes an app and the problem remains); not involving the operator; buying sensors and connectivity before having a defined process; leaving the data without an owner; and forgetting that the system needs maintenance after it goes live.
A lean roadmap, which we adjust to each plant's reality: in the first month, map and choose a single process; in the second, prototype and run a pilot on one line or department; in the third, fix what the pilot revealed and turn on the integration in read-only mode. Only then is it worth expanding. Actual timelines depend on the number of processes, the quality of the legacy system and the availability of the people who operate it.
From our experience
SKAD's founder received Embraer's Innovation Day award for the Portal DEM and the Pick-by-Light system. It's recognition that comes from inside a demanding industrial environment, and we bring that field perspective to the projects we build.
Today we have systems in operation that handle the kind of process described in this article: DEM, for equipment inspection, which has been modernized; Inspefire, for fire safety inspection; and SyncStop, for tracking industrial stoppages and projects, also modernized. In all of them, the focus is on recording field work in a reliable, searchable way and keeping the system evolving after it goes live.
Frequently asked questions
Where should digital transformation start in a small manufacturer?
With a repetitive process that involves a lot of paper and little risk to production, such as inspections, checklists or maintenance orders. Map it, prototype it and run a pilot on one line before expanding.
Do I need to replace my ERP to digitize the plant?
In most cases, no. You can capture data on the shop floor with an app and integrate it with your existing ERP, starting in read-only mode and moving forward gradually.
Does Industry 4.0 require sensors and IoT from the start?
No. Sensors make sense when the process is already clear and manual recording has become the limit. For most companies, the first gain comes from properly recording what people already do.
How can you digitize without stopping production?
By choosing processes at the edge of operations, testing in a separate environment, integrating read-only at first and keeping the old method running in parallel during the pilot. Changes to critical systems should happen outside peak hours.
How do you measure the return on digitizing industrial processes?
Measure metrics before and after, such as time to detect a failure, time to put together reports, overdue items and rework. Without a prior baseline, there's no way to claim a gain.
Conclusion
Don't start with an eighteen-month strategy. Pick one process that currently depends on paper or a spreadsheet, bring the real form and listen to the person who fills it out. With that in hand, the cheapest way to test the idea is to see the screen working before deciding anything. Ask for the free app prototype: in a few days you'll see how the process would look on the operator's phone and you can decide based on something concrete.