Machine retrofit data collection is the practice of capturing run/stop status, output counts and condition data from legacy equipment without replacing it. Clamp-on current sensors, part counters, or vibration and temperature sensors are fitted externally, or data is read directly from the PLC (programmable logic controller) where one exists. This approach, called retrofitting, usually leaves the control system untouched.
Key takeaways
- Getting data from an old machine does not require replacing it; a current sensor, part counter, vibration sensor or PLC read is enough.
- A current sensor shows that a machine stopped and an operator terminal shows why; the best results come from combining methods.
- According to a 2024 Siemens report, the world's 500 largest companies lose almost $1.4 trillion a year to unplanned downtime.
- A retrofit pilot should start with one to three bottleneck machines, a concrete question and success criteria written down up front.
- The gateway belongs on a dedicated production network segment from day one; the data becomes decisions through OEE, downtime Pareto charts and MES/ERP integration.
Data collected from legacy machines flows through a gateway to a dashboard, so a twenty-year-old press and last year's CNC can be monitored on the same screen. Why that gateway filters and pre-processes data is explained in edge computing in manufacturing.
Why collect data from legacy machines?
Many small and mid-sized factories run a mixed fleet. A few newer CNC machines can be networked; the rest are presses, injection moulding machines, lathes and looms from the 1990s and 2000s that either have no display at all or show information only on their own panel. These machines are still the backbone of production, and replacing them is neither economical nor necessary. Which data to read from networked machine tools is covered separately in CNC machine monitoring.
The trouble is that most of what is known about legacy machines lives in the head of an experienced operator or on a form completed at the end of the shift. Ask "how long was press 3 down today?" and the answer is usually an estimate.
Reasons for stoppages go unrecorded, short stops are never written down and counts are taken by hand at shift end. On the management screen, an unexplained gap remains between what the ERP planned and what the shop floor actually produced.
When Industry 4.0 comes up, the first reaction is often "our machines are too old for that". In fact, the first and most valuable step in digitalisation is simply knowing, accurately, when an old machine is running and when it is not. Where that step sits in the bigger picture is shown in our Industry 4.0 roadmap for SMEs.
How much does unplanned downtime cost a factory?
Unplanned downtime is one of the most expensive items in manufacturing, and it cannot be managed while it goes unmeasured.
Siemens estimates that the world's 500 largest companies lose almost $1.4 trillion a year to unplanned downtime, equivalent to 11% of their revenues. (Siemens – The True Cost of Downtime 2024)
The Siemens figure covers large corporations, but the logic holds for SMEs too: the plant that sees downtime, knows its cause and prevents it recurring wins. The same report notes that nine out of ten respondents do some form of condition monitoring and almost half now have dedicated predictive maintenance teams. Data collection is becoming the default for manufacturers that compete.
The payoff from data-driven maintenance has been measured as well. Citing independent surveys, the US Department of Energy's O&M Best Practices Guide reports industry-average results for plants that adopt a functioning predictive maintenance programme: downtime down 35–45% and maintenance costs down 25–30%. No single plant should expect those averages exactly, but none of them is reachable without data from the older machines.
There is a similar gap on energy. According to the IEA Energy Efficiency 2025 report, around two-thirds of global final energy demand growth since 2019 has come from industry, where the annual rate of energy intensity improvement has fallen below 0.5%. Energy metering per machine is the simplest data to show where improvement should start, and the same metering exposes the risk of a reactive power penalty before it reaches the bill.
What are the methods for collecting data from legacy machines?
Six core methods are used to collect data from legacy machines; each differs in what it reveals and how much intervention it needs:
| Method | What it tells you | Intervention on the machine | Best suited to |
|---|---|---|---|
| Current sensor (clamp-on CT) | Running / idle / stopped, energy use | None; clips round the cable | Machines with no electronic interface |
| Digital input (contacts, stack light signal) | Running, alarm, cycle signal | Very low; existing signal read in parallel | Machines with a stack light or cycle signal |
| Part counter (inductive / optical sensor) | Output count, cycle time | Low; sensor mounting | Lines where counts are taken by hand |
| Vibration and temperature sensors | Bearing, motor and gearbox condition | Low; surface mounting | Critical equipment with costly failures |
| Direct PLC read | Counts, recipe, alarm codes, speed | Medium; communication set-up needed | Machines with a PLC |
| Operator terminal | Downtime reason, scrap, work order | None | Every machine, for context sensors cannot see |
None of these retrofit methods changes the machine's capacity or cycle time; they only make visible what it already does. That is why retrofitting can usually be done without stopping production, in short windows such as planned maintenance or between shifts. The best results usually come from combining methods: the current sensor says the machine stopped, the operator terminal says why. Rugged industrial data terminals let operators choose a downtime reason with two taps. How the collected data connects to a wider industrial IoT architecture is explained in what is IIoT.
As the number of machines grows, data from PLCs and sensors needs to come together in one place, with alarms and trends on a shared screen. That collection and monitoring layer is the job of a SCADA system; signals captured by retrofitting can feed SCADA or a remote monitoring platform in the same way. To read PLC-equipped machines from different vendors in one common language, see our guide to the OPC UA standard.
How do you digitalise legacy machines? Six steps
- Start with a question. Not "let's measure everything", but something concrete such as "why does line 3 keep stopping?" or "why do the compressors run at night?"
- Pick pilot machines. One to three machines that are bottlenecks, expensive to lose or the source of most complaints make a good start. Write down the success criteria up front.
- Match the method to the machine. Use the methods comparison table to choose the least intrusive method for each machine. Collecting data without touching the control circuit matters for safety and for warranty and service relationships.
- Plan connectivity and network security. The gateway should sit on a network segment reserved for production devices, and wireless coverage and electrical noise should be tested on site. For remote points, LoRa vs NB-IoT vs cellular explains which technology fits.
- Turn data into meaning. A raw current reading tells nobody anything. Once running thresholds, the shift calendar and ideal cycle times are defined, the data becomes management language such as an OEE calculation and a downtime Pareto chart.
- Scale up. If the pilot hits its targets, roll the same approach out to other machines, then integrate with MES, ERP or your maintenance system; MES vs ERP clarifies which data belongs in which system.
The retrofit approach works for utilities too. Our article on compressed air leak detection shows how to add external flow, pressure and power measurement to older compressors that share no data.
For manufacturing SMEs, the software and hardware needed for a retrofit can also be financed with public support; see SME digitalisation support programmes in Turkey for the details.
How Digital Bridge runs retrofit projects
The most common obstacle in legacy machine projects is that off-the-shelf sensors and boxes do not fit the site: there is nowhere to mount them, the environment is dusty or oily, or the machine's signal is non-standard. As a company whose software and hardware come from the same team, this is where we make the difference.
- A free site assessment. Our Industry 4.0 projects begin with a walk round the site to map the machine fleet, the available signals and your priority questions. The assessment is free.
- Custom hardware when needed. Where no ready-made product fits, we design the circuit board, embedded software and mechanical enclosure ourselves through our IoT device manufacturing service. Technical feasibility studies for hardware are free.
- Pilot first. For predictive maintenance projects we pilot on one to three critical machines and show the value in metrics within the first 30–60 days. For more on the maintenance side, see our predictive maintenance service and our predictive maintenance guide.
- One dashboard for everything. Data flows into our remote monitoring platform, with dashboards, threshold alarms and historical reports in one place. Companies with several plants see all their machines on the same screen.
- Connected to production management. Downtime and output data are linked to work orders through MES production management; our MES integrates with Logo, SAP, Mikro (widely used Turkish ERP packages) and bespoke ERPs. Where process control is needed we build SCADA/HMI systems, and with a custom SCADA the source code belongs to you, with no additional licence fees (commercial vs custom SCADA).
After a needs analysis, every project gets a written proposal setting out scope, stages and cost. We do not sell packaged solutions, because no two factories have the same machine fleet.
Common mistakes in retrofit projects
Fitting the same sensor to every machine. A current sensor works perfectly on a press but can produce meaningless data on a fan that runs continuously. The method has to suit the machine.
Leaving the operator out. Sensors see a stoppage but not its cause. A system that never asks the operator why can tell you how long you stopped, but not why. An andon system where operators flag a stop with one button and pick its reason closes that gap.
Leaving network security for later. A gateway that collects machine data is a new door into the factory network. Failing to separate production devices from the office network, leaving default passwords unchanged or exposing remote access to anyone turns a data project into a security risk. Network security measures belong on day one; for the full framework, see OT security for SCADA.
Not writing down what success looks like. Pilots that start with "let's see what we find" tend to end in limbo, with no basis for deciding whether to continue. Define up front what change in downtime, response time or energy use counts as success.
Sending data to a screen nobody looks at. Unless you decide up front where the data will be used, in the morning meeting, at shift handover or in maintenance planning, the project will be forgotten within months.
Data from legacy machines pays off fastest when it is combined with work orders, showing how long each job took on which machine. For details, read our guide to production work order tracking.
Where should you start collecting data from legacy machines?
Starting to collect data from legacy machines does not require a large budget or a factory-wide programme. One bottleneck machine and a clear question are enough. Share your machine list and your biggest pain point, and our free site assessment will help decide which machine to start with and which method to use. You can reach us through our contact page.
For the steps that follow data collection, browse our complete Industry 4.0 guide.