MES Without Replacing Machines: Connecting Existing CNCs and Older Equipment into One Data Flow

Almost all content about manufacturing digitalisation assumes a new factory. A real factory looks different: a 2019 CNC stands next to a 1996 press, and neither of them is going anywhere in the next five years.

The good news is that connecting those machines now costs less than the reporting being done manually because of them.

The first question is not "which MES" but "what can the machine actually say"

Before choosing any software, walk the shop floor and sort the machines into three levels.

Level 1: the machine already talks. A modern controller with an OPC UA or MTConnect interface. No hardware needed, only configuration and signal mapping.

Level 2: the machine has an output, but an old one. A PLC, Modbus, a serial port, sometimes only relay outputs. You need a gateway to translate.

Level 3: the machine has nothing. Retrofit with sensors: a current clamp on the motor, a cycle sensor, a piece counter, a stack light reader. This gives you state and takt time, not the program.

A systematic review of digital retrofitting of legacy manufacturing systems, published in Manufacturing Review in 2022, found that 62.5% of the studied implementations use exactly this IoT hardware approach (gateways and sensors), 25% embed a gateway in the existing controller, and 12.5% use packaged starter kits. Among legacy systems Modbus (39%) and Ethernet (36%) dominate, while OPC appears in 75% of the cases reviewed.

Translated: for most old machines the solution is known and repeatable. You are not inventing anything.

Be honest with yourself about what each level delivers

Level 3 gives you: running or stopped, cycle time, utilisation, micro-stops. That is the basis for OEE.

Level 2 adds: piece counts, alarms, basic process parameters.

Level 1 adds: which program, which tool, offsets, quality data.

The important conclusion is that roughly 80 percent of the value of your first report comes from level 3. Do not wait for every machine to reach level 1 before you start, because that wait lasts years and delivers nothing in the meantime.

Why retrofitting pays back faster than expected

The mistake is comparing the cost of connecting a machine with the price of a new one. The right comparison is with the person currently retyping the data.

If a shift leader spends thirty minutes per shift filling in a report, across three shifts that is an hour and a half a day. Over a year, on one line, that is around three person-weeks - and it produces data that, being manual, is rounded and reconstructed after the fact.

What an MES has to do with those signals for any of it to matter

A raw "machine running" signal has no value. Value appears when it is connected to four things.

Downtime with coded reasons. The operator picks from a list on a tablet in five seconds, and the list has at most fifteen entries. Free text cannot be summed, and a long list does not get used.

A link to the work order. Without it you have "the machine is running" when what you need is "the machine is running order 4711". The first is interesting; the second is usable.

OEE broken into its three factors. Availability times performance times quality. A combined OEE of 62 percent does not tell you what to do; 91, 78 and 87 percent does.

Scrap tied to an operation and a material batch. Otherwise you know you have scrap but not where it comes from.

A sequence that works

1. One line or cell, not the whole plant. Three to six machines, ideally the ones forming the bottleneck.
2. Measure only, for the first two weeks. State and cycle time, without changing a single process. The goal is to see how far reality differs from what everyone in the plant believes is happening. That gap is usually the biggest surprise of the entire project.
3. Add downtime entry on a tablet. The single largest jump in value per euro spent in the whole programme.
4. Connect the work order. Only now do the numbers become comparable between orders and products.
5. Then quality and traceability, if regulation or customers require it.

Three mistakes that keep repeating

Collecting every signal because you can. Two hundred tags per machine that nobody looks at, a database growing for no reason, and a team exhausted before it has delivered one usable report. Start with five signals per machine.

A screen on the wall with no owner. A chart nobody opens in a meeting stops being accurate within a month, because nobody notices when a sensor fails.

Publicly comparing shifts before the data is reliable. If in week one you publish a table saying the second shift performs worse, and the data is incomplete, you have lost the operators. After that no amount of technical quality helps, because people learn to work around the system.

If you have a mixed machine park and want to know what is possible

APPARO builds MES and ERP systems for manufacturing, including plants with a mixed park where old and new controllers have to end up in the same report. Our work on a unified ERP, MES and CRM platform for pergola manufacturing came out of exactly this problem.

We start with a walk-through and an inventory of machines against the three levels described above. That inventory shows what can be connected in a week, what needs a gateway, and what genuinely is not worth it, and the findings stay with you regardless of who you build with afterwards.

Write to office@apparo.rs or book an intro call at apparo.rs/en/start-project.