MES system: what it is and when a plant really needs one

What is an MES system?

MES is the software layer between business planning and the machines: it receives work orders, assigns them to workstations, collects production data and sends the result back to the system that runs the books.

It's usually described in two ways. The first is the ISA-95 standard (the same standard is also numbered IEC 62264). It splits a plant into levels: from 0 (the physical process and machines), through sensors and PLC controllers, up to level 4, business planning and logistics, where ERP lives. MES sits at level 3, manufacturing operations management. The standard deals mainly with how level 3 talks to level 4.

The second comes from the MESA International association, which defined 11 functions that set the scope of MES: from detailed scheduling, through data collection and quality control, to product tracking and document control. It's an old model, but it still works as a checklist. Few plants need all eleven. Treat it as a menu, not a requirement.

What does an MES do in practice?

An MES turns guesswork on the floor into a record: who, what, on which machine, how many and why not more. Here are the functions used most often and the questions they answer.

The most value usually comes from recording downtime with a cause. That gives you availability, the first of the three parts of OEE (overall equipment effectiveness: availability × performance × quality), and it shows where the time really goes.

MES, ERP, APS and SCADA: how do they differ?

These systems split the work by the question they answer. ERP plans and accounts, APS builds the detailed schedule, MES runs and records execution, and SCADA supervises the machines.

The borders are blurred. Many ERPs have a production module and many MES products have a simple scheduler. Work orders usually come from ERP or APS, and MES sends quantities made and material consumed back to ERP. What matters is that each piece of information has one source. If quantities are typed into ERP and MES separately, after a month you have two versions of the truth.

How do you know a plant needs an MES?

A plant needs MES when production decisions are made on day-old data or none at all. Go through this list: three or more hits are a good reason to look into it.

• The weekly plan drifts from the output and everyone gives a different reason why.
• Shift sheets are retyped into a spreadsheet the next day, or never.
• The manager hears about a breakdown or a material shortage at the end of the shift.
• A customer or auditor wants batch history (which raw material, which machine, who inspected it) and rebuilding it takes hours.
• You don't know how long a changeover really takes, so the planner adds slack just in case.
• You cost products from standard times because nobody records actual machine time.
• The plant is growing and the same supervisor has more and more workstations to track from memory.

One hit rarely calls for an MES-class system. Sorting out one process, such as reporting downtime, is usually enough.

Off-the-shelf MES, ERP module or your own production board?

You have three routes, and each one suits a different situation. An off-the-shelf MES fits plants with many machines and strict batch-tracking demands. The ERP module is enough when production is simple and you already have the ERP. Your own production board makes sense when the process is unusual and you need a few functions, not eleven.

I go through the general build-or-buy decision in more detail in custom system or off-the-shelf tool. With MES there's one extra point: the more machines you want to connect directly to their controllers, the bigger the advantage of a ready system with ready connectors. The more data comes from people rather than machines, the easier it is to build your own board.

If you're comparing offers, send the same request to off-the-shelf MES vendors and to companies that build systems. You can put it together with the request-for-quote generator: prioritised scope, integrations, data ownership and questions for the vendor.

How do you start collecting shop-floor data in 30 days?

Start with one thing: recording downtime with a cause on the few machines that are the bottleneck. Thirty days of that data will tell you more about the system you need than any vendor demo.

1. Pick 2–4 machines that set the output of the whole line.
2. Agree a short list of stop causes, six at most: changeover, no material, breakdown, no operator, quality check, other. A long list ends with everyone choosing “other”.
3. Give operators the simplest way to report: a tablet at the machine, a phone or a QR code. It must take a few seconds, or nobody will do it.
4. Record the time automatically. The person picks the cause; the start and end times save themselves.
5. Review the data with supervisors every week. One list: the five longest stops and what was done about them.
6. After 30 days, do the sums. How much time the plant loses, for what reasons, and which of them a system will fix and which need an organisational change.

A worked example (made up, to show the method)

A plant runs 8 machines on two 8-⁠hour shifts, which is 480 minutes per machine per shift. After 30 days of recording, the average stop time is 95 minutes per machine per shift: 40 minutes of changeovers, 25 minutes waiting for material, 20 minutes of breakdowns and 10 minutes of other causes.

The conclusion from a table like this often surprises people. Almost 7 machine-hours a day are lost to missing material, and that's a problem of getting material to the workstation, not of the machines. An MES alone won't fix that: it flags the shortage sooner, but the fix is in internal logistics. Changeovers, in turn, are a matter of standard work and order sequence, that is, of scheduling. Only with this kind of sum do you know which system functions are worth paying for.

What Reefa taught me, even though it isn't a factory

Reefa is my cleaning company, not a factory, but it has the same problem as a shop floor: the work happens far from the office and the plan and the reality like to drift apart. So I built something like an MES for field teams.

Staff start and finish shifts in a Telegram bot, with the location tied to the site, and send photos of the work. If a shift hasn't started on time, the system raises an alert at 20, 40 and 60 minutes. A shift someone forgot to close is closed automatically after 14 hours, using the median length of that person's recent shifts on that site. Through the bot we logged 848 hours in 251 shifts in June 2026, 960 hours in 329 shifts in July and 1,100 hours in 364 shifts in August (counting shifts of up to 12 hours). There's more on this project in the case study a bot for the field team.

Three lessons carry straight over to production:

• Collecting data is a rollout too. In June roughly half of the planned visits went through the bot. The system worked; the habits weren't there yet. Measure coverage from day one instead of assuming the data is complete.
• Data settles arguments. When photos went missing on one site, the logs showed no errors at all. Only the data showed that in 17 of 24 shifts the last photo arrived 0–1 seconds before the “Finish” button, so closing the shift overtook the upload. Without timestamps nobody would have found it.
• Rules in the system, not in the manual. Auto-closing shifts and alerts work on their own, with nobody having to remember them. On a shop floor the same goes for reporting downtime.

When doesn't an MES pay off?

An MES doesn't pay off when the plant is small, production is repetitive and stable, and the biggest losses lie outside the shop floor. Then the money is better spent on sorting out the process or on a simpler tool.

• A few machines on one shift: the manager can see the floor without a system, and a simple downtime log is enough.
• No master data: if routings and standard times exist only in people's heads, an MES has nothing to compare output against. Write the process down first; process mapping in a company helps with that.
• Nobody will review the data: an MES whose reports nobody reads is an expensive recorder.
• The problem is elsewhere: if orders are late because of sales or purchasing, a shop-floor system won't change much.

To be honest, a system won't raise output by itself. It shows where time goes. The improvement comes from people who look at that data every week.

Where to start with MES in your plant

Start with 30 days of downtime recording on the bottleneck machines, and work out what today's gap between plan and output costs you. With that sum you'll know whether to look for an off-the-shelf MES, an ERP module or a simple production board of your own.

If you want to see which processes in your company are worth automating first, take the AI readiness assessment. If you already know you need your own system, scope and pricing are on the custom business systems page: a build with me starts from PLN 60,000 excl. VAT. And if you'd rather work through it together, book a 30-⁠minute call and bring your shop-floor data, even if it's on paper.