Skip to main content
Guides & How-To

MES systems: what is manufacturing execution, and do you need one?

An MES follows and documents production on the shop floor, between the ERP and the machines. Here is what it does, when it makes sense, and what drives the price.

Author: Agent360 editorial team
Published: 8 October 2026
Last updated: 7 October 2026
Reading time: 9 min read

In short: an MES (Manufacturing Execution System) is the system that follows production on the shop floor: which orders are running, on which machines, with which materials and with what result. It sits between the ERP system, which plans orders and finances, and the controllers that make the machines run. If you need to be able to trace every batch, see stops and waste as they happen, and do away with paper slips, an MES is often the answer. If production is simple, ERP and good routines can go a long way.

If you've ended up here, it's probably because someone mentioned MES in a meeting, a supplier has suggested it, or a customer has asked whether you can document exactly what went into a particular delivery. Here you get a down-to-earth explanation of what an MES does, when it makes sense for a smaller manufacturer, and what to watch out for before you buy.

What is an MES, and what does it do?

An MES is software that follows and documents how raw materials become finished goods. The system produces what is called an "as-built" record: a record of the data, processes and results that actually happened in production, not just what was planned. This is especially important in industries such as food and medicine, where documentation of processes and events may be required.

The international standard ISA-95 (also known as IEC 62264) divides a manufacturing business into layers. At the bottom are the physical machines and the sensors. Above them are the controllers, for example PLCs (programmable logic controllers). On layer 3 sits the MES, and on layer 4 sit the business systems such as ERP. The standard is especially concerned with the interface between layers 3 and 4, in other words how production and business talk to each other.

An MES typically covers:

  • Product definitions: recipes, bills of materials, process settings and other data on how a product is to be made.
  • Resources: which machines, employees and materials are available and have the right properties.
  • Order scheduling and dispatch: work orders from the ERP are distributed to the individual workstations and adjusted when something unforeseen happens.
  • Data collection: process data, machine status, batch numbers and production logs.
  • Performance analysis: an overview of work in progress and key figures such as OEE (Overall Equipment Effectiveness).
  • Traceability: the full history of a batch, an order or a machine.

When do you need an MES?

Signs that the time has come:

  • It takes hours or days to answer which batch of raw material ended up in which customer order.
  • Production runs on paper slips that are keyed in afterwards, if they're keyed in at all.
  • The ERP shows what was planned, but nobody knows exactly what was produced, scrapped or used.
  • Stops and changeovers aren't recorded, so nobody can say where the capacity disappears.
  • Customers, auditors or authorities ask for documentation that you find hard to deliver.

Food is a good example. The Danish Veterinary and Food Administration (Fødevarestyrelsen) requires a food business to be able to trace its goods one step back (who you've bought from) and one step forward (who you've sold to), and the traceability system must be in place before you start. The rules don't require an MES. Invoices and delivery notes can be the documentation. But the authority recommends considering limiting batch sizes, because it makes a recall more targeted. The smaller and better recorded the batches are, the less has to be recalled, and that is exactly the kind of recording an MES is built for.

It may be too early if you have few item numbers, one shift and a simple flow, or if your master data in the ERP (items, bills of materials, routings) isn't yet in order. An MES builds on that data. If it's wrong, you're just digitising a mess.

The MES in relation to neighbouring systems

Layer (ISA-95) Typical systems What it sends to the MES What it gets from the MES
Layer 4: business ERP (finance, orders, inventory) Production plan and order requirements Production results, material produced and consumed
Layer 4: business PLM (product development) Product definitions, routings, work instructions, machine settings Test results from production
Layer 3: operations LIMS (laboratory), WMS (warehouse), CMMS (maintenance) Test results, material ready in stock, maintenance plan Sample requests, material requirements, machine data
Layer 2: control PLC and other control systems Process values, alarms, production results Work instructions, recipes, settings

If you're unsure what the ERP covers, read CRM or ERP: what do I need?

What does it cost?

The price depends more on your production than on the system itself. Ask about these cost drivers when you collect quotes:

  • Licence: is it charged per user, per machine, per station or per factory? And what happens to the price when you add a line?
  • Machine connection: the industry standard for getting data from the equipment is now OPC UA. Machines that can't deliver data require extra equipment or manual recording.
  • Integration with the ERP: orders have to go down, and results have to go up. The more systems, the more interfaces to build and maintain.
  • Master data: recipes, bills of materials and routings have to be cleaned up and entered.
  • Training: operators have to use the system every day. If they don't, the data turns out poor.
  • Operations: updates, support and either your own servers or a subscription.

Buy an off-the-shelf system or build your own?

For most, an off-the-shelf system is the right choice. MES functions are well described in standards such as ISA-95, and the suppliers have solved the hard parts: machine connection, traceability and reporting. At the same time, the market is moving from large on-premises installations towards modular platforms, where heavy analysis can sit in the cloud while the time-critical parts stay at the factory, and where your own engineers can build screens and workflows with low-code tools.

A custom solution can make sense if you only need a small part, for example digital work slips and stop recording on one line, or if your process is so special that an off-the-shelf system would have to be bent a long way. Whichever you choose, the integration is often the hardest part. ISA-95 has its own data format, B2MML, for connecting the MES to the business systems, and many use an integration layer (middleware) to move the messages. An integration layer is also the way to go if you want to connect ERP, MES and your quality system without replacing anything. The quality side, for example deviations, document control and audits, typically belongs in a quality management system (QMS).

Where does an AI layer fit in?

An MES collects data but doesn't act on it itself. An AI layer on top can take over some of the tasks that today require a person to read and write:

  • Summarising: producing a short shift report from stops, scrap and operator notes, so the next shift knows what happened.
  • Sorting: reading free text at stops ("the belt jumped off again") and suggesting a stop reason, so the statistics become usable.
  • Answering: finding out where a particular batch ended up by looking in the MES and ERP when a customer calls.
  • Filling in: reading batch numbers and quantities from the supplier's delivery notes and suggesting them at goods receipt.
  • Preparing: writing a first draft of a deviation report, which the quality manager corrects and approves.

What AI can't do: control the machines, replace sensors that aren't there, or invent data that was never recorded. In regulated production, release and decisions about deviations should still rest with a person. And an AI layer is never better than the master data it reads.

How to go about it

  1. Start with the question you can't answer today. Is it traceability, waste, stops or delivery capability?
  2. Map one line. What data exists, where does it come from, and what is still written on paper?
  3. Clean up master data in the ERP. Items, bills of materials and routings must be right before they're passed on.
  4. Define a pilot on one line or one product with a clear goal.
  5. Choose a system based on the integration, not just on the screens. Ask how it talks to your ERP and your machines.
  6. Expand when the pilot delivers the data you were missing.

Agent360 can help you choose a system or build the integration layer between MES, ERP and your other systems, so contact us if you'd like to talk it over.

Sources

Frequently asked questions

What does an MES system cost?

The price depends more on your production than on the system itself. The price is driven by the licence model, which can be per user, machine, station or factory, the connection to the machines, the integration with the ERP, clean-up of master data, training of the operators and ongoing operations.

Does the Danish Veterinary and Food Administration require an MES?

No. The Danish Veterinary and Food Administration requires a food business to be able to trace its goods one step back and one step forward, but the rules don't require an MES, and invoices and delivery notes can be the documentation. An MES can, however, make the batches smaller and better recorded, so that a recall becomes more targeted.

What is the difference between MES and ERP?

ERP plans orders, finances and inventory, while MES follows what actually happens in production on the shop floor. The ERP sends the production plan and order requirements down to the MES, and the MES sends production results and material produced and consumed back up. In the ISA-95 standard, the MES sits on layer 3 and the ERP on layer 4.

#mes #production #traceability #erp #integration

Which of your manual steps could disappear?

We map where data is moved by hand between your systems, and what can realistically be automated.