ERP and MES differences and integration

What is the difference between ERP and MES?

The ERP knows the plan, the MES knows the shop floor. The ERP (Enterprise Resource Planning) is the management system that holds the company's processes together: customer orders, stock, purchasing, invoicing, costs and often production planning. The MES (Manufacturing Execution System) is the system that runs production while it happens: it receives work orders, assigns them to machines and people, records progress, times, stops and scrap, and keeps track of lots.

The boundary between the two is described by ISA-95, the international standard also adopted as IEC 62264. It divides the company into levels: business systems sit at level 4, the systems that manage manufacturing operations at level 3, and the standard defines the information that passes between them. Below, at levels 1 and 2, are the controls of the single machines and the supervision of the plants.

The ERP is almost always there. In 2025, 48.8% of Italian SMEs with at least 10 employees used one, according to ISTAT. According to Tecnelab's guide, what is almost always missing in Italian manufacturing SMEs is level 3, the MES. That is why the question is rarely which of the two to choose: it is whether to add an MES to the ERP you already have.

The boundaries between products are not sharp, though. Many ERP systems have a production module with progress reporting, and many MES products schedule the sequence of jobs. Before buying a new system, look at what the production module you already paid for does: compare functions, not product names.

Do I need an MES if I already have an ERP?

Not always. It depends on the question nobody in the company can answer today. If the problem is that shop floor data reaches the ERP by hand and late, collecting it and connecting it is often enough; if the shop floor has to decide on what is happening now, you need an MES. In between there is a third way, which some AI assistant answers call a lightweight MES: shop floor terminals and data collection from the machines, connected to the ERP, to be extended later.

No signal decides on its own. What counts is which side most of them fall on, and how much not knowing costs you today: the hours of downtime, the scrap found late, the final costings rebuilt at the end of the month. Do that sum with your own numbers before looking at a quote; if you do not have them, that is already an answer, because the first step is measuring them.

An MES does not fix an ERP with wrong data, either. If item codes, bills of materials and routings in the ERP are not reliable, the MES will execute wrong orders more precisely: the master data gets fixed first. And whether the MES is better bought or built is the choice in the guide to custom versus off-the-shelf software. A product suits the part of the shop floor that works like everyone else's, custom software the part that sets you apart.

What data do ERP and MES exchange?

From ERP to MES go production orders, bills of materials, routings and master data; from MES to ERP come back good quantities and scrap, times, consumption, lots and stops. Each direction has its own timing.

For each piece of data we decide which system owns it and which systems read it: usually the ERP owns items, bills of materials and orders, the MES what happens on the shop floor. It is the choice that avoids two versions of the same master data, and the first step of the method we use to connect two systems.

Not everything goes in real time. A machine's status is needed within seconds by whoever is on the shop floor; material consumption can reach the ERP at the end of the operation; costs at the end of the shift. The signals machines send every second stay in the MES, and the ERP receives the summary it needs: an ERP that receives every signal slows down, and nobody in the office reads them.

How are ERP and MES connected?

There are five ways, and a factory often uses more than one.

A vendor connector is the shortest way, when there is one. Microsoft, for example, documents for its ERP Dynamics 365 Supply Chain Management the integration with MES products from other vendors. It describes events and interfaces to release orders and receive consumption and progress, and a dashboard where failed exchanges are visible and get fixed. When there is no connector, or it does not cover the data you need, you move to APIs or a middleware.

One way to avoid is writing directly into the ERP's database, if the vendor does not provide for it. The rules of the management system, such as the checks on stock or lots, live in the program and not in the tables, and a direct write skips them.

On the machine side there are two open protocols. OPC UA has a specification dedicated to numerically controlled machines, OPC 40502, and the OPC Foundation also publishes an OPC UA model for the information described by ISA-95, OPC 10030. MTConnect is an open, royalty-free standard for machine tool data. Older machines without an interface are connected with a gateway or with sensors, or the operator reports progress from a terminal. We connect to what each system and each machine already exposes, and for a proprietary protocol we write an adapter, without touching the hardware in the field.

Why does an ERP and MES integration fail?

Almost never because of the technology. The problems that show up after go-live are usually these:

• the master data does not match: item codes, units of measure or bill of materials revisions differ between the two systems, and a production report does not find its item;
• the same event arrives twice, and the ERP deducts stock twice;
• a production report rejected by the ERP is silently lost: the MES closed the operation, the ERP does not know, and the stock count finds out;
• two systems change the same data, and whoever wrote last wins;
• nobody compares the quantities produced in the MES with those recorded in the ERP, and the gap grows every day;
• the network on the shop floor goes down, and the data from that period never arrives.

They are prevented at design time, not afterwards. We decide with you what happens when a system does not respond, so that no data gets lost silently. Messages are queued and realigned when the network comes back, as we do with the unstable networks of workshops, and every rejected message ends up in a list someone looks at.

Before signing with an MES vendor, ask them to show you a full cycle with your data. It should start from an order released by the ERP, machined on one of your machines with a stop and a scrapped part, and end with the production report going back into the ERP and the stock adding up. A demo with sample data says little about how it will behave with yours.

How much does an MES cost, and connecting it to the ERP?

According to Tecnelab's guide of 1 June 2026, an MES for an SME with 30-80 employees costs roughly between 30,000 and 120,000 €, including configuration and training. According to the same guide, integration with the ERP, if the two systems have standard interfaces, adds 20-30% to the cost of the MES. Data collection alone on a department of 10-15 machines, with gateways, sensors and a platform to view the data, costs between 15,000 and 40,000 €, again according to Tecnelab. Public figures are estimates by whoever publishes them, not price lists. An off-the-shelf MES also carries a licence, which depending on the product is paid once, with yearly maintenance, or as a fee per machine, per user or per station.

With us the thresholds of every custom project apply, the same as in the guide to custom software costs. A first module, meaning a single function in production, starts at 10,000 € with 6-8 weeks of work. A platform with several departments and roles and integrations with ERP and CRM starts at 45,000 €. Which threshold applies to your connection is set by the analysis of the systems you have, before the quote, and the top of the range is written into the contract.

The return is calculated on today's losses, not on a vendor's percentages: the hours of downtime, the scrap, the hours spent retyping and reconciling data, each at its own cost. If the integration or the MES does not remove enough of them to pay for itself, it is not worth it, and it is better to know before.

An example: 40 employees and 15 CNC machines

A contract machining company has 40 employees and 15 numerically controlled machines. The ERP handles orders and stock; progress on the shop floor is recorded on paper. On 23 October 2026 we asked Google, ChatGPT and Perplexity whether it needs an MES or whether connecting the machines to the ERP is enough, and they do not agree.

Google's AI Overview and AI Mode answer that an MES is needed: connecting the machines brings raw data to the ERP, without the setup or the cause of the stop. ChatGPT and Perplexity answer not to start from a full MES: first shop floor data collection connected to the ERP, tested on 2-3 machines according to ChatGPT, on 3-5 according to Perplexity.

All three agree on one point: connecting the machines alone is not enough. A machine knows how many parts it made and when it stopped, but not for which order or why. That data is added by the operator, from a terminal next to the machine. From there, the way that pays off is this:

1. the map of the 15 machines: make, numerical control, year, what they expose, OPC UA, MTConnect or nothing;
2. progress reported from the terminal instead of on paper: start and end of each operation, good parts, scrap, cause of stops;
3. a trial on 2-3 machines with different controls, for a few weeks, on real jobs;
4. production reports reaching the ERP without being retyped, with quantities reconciled every day;
5. the decision with the data in hand: if rush jobs, setups and job sequencing are still what holds the shop floor back, the MES; if the problem was the paper, it is already solved.

For a department of this size, Tecnelab's estimate for data collection alone is between 15,000 and 40,000 €. If data collection connected to the ERP is enough, we tell you before the quote; if an MES is needed, you choose it among off-the-shelf products and we connect it to the systems you have.

When is an MES not needed?

When production is simple: few machines, mostly manual work, jobs with short routings, and the ERP's production module is enough. When nobody will look at the data: according to Tecnelab, buying the technology before mapping the processes is the most common mistake, and an MES installed without clear flows becomes data collection nobody uses. When the problem lies upstream, in the ERP's master data and bills of materials. And when the ERP itself is old and needs replacing: then the question is a different one, and you find it in the guide to legacy software.

In all these cases the first job is smaller than an MES: connecting what is there properly, starting from the part that is handled worst.

Sources and further reading

• ISA, “ISA-95 Standard, Enterprise-Control System Integration”isa.org, accessed 23 October 2026
• ISTAT, “Imprese e ICT, anno 2025” (Enterprises and ICT, 2025), 15 December 2025istat.it, in Italian, accessed 23 October 2026
• Tecnelab, “Manifattura 4.0, digitalizzazione della fabbrica, la guida operativa” (Manufacturing 4.0, digitalizing the factory, the operational guide), 1 June 2026tecnelab.it, in Italian, accessed 23 October 2026
• Microsoft, “Integrate with third-party manufacturing execution systems”, Dynamics 365 Supply Chain Managementlearn.microsoft.com, accessed 23 October 2026
• OPC Foundation, “OPC 10030, OPC Unified Architecture, Common Object Model, ISA-95”, 6 November 2013reference.opcfoundation.org, accessed 23 October 2026
• OPC Foundation, “OPC 40502, OPC UA for Computerized Numerical Control (CNC) Systems”reference.opcfoundation.org, accessed 23 October 2026
• MTConnect, the open standard for machine tool datamtconnect.org, accessed 23 October 2026