MES ERP Integration: Connecting ERP and MES on the Plant Floor
The short answer
What is MES ERP integration?
MES ERP integration keeps the ERP plan and the plant floor in agreement. Many manufacturers do not need a full MES. They need a narrow connection to the ERP they already own, built so production keeps running.
MES ERP integration is the work of getting what is actually happening on the line, which job is on which machine, what the operator recorded, and why the line stopped, into the system that plans and accounts for the work, and of sending a trustworthy schedule and order data back the other way, without taking the plant down. An ERP plans and accounts for work: what was ordered, what it should cost, and what was scheduled. A manufacturing execution system deals with what is happening right now on the floor. Many manufacturers do not need a full MES. They need a narrow piece of that visibility connected to the ERP they already own, which is far cheaper and lands far faster than replacing either system. Active Logic is a 100% U.S.-based custom software company. Our manufacturing software development includes ERP and MES integration and the modules an ERP does not cover, delivered by senior U.S. engineers under a Director of Engineering.
Replacing an ERP that already runs the business is unreasonable in most cases. The usual work is software that links the systems a company already owns, and adds the piece those systems were never going to cover. Software should adapt to the business. In some cases a replacement might be the way to go. It is the wrong default.
What is MES ERP integration?
MES ERP integration is the interface that keeps the plan and the floor in agreement. The ERP holds the order, the expected cost, and the schedule. The MES, or whatever is standing in for one, holds the job on the machine, what the operator recorded, and why the line stopped. When those pictures diverge, the plant runs on the whiteboard, the traveler, or the one person who knows why the job is late, and finance runs on a report that was already old when it printed.
Business planning sits above manufacturing operations. The hard work is the interface between them.
Why do the ERP and the plant floor drift apart?
They drift because the floor already has a system, and it is rarely the one the office bought. A spreadsheet, a whiteboard, and a paper traveler are an unofficial MES. Someone retypes a completion into the ERP at the end of the shift. A batch update looks fine before the first changeover and is wrong by the second. The machine speaks a different protocol than the office system, so the count that matters never arrives on its own.
McKinsey, writing in June 2022 about why IT and operations technology fail to scale together, names the same obstacles: manual translation between the enterprise layer and the operations layer, often through a spreadsheet; data that is never collected or sits fragmented in silos; and no integrated execution system, or several systems with no connection between them. The CIO usually owns enterprise IT. The COO, the plant leader, or a business-unit manager usually owns the shop floor, the plant systems, and maintenance. Two owners, two pictures, and nobody whose job is the seam.
A disconnected ERP and a disconnected floor are an operational liability. The report says the job is on time. The board by the line says it is not. Until the plant decides which picture it will trust, every project stacked on those systems inherits the argument.
McKinsey’s April 2022 Industry 4.0 report says a large majority of manufacturers remain unable to scale a pilot or show a satisfactory return. Where a program does scale, McKinsey reports it is not uncommon to see 30 to 50 percent less machine downtime, 10 to 30 percent more throughput, 15 to 30 percent better labor productivity, and forecasting on the order of 85 percent more accurate. Those figures describe scaled programs, not one interface, and they are not a result we are promising. Use them as the reason a connection has to serve a result the plant and the finance team both recognize. The same report tells manufacturers to put IT and operations-technology architecture on a three to five year road map. The interface belongs on that map. It is not the whole map.
What does ERP and MES integration actually involve?
Manufacturing ERP integration starts with a map of how a work order, a material move, a downtime event, and a completion actually travel today, including the steps that never make it into a process diagram. ERP MES integration is that same map, plus a decision about which objects have to be kept in agreement: orders, inventory, genealogy, labor, and quality. From there the work is practical. Build or adapt the interfaces. Handle older systems with no modern API. Test against a live plant. Cut over in phases so production keeps running.
Most of the manufacturing work we take on involves the ERP the company already runs. That includes SAP, Epicor, Infor, Microsoft Dynamics, and the engineer-to-order platforms mid-market plants actually use, including older systems with no modern API. The shape is consistent: get shop-floor reality into the system of record without disrupting production, build the custom ERP modules the ERP was never going to cover, and, in the other direction, surface ERP records inside the tools people already use so nobody is retyping between screens.
The physical end of the same problem is the machine. When the count lives on a controller, the software has to speak Modbus or OPC UA. A protocol mismatch is another way the floor and the system of record stop sharing a number.
There is no published percentage split for where the money goes on this specific kind of connection. Complexity moves the number more than the label on the system. The hours hide in the gap between what the process actually does and what either system assumed, in testing, and in third-party interfaces that do not behave the way the documentation claims. The line buyers overlook most often is QA, the fixes after the first release, and maintenance. A focused tool that replaces one spreadsheet or automates one document process typically ships in three to five months. A platform that spans the shop floor, the ERP, and the customer typically runs six to twelve months, in phases, so people get working software before the last interface is done. We do not publish a rate or a project price for this. The figure depends on decisions we have not made together yet.
Do you need a full MES, or a narrower connection to the ERP you already have?
Often you need the narrower connection. Many mid-market plants do not need a multi-year MES program. They need a live job board, digital travelers, downtime capture, or a single inventory count, tied to the ERP they already run. Those are pieces a full MES product would have covered, built only where the current process is blind. None of them require throwing out the ERP.
If the ERP does the core job, connect the missing floor piece instead of replacing the system. How to choose among a refactor, a rebuild, and a replacement is covered in legacy system modernization. Heavily customizing SAP, Dynamics, or a similar system can cost more than building the missing piece. When the way the plant builds has real nuance, that piece is usually better as custom software the company owns. See build vs buy software. Leaving the whiteboard and the traveler in place for a while can also be the right call, when the business is not ready to fund the connection and the workaround is still the cheaper option. Accept the workaround until the business can support the investment. Do not fund a bad project just to say the floor is “live.”
The question to ask before you buy an MES, or before you hire anyone to connect the floor to the ERP, is the one the first conversation should settle: do you need a full MES, or a narrow piece of visibility on the ERP you already own? People who have lived in the same software for years tend to ask only for what they believe that software can already do. The missing piece never gets named, so the project becomes a replacement.
Two public examples are about the ERP and the work in front of people having to agree.
Horizon Systems designs and builds engineer-to-order equipment. We built a custom CRM tied to the ERP so an account resolves to the same customer and vendor records that run the job, and job history is visible from the account. A team of six senior engineers took it to production in about seven months. If the account and the job are not the same record, nobody trusts either system.
PTMW is the document side of the same problem. A bill of materials goes in and a finished, ordered submittal packet comes out, and the exceptions show up while there is still time to fix them, rather than at submittal. Five senior engineers, an eleven-month build, in production. The official record and the packet in front of the engineer had to agree.
The same pressure shows up past the plant gate. BG Products reaches the market through distributors and independent shops, and the software had to live inside Tekmetric, the shop system those shops already run, rather than asking them to open something else. Manufacturing software sometimes has to meet the next person in the system they already have open. That is channel work, not a plant-floor execution system, and it belongs in the same conversation only because the record still has to match.
What should a buyer expect a partner to use as a reference for the interface?
Ask for a reference, not a stack you are required to buy. ISA-95 (ANSI/ISA-95, also published as IEC 62264) is the international standard for integrating business systems with manufacturing control systems. Level 4 is business planning and logistics, where an ERP sits. Level 3 is manufacturing operations, where an MES sits. The standard deals primarily with the interface between those two levels, and one stated goal of that work is to reduce the risk, cost, and errors of implementing the interface.
NISTIR 8107 calls ISA-95 a commonly used reference model for automated interfaces between enterprise and control systems, for batch, discrete, and continuous processes. It describes B2MML, published by MESA, as an implementation of IEC 62264 used to link ERP and supply-chain systems with an MES. A companion paper in the Journal of Research of NIST adds that OAGIS is widely used in discrete manufacturing for ERP and MES integrations, that MTConnect and OPC UA collect machine-tool data, and that the ISA-95 parts focused on ERP and MES have existed since 2000 with abundant vendor support. The pyramid diagram is a common implementation, not a requirement. Dennis Brandl and Charlotta Johnsson, writing in ISA InTech in October 2021, point out that the same activities can sit in a networked architecture, including on edge devices, without abandoning the standard. You do not have to purchase five layers of software to take the interface seriously.
Where does AI help, once the data is actually flowing?
After the numbers are connected and the plant trusts them. A plain-language question of the ERP (“What shipped late last month, and why?”), a scrap or defect trend, or a match between an invoice and a receipt is only as good as the record underneath it. If the floor and the ERP disagree, a model will answer with confidence from the wrong picture. That is the hard part handed to a machine, which is the argument in The Part AI Does Was Never the Hard Part.
Deloitte’s 2025 Smart Manufacturing and Operations Survey asked 600 executives at manufacturers with U.S. headquarters or operations, $500 million or more in revenue, and more than 1,000 employees. The survey was fielded in August and September 2024. At facility or network level, 29% were using AI or machine learning, and 24% were using generative AI at that scale. Those companies are larger than the mid-market plants we usually work with, and even in that group facility-level AI is a minority. The useful move is not to bolt AI onto a plant that still cannot say which job is on which machine. It is to put AI on data someone on the floor will stand behind. Asking the ERP a question in plain English is a fair illustration of what becomes possible once that is true. It is not a system we are claiming to have shipped as a plant product.
How do you connect these systems without taking the plant down?
In phases. The new path runs alongside the old one where you need a comparison, and a flip-the-switch cutover is not the default. The aim is to get shop-floor reality into the system of record without disrupting production, with working software in people’s hands before the full scope is done. New components run beside the system that is still shipping product. That is the same posture behind our application modernization services: operations continue while the new path is proven.
The pattern I want from a cutover is the one on the PTMW work. Exceptions surface while there is still time to fix them, not after the packet, the shipment, or the close. A Director of Engineering owns the engagement from architecture through production. The engineers are senior, full-time, and 100% U.S.-based, with no offshore or nearshore teams. Some of this work legally requires U.S. engineers, including ITAR programs, many government contracts, and security reviews that ask where the developers sit. There is no offshore handoff in the middle of a live plant.
What usually goes wrong?
The expensive mistakes are decisions made before anyone designs an interface.
- Buying a full MES when a narrow visibility layer would do the job.
- Replacing a working ERP because the floor is hard to see from it.
- Point-to-point interfaces that nobody owns after the project team leaves.
- Integrating from a conference-room diagram without walking the path a work order actually takes.
- Pointing AI at reports the plant does not trust.
- Treating the connection as an IT project the operations team hears about at the end.
Many manufacturers have no development staff of their own, and they should not have to. That is a reason to be precise about the missing piece. It is not a reason to hand the plant’s process to a product that covers someone else’s.
In that same Deloitte survey of large manufacturers, 51% said these initiatives are owned and driven by operations leaders (a COO or a director of operations), and 38% said technology owners such as a CTO are leading. 65% ranked operational risk as the first or second concern. The ownership belongs with the people who feel a late job. Projects that succeed are the ones operations decides, not the ones operations is told about after the interface is already built.
When the ERP and the floor last disagreed, which picture did you trust? If the honest answer is the board, because the system is behind, you already know the scope. Connect the systems you run. Add the piece that is missing. Schedule a call with a knowledgeable team member and get your project scoped.
Frequently asked questions
What is MES ERP integration?
MES ERP integration connects shop-floor reality (which job is on which machine, what the operator recorded, labor, materials, quality, and why the line stopped) to the ERP that plans and accounts for the work, and sends schedule and order data back, usually through a defined interface rather than retyping. An ERP plans and accounts for work: what was ordered, what it should cost, and what was scheduled. A manufacturing execution system deals with what is happening right now on the floor. The integration is the interface that keeps those two pictures in agreement.
What is manufacturing ERP integration?
Manufacturing ERP integration is the same work: connecting the plant floor to the system of record the company already runs, and sending schedule and order data back. An ERP plans and accounts for the work. A manufacturing execution system deals with what is happening on the floor right now. Many plants need a narrow piece of that visibility connected to the ERP they already own, not a second system with a new name.
Do I need a full MES if I already have an ERP?
Often no. Many manufacturers need a narrow piece of floor visibility connected to the ERP they already own, such as a live job board, digital travelers, downtime capture, or one inventory count. A packaged system is a fit for a generic need. When how the plant builds has real nuance, that piece is usually better as custom software the company owns. See build vs buy software.
Can you integrate with SAP, Epicor, Infor, or Microsoft Dynamics?
Yes. Most Active Logic manufacturing engagements involve the ERP already in place, including SAP, Epicor, Infor, Microsoft Dynamics, and the engineer-to-order platforms mid-market manufacturers run. That includes older systems with no modern API. The usual shape is to get shop-floor reality into the system of record without disrupting production, and to build the pieces the ERP was never going to cover.
What standard is commonly used as a reference for ERP and MES integration?
ISA-95, also published as IEC 62264, is the widely used standard for integrating business systems with manufacturing control systems. NIST describes ISA-95 as the enterprise-to-control-system integration standard in the MES area, notes that OAGIS is widely used in discrete manufacturing for these integrations, and describes MESA's B2MML as an XML implementation of IEC 62264 used to link ERP and supply-chain systems with an MES.
How long does a plant-floor-to-ERP integration take?
A focused tool often ships in three to five months. A broader platform spanning the shop floor, the ERP, and the customer often runs six to twelve months in phases, so working software reaches people early. Active Logic does not publish a rate or a project price for this work. The number depends on how many systems have to connect and how specific the process is.
How do you do this without disrupting production?
Phased delivery, with the new path running alongside the old one where a comparison is needed, and no flip-the-switch cutover as the default. A Director of Engineering owns the engagement. The engineers are senior and 100% U.S.-based. Exceptions should surface while there is still time to fix them.
Where does AI fit in ERP and MES integration?
After the data is connected and the plant trusts it. Plain-language questions of the ERP, scrap or defect trends, and reading a document or an invoice are useful only then. Pointing AI at reports the plant does not trust hands the hard part to a machine. That argument is developed in The Part AI Does Was Never the Hard Part.
Should we replace our ERP to get plant-floor visibility?
Usually no. If the ERP covers most of what the business needs, keep it and connect or add the missing piece. Replacing a working ERP is unreasonable in most cases, though in some cases it might be the way to go. When the process has real nuance, the specific piece is usually better as custom software the company owns.

