The practical point. A manufacturing execution system (MES) can be modernized while the plant keeps running, but not in a single weekend cutover. The safer path is phased: learn what the old system really does, migrate one bounded area at a time, run old and new in parallel until the numbers agree, and validate in proportion to risk. In life sciences, both industry guidance and the United States Food and Drug Administration (FDA) now favour that risk-based approach [1], [2].
Why the MES is the hardest system to replace
Of all the systems in a plant, the MES is the one I would least like to replace in a hurry. In the ISA-95 model from the International Society of Automation (ISA), it sits at Level 3, between the business systems above it and the control systems below it [3]. Part 3 of the standard describes the activities at this level in four groups: production, quality, maintenance and inventory [4]. In plain terms, the MES turns a plan into production. It dispatches orders, enforces the recipe or routing, tracks material, records quality results and keeps the record of who did what, and when.
That position is exactly what makes it hard to replace. An enterprise resource planning (ERP) system can usually tolerate a few hours of delay. A control system keeps the process running on its own. The MES sits in the path of every order on the shop floor, on every shift. When it stops, operators fall back to paper, and in a regulated plant paper brings its own burden of review and reconciliation.
Over a long career spent integrating plant floor and business systems and modernizing MES platforms, the pattern I have seen most often is that the risk in an MES replacement is rarely the new software. It is the undocumented behaviour of the old system: the report a supervisor relies on at shift change, the check someone built into a screen years ago, the workaround the night shift uses and nobody wrote down. Those hidden functions are what stop a plant on go-live morning.
The cost of getting it wrong is real. Siemens estimated in 2024 that unplanned downtime costs the world’s 500 largest companies 11% of their revenues, almost 1.4 trillion US dollars a year, and that an hour of downtime in a large automotive plant now costs about 2.3 million US dollars [5]. A failed cutover is unplanned downtime by another name.
A phased path
The approach I trust has five stages. None of them is novel. The discipline lies in refusing to skip any of them.
- Inventory what the old system actually does. Interview operators, supervisors, quality staff and maintenance crews on every shift, not only the system owners. Capture every report, interface, screen and manual workaround. This inventory is the real requirements document.
- Define the information model first. Use the ISA-95 vocabulary for equipment, material and personnel so that the interfaces to ERP above and to the control systems below are explicit [3]. A clean model makes each later phase smaller.
- Migrate in bounded slices. Choose a line, an area or a product family with clear edges. Move it, stabilize it, then move the next. Never choose a slice that splits a single batch or work order across two systems.
- Run in parallel before you cut over. For a defined period, record production in both systems, or replay real transactions through the new one, and reconcile the results every day. Cut over only when counts, genealogy and quality results agree.
- Keep a tested way back. Write and rehearse the fallback procedure, including paper forms where needed, before go-live. A fallback nobody has practised is not a fallback.
Validation in regulated plants
In life sciences, an MES creates regulated records, so modernization has a second dimension. In the United States, Part 11 of Title 21 of the Code of Federal Regulations (21 CFR Part 11), the FDA rule on electronic records and electronic signatures, was published in March 1997 and took effect in August 1997 [6]. It sets the criteria under which the FDA treats electronic records and signatures as equivalent to paper records and handwritten signatures [6]. In 2003 the FDA issued guidance that narrowed its reading of Part 11’s scope. It said the agency would exercise enforcement discretion on certain requirements, including validation, audit trails, record retention and record copying, while the underlying good manufacturing practice rules, which the FDA calls predicate rules, continue to apply [7].
Industry guidance has moved in the same direction. The International Society for Pharmaceutical Engineering (ISPE) published the second edition of GAMP 5 (Good Automated Manufacturing Practice), subtitled “A Risk-Based Approach to Compliant GxP Computerized Systems”, in July 2022, replacing the first edition of 2008 [1]. GxP is shorthand for the family of good practice regulations, such as good manufacturing practice. For medical device makers, the FDA issued final guidance on computer software assurance for production and quality system software in September 2025 [2]. It recommends a risk-based approach to gaining confidence in that software and supersedes Section 6 of the agency’s 2002 guidance on the general principles of software validation [2]. In Europe, the European Commission consulted from July to October 2025 on a revised Annex 11 on computerized systems, a revised Chapter 4 on documentation and a new Annex 22 on artificial intelligence, all within the good manufacturing practice guidelines of the European Union (EU) [8]. Until a revision is adopted, the current Annex 11 remains the one that applies.
| Reference | Year | What it means for an MES project |
|---|---|---|
| 21 CFR Part 11 | 1997 | Controls for electronic records and signatures |
| FDA Part 11 guidance | 2003 | Narrower scope; predicate rules still apply |
| ISPE GAMP 5, 2nd ed. | 2022 | Validation effort scaled to risk |
| FDA software assurance | 2025 | Risk-based assurance for device makers |
| EU Annex 11 draft | 2025 | Revision under review |
For a phased migration, the practical consequence is that validation should follow the slices. Each slice gets its own risk assessment, its own test evidence and its own release, with the effort concentrated on the functions that affect product quality, patient safety and data integrity. The parallel run can then do double duty. It is a technical check, and, if it is planned and documented as such, it is also part of the validation evidence. Data migration deserves particular care. Historical batch records must remain complete, readable and attributable for their full retention period, whether they move into the new system or stay in an archived copy of the old one.
People carry the plant through the change
A new MES changes how operators work, minute by minute. Screens move, steps change order and exceptions are handled differently. The people who run the plant should test the new system with real orders before go-live, and their training should happen close to go-live and on the shift schedule, not in a classroom two months earlier. Supervisors need extra support in the first weeks, because they will absorb most of the questions. Maintenance crews and engineers need to know how the new system talks to the equipment, and what to do when an interface fails at three in the morning.
Practical advice for plant and operations leaders
- Fund the inventory properly. Discovering what the old system really does deserves real budget and time on every shift.
- Choose slices by risk, not convenience. Start where the process is well understood and the regulatory exposure is lower, then build confidence.
- Plan the parallel run as evidence. Define in advance what must reconcile, for how long, and who signs off.
- Agree the validation strategy with quality early. Write down a risk-based approach, in line with GAMP 5, before the first test script.
- Protect the history. Decide early where legacy records will live and how they will be retrieved during an inspection.
- Put operators on the test team. They will find the hidden behaviour that requirements documents miss.
An MES is the plant’s memory and its daily discipline. Replacing it is a long and careful job, and the plants that do it well treat it as an operations project supported by information technology, not the other way round.
References
- International Society for Pharmaceutical Engineering, “GAMP 5: A Risk-Based Approach to Compliant GxP Computerized Systems,” 2nd ed., ISPE, July 2022.
- U.S. Food and Drug Administration, “Computer Software Assurance for Production and Quality System Software,” final guidance, Sept. 2025.
- ANSI/ISA-95.00.01-2000, “Enterprise-Control System Integration, Part 1: Models and Terminology,” International Society of Automation, 2000.
- ANSI/ISA-95.00.03-2013, “Enterprise-Control System Integration, Part 3: Activity Models of Manufacturing Operations Management,” International Society of Automation, 2013.
- Siemens, “The True Cost of Downtime 2024,” Siemens AG, 2024.
- U.S. Food and Drug Administration, “21 CFR Part 11: Electronic Records; Electronic Signatures; Final Rule,” Federal Register, vol. 62, no. 54, pp. 13430-13466, Mar. 20, 1997.
- U.S. Food and Drug Administration, “Guidance for Industry: Part 11, Electronic Records; Electronic Signatures, Scope and Application,” Aug. 2003.
- European Commission, “Stakeholders’ consultation on EudraLex Volume 4, Good Manufacturing Practice guidelines: Chapter 4, Annex 11 and new Annex 22,” July 2025.