Six months after the MES go-live the customer had a system that was collecting data, but nobody was using it. Modules configured, ERP integration accepted and signed, tests passed. And on the shop floor the supervisors were still running their own Excel files, because "it is faster that way and you know what is going on". This is not a story about bad software. It is a story about the fact that a technical implementation and an organizational implementation are two completely different projects. The second one is rarely planned as carefully as the first.
Mistake 01
No designated data owner
The most common and hardest to fix mistake is not technical. It is organizational: there is no one in the MES or APS system who is formally responsible for the quality and currency of the data. The help desk handles outages, but nobody owns the fact that routings are out of date, operation times drift from reality, and inventory levels in ERP and MES differ by 15%.
The data owner is not the system administrator. It is a business role: usually the production manager or a process lead, who decides what is true in the system and reacts when data starts drifting from reality.
Mistake 02
System accepted on demonstration data
UAT (User Acceptance Testing) should run on real production data: work orders from the last three months, current routings, real machine parameters. In practice testing often runs on data prepared by the vendor, which is clean, complete and free of the edge cases typical for a specific plant.
Result: the system passes testing perfectly, and a month after production go-live it turns out it does not handle variants that did not exist in the demo - for example cooperation orders, products with batch quality certificates, or multi-level BOM structures.
Require the vendor to run at least one week of acceptance testing exclusively on production data from the previous quarter. Document every anomaly: that is the backlog of fixes for before Go-Live.
Mistake 03
Shadow IT: Excel next to the MES system
Production supervisors and planners, who built their spreadsheets over years, will not drop them overnight. And it is not bad will. Those Excel files often contain business logic that nobody transcribed into the MES, or they act as a "safety buffer" against system outages.
The problem starts when Excel becomes the single source of truth, and the MES collects data that nobody reads or corrects. The system reports 98% OEE, Excel shows 74%, and management does not know who to trust.
Mistake 04
ERP-MES integration treated as a post go-live item
The integration with ERP (SAP, Comarch, Epicor, Infor...) is usually the most technically challenging part of an MES implementation, and the one most often deferred. The argument: "Let us launch MES standalone first, and do the integration in a later phase." It is a trap.
Without two-way synchronization with ERP, MES runs as an isolated data island. Work orders are entered manually, confirmations do not reach finance, inventory levels are inconsistent. Every day of standalone operation builds up technical and organizational debt.
ERP-MES integration should be part of the system architecture from day one of the project, not a separate project "to do after acceptance". Lack of integration in the implementation schedule is a warning sign. It is worth asking the vendor for a concrete plan and implementation references with the same ERP.
Mistake 05
"Temporary for go-live" permissions that stay forever
During testing and pilot phases consultants and implementation administrators receive broad permissions: access to all modules, the ability to edit historical data, visibility into all work orders. Temporarily. After system acceptance no one formally revokes those permissions.
One year after go-live there are three accounts with global admin permissions in the system. Two belong to people who no longer work at the company, and one is a vendor service account that is active 24/7. This is not just a security problem, it is also a potential compliance problem (ISO, audits).
Mistake 06
No escalation procedure when the data drifts from the shop floor
The operator enters an operation time into MES: 45 minutes. Actual time: 90 minutes, because the machine was misaligned. No one corrects it, because there is no procedure for reporting discrepancies. MES accumulates wrong data, APS plans based on it, schedules become unrealistic, production managers stop trusting the system.
Discrepancy between system data and shop floor reality is a normal state - production is dynamic. The problem is not the discrepancy itself, but the lack of a mechanism to catch and correct it.
Mistake 07
Loss of a key system user
In every MES/APS implementation there are one or two people who really understand the system logic: scheduling configuration, prioritization rules, ERP mappings. When that person leaves the company 3-6 months after go-live (which is statistically very likely), the organization is left with a system that runs, but no one knows why.
Knowledge about the system should be distributed and documented, not held only in the head of one expert, even the best one in the department. The same applies to ITSM systems. Read how to convince the board to invest in ITSM, so that funding and support continue after go-live.
- Designated Data Owner with formal responsibility for data quality
- UAT run on production data (not demo)
- Shadow IT spreadsheets identified and addressed
- ERP-MES integration tested end-to-end in both directions
- Permissions reviewed and reduced after the implementation phase
- Escalation procedure for shop floor data discrepancies
- Configuration documentation in the internal wiki (not only with the vendor)
- At least 2 people trained as power users
- Knowledge succession plan for key users
- Weekly schedule deviation review for the first 90 days
Frequently asked questions
MES implementation cost depends strongly on plant size, number of machines, functional scope and number of integrations - which is why prices range widely and a reliable quote is always prepared for a specific project. Keep in mind that a significant share of IT project cost lands in post go-live fixes and interventions - the better the implementation phase is planned, the lower the total cost of the project.
A standard APS implementation in a manufacturing plant takes from 3 to 9 months. The pilot phase is typically 6-8 weeks, followed by rollout to the rest of production. Hidden problems most often surface around 6 months after go-live, when production returns to its normal rhythm after the implementation phase with its heightened oversight.
MES (Manufacturing Execution System) manages production execution in real time - it tracks work orders, machines, operators and quality. APS (Advanced Planning and Scheduling) plans and schedules production ahead of time, optimizing resource availability and delivery dates. In practice the two systems work together: APS plans, MES executes and reports. A data error in MES directly degrades APS planning quality.
Yes, ERP-MES integration is key to getting value out of the system. Without it, data on orders, materials and costs stays in silos. MES runs on different data than ERP and the differences quickly add up. Plan the integration during the architecture phase, not as a separate post go-live project. Every month of running without integration is growing data debt.
The optimal time is 3-6 months after production go-live. By then real usage patterns are visible, first data discrepancies appear, and configuration can still be fixed relatively cheaply. An audit run after 12+ months is usually more expensive - both technically and organizationally - because habits and workarounds have had time to take root.
Free post go-live MES/APS audit
We will check whether your MES or APS system really works, or just collects data that no one uses. 60 minutes, concrete conclusions, zero obligations.
Related articles
ManageEngine - ERP integration: practical guide Help desk for a manufacturing company ITSM for manufacturing: deploying a help desk in a factoryPlanning an ERP or MES implementation? See what an implementation with Rotech Group looks like →