1.3 System landscape & RACI

A manufacturing floor operating without clearly defined system boundaries is at significant risk of data corruption. Imagine a scenario where an ERP system tries to manage millisecond-level sensor data, or a PLC attempts to query financial ledgers. This creates a kind of “split-brain” architecture where financial records can end up contradicting physical reality. To prevent this, it’s essential to define architectural layers based on Data Granularity and Functional Domain, ensuring every piece of data has one, and only one, designated Master Owner.

System RACI

In a digital ecosystem, the RACI framework applies not to people, but to Databases.

• Accountable (A): This is the System of Record. It holds the definitive “Golden Copy.” If this database is corrupted, the single source of truth is lost. There can be only one Accountable system for any given data object.
• Responsible (R): This is the system that executes the transaction or generates the raw data in the first place.
• Consulted (C): These are systems that provide validation or constraints, such as a PLM system supplying design limits to an MES.
• Informed (I): These are systems that subscribe to data updates in a read-only capacity.

The RACI matrix

Data Object	ERP	PLM	MES	QMS	WMS
Bill of Materials (BOM)	I	A	I	C	–
Routing / Process	–	A	I	C	–
Production Order (WO)	A	–	R	–	I
Material Lots (Inventory)	A	–	R	–	R
Serial Numbers (UIDs)	I	–	A	–	–
Defects / NC	–	C	R	A	–
Rework Loops	–	C	A	I	–

The core trinity: ERP, MES, PLM

Modern manufacturing relies on three pillar systems. The key is to use each system for its intended purpose, rather than forcing one tool to perform the function of another.

ERP (enterprise resource planning)

• The Domain: Finance, Inventory, Procurement, and Order Management.
• The Core Question: “Why are we building this, and how much does it cost?”
• The Boundary: ERP’s authority typically stops at the warehouse door. It manages aggregate inventory (e.g., “we have 1,000 screws available”) but does not track which specific screw was installed into which specific product on the assembly line.

PLM (product lifecycle management)

• The Domain: Engineering Design, BOMs, CAD data, and Revision Control.
• The Core Question: “What are we building?”
• The Boundary: PLM owns the product definition. It creates the “Digital Twin” and pushes specifications and revisions down to ERP and MES. It is a system of definition, not execution.

MES (manufacturing execution system)

• The Domain: Production Control, Quality, Genealogy, and Machine Connectivity.
• The Core Question: “How is it being built right now?”
• The Boundary: MES executes the work on the shop floor. It acts as the essential bridge, translating the digital definition from PLM into physical reality and reporting that reality back up the chain.

Critical logic & boundaries

The RACI matrix provides a framework, but you need to implement the following logic to enforce the model effectively on the floor.

1. Static data (BOM & routing)

• Owner (A): PLM.
• The Risk: It’s common for Manufacturing Engineers to adjust routings directly within the MES to “make it work” on the floor. While pragmatic, this severs the revision control loop with the Engineering team, creating undocumented changes.
• The Guideline:
  • Any required process change must first be formally released through the PLM system via an Engineering Change Order (ECO).
  • When the MES detects a mismatch between its instructions and the latest PLM revision, it should enforce a controlled stop, preventing the unit from shipping until the authorized update is pushed from PLM.

2. Demand & work orders (WO)

• Owner (A): ERP.
• The Risk: Production planners might create “phantom orders” directly in the MES to keep production lines running without proper financial approval or material allocation.
• The Guideline:
  • The ERP system must be the sole originator of the Production_Order_ID.
  • The MES is Responsible (R) for updating the order’s status (e.g., from Scheduled to In_Progress to Complete).
  • The MES does not create Work Orders; it only consumes and executes them.

3. Material lots & inventory

• Owner (A): ERP (for Financial Value).
• Executor (R): WMS (for Physical Storage) / MES (for Shop Floor Consumption).
• The Logic:
  • While material is in the warehouse, the WMS controls its precise location and status.
  • Once material is issued to the shop floor, the MES assumes custody and tracks its consumption.
  • When material is consumed in an assembly or scrapped, the MES must actively trigger the deduction transaction.
  • The ERP then passively receives this inventory update. Relying solely on the BOM to “back-flush” consumption, especially for high-value components, is risky and should be avoided, as it doesn’t account for scrap or errors.

4. Traceability (serials & defects)

• Owner (A): MES.
• The Risk: Storing detailed serialization and defect data in the ERP can bloat the database and slow down critical financial closing processes.
• The Guideline:
  • The MES should generate and manage the unique serial number (UID) for each unit.
  • The QMS is Accountable for the final disposition of a defect (e.g., Use As Is or Scrap), but the MES is Responsible for physically blocking the non-conforming unit from progressing down the line.

Critical Factor

Avoid pushing every single process result (like each screw torque reading) to the ERP. Instead, push a consolidated “Genealogy Summary” (showing parent/child component links) only when the finished unit is shipped.

Data flow strategy: push vs. pull

The direction of data movement must be explicitly defined to prevent synchronization delays and inconsistencies.

Master data (push)

Static data, such as BOMs, Part Numbers, and User Profiles, requires a Push methodology. The owning system transmits updates to all subscribing systems immediately upon release.

• For example, when PLM releases a new BOM revision, it should actively push that update down to both the ERP and the MES simultaneously.

Transactional data (event-driven)

Dynamic, event-based data, like Inventory Consumption or Machine Status Updates, operates on an Event-Driven logic.

• When an operator completes a unit (triggering an event in the MES), the MES instantly updates its local Work-in-Progress (WIP) count.
• The MES then accumulates these events and transmits a bulk summary (often called a “Production Declaration”) to the ERP at defined intervals—for instance, hourly or at the end of a shift.

Shadow IT containment

Resourceful engineers often build isolated applications—like PowerApps, SQL scripts, or complex Excel macros—to solve immediate, localized problems. While useful, these tools can quickly become “Dark Data” silos, invisible to enterprise systems and compliance audits.

The Containment Rule:

• If a localized application generates data that is required for formal audits (e.g., Quality Pass/Fail records or Traceability links), it must be integrated to write that data directly into the MES, ensuring it is captured in the official system of record.
• If an application is intended purely for visualization or analysis (e.g., a custom Dashboard), it should be restricted to executing read-only queries against a central Data Lake or warehouse.

---

Recap: System Landscape & RACI

Data Object	System of Record (A)	Responsible System (R)	Critical Boundary / Action
Bill of Materials (BOM) & Routing	PLM	–	Changes require PLM ECO. MES must enforce stop on revision mismatch.
Production Order (WO)	ERP	MES	MES consumes ERP-generated WO ID; updates status only.
Material Lots (Inventory)	ERP (Financial)	WMS (Warehouse) / MES (Shop Floor)	MES triggers consumption/scrap transactions; ERP receives updates.
Serial Numbers (UIDs)	MES	–	MES generates UID. ERP receives only consolidated genealogy summary upon shipment.
Defects / Non-Conformances	QMS (Disposition)	MES (Execution)	QMS owns final disposition. MES physically blocks unit progression.