Key takeaways
- No single system runs a pharmaceutical company. A plant is a set of systems, and the important question is not how many there are but which one owns which data.
- Ask six different questions about any piece of data: where is the official record, who takes the decision, who handles execution, what does the physical control, where is the detailed history, and what does the analysis.
- One batch appears in many systems, and each sees it through its own responsibility. That is normal. Problems start when two systems both try to be the master.
- The function that decides is not always the system that records. The quality unit decides whether a batch is released; that decision is recorded as a status that ERP and WMS must respect.
- Sharing a platform is not sharing ownership. Systems can run on the same infrastructure and still need clear boundaries for who creates, changes and approves a record.
- Do not push everything into ERP. High-frequency process data belongs near the equipment, execution evidence in the manufacturing system, and business facts in ERP.
On this page
- 1. Two ideas: boundary and system of record
- 2. Six kinds of responsibility
- 3. The layers of a pharmaceutical plant
- 4. One batch, seen by six systems
- 5. Pairs that are easy to confuse
- 6. Who typically owns what
- 7. Keep data in its own layer
- 8. How the systems talk to each other
- 9. When to split a system out of ERP
- 10. The regulated-industry angle
- 11. Build your own ownership matrix
- FAQ
- Sources
Pharmaceutical manufacturers rarely run on one piece of software. Between a purchase order and a released batch there are typically an ERP, a warehouse system, a manufacturing system, a laboratory system, a quality system and the control layer on the equipment, plus document, training and maintenance systems around them.
Buyers, QA teams and IT leads often argue about which system should do what. The argument gets easier when you stop asking "does our ERP do this?" and start asking "which system is allowed to be the official source for this?". This guide gives you the vocabulary and a map to answer that question. It is general and vendor-neutral; every company draws the lines a little differently.
1. Two ideas: boundary and system of record
Boundary is the line that says what a system is responsible for, what data it is allowed to create or change, and what belongs to another system.
System of record (SoR) is the system designated as the official source for a type of data. Other systems may read, reference or cache that data, but the right to create, change and approve the official record sits in one place.
An example from a tablet plant: the technical life of a tablet press is owned by the asset-management system. The manufacturing system refers to the press when it runs a batch, the quality system refers to it in a deviation, the historian attaches its sensor data to it, and ERP refers to it for costs. If the press is renamed, the change starts with the owner and is then distributed. It should not be edited separately in five systems until five names exist.
Two clarifications prevent common misunderstandings:
- A domain boundary is not a server or a database boundary. Several systems can share a platform, a codebase or even a database cluster and still keep logically separate ownership.
- Shared platform is not shared ownership. Using the same identity, workflow or audit services does not entitle the ERP to edit a laboratory result, or the quality system to edit a production record.
2. Six kinds of responsibility
"Which system manages this?" is too vague. Split it into six questions:
| Responsibility | The question | Example |
|---|---|---|
| Record | Where is the official record? | Laboratory results sit in the laboratory system |
| Decision | Who makes the decision? | Batch disposition is a decision of the quality unit (in the EU, certified by a Qualified Person), based on evidence from MES, LIMS and QMS |
| Execution | Which system coordinates the work? | Batch execution runs in the manufacturing system |
| Control | What directly controls the physical process? | Motors, valves and interlocks are controlled by PLC or DCS |
| History | Where is the detailed history kept? | Process time-series sit in the historian |
| Analytics | What aggregates, analyses and predicts? | BI, a data platform or AI |
Take one drying step. The PLC reads the sensor and drives the heater. SCADA shows the current state and alarms. The historian keeps the temperature time-series. The manufacturing system ties those readings to the batch, operation and equipment. The quality system handles a deviation if a limit was exceeded. ERP receives only the business facts it needs. A BI or AI tool may look at trends. The same physical event passes through many systems that do not have the same job.
3. The layers of a pharmaceutical plant
The table lists the layers you will meet and the question each one answers. Names and acronyms vary by vendor; the roles are what matter.
| Layer | Typical systems | The question it answers |
|---|---|---|
| Enterprise transactions | ERP (finance, purchasing, sales, inventory value, MRP, costing), HCM | What did we plan, buy, sell, hold and pay for? |
| Warehouse and logistics | WMS (physical stock, locations, picking), TMS (transport) | Where is the stock physically, and what moves next? |
| Manufacturing operations | MES, electronic batch record (EBR), scheduling | How was this batch actually made? |
| Plant control and OT data | PLC, DCS, HMI, SCADA, historian, edge gateway | What is the process doing right now, and what did it do? |
| Quality and laboratory | QMS (deviation, CAPA, change control, complaints), LIMS (samples, tests, results), CDS for chromatography data | What quality events occurred, and what do the test results say? |
| Controlled documents and training | eDMS (SOPs, specifications, protocols), LMS (training, qualification) | Which version is effective, and who is trained to do this? |
| Product and regulatory | PLM, RIM, artwork and label management | What is the approved product, and where is it registered? |
| Serialisation and traceability | Line-level printing and verification (often called L1/L2), site serialisation system (L3), enterprise repository (L4), connection to national or trading-partner systems (L5) | Which serial numbers were assigned, packed into which case and pallet, and reported where? |
| Assets and facilities | CMMS/EAM, calibration, BMS, environmental monitoring | What is the condition and history of the equipment and the facility? |
| Data, integration and security | Master data management, API gateway, event bus, data warehouse, BI/AI, identity and access management | How do systems connect, and who may do what? |
The L1–L5 numbering for serialisation is an industry convention rather than a formal standard; the requirement behind it comes from rules such as the EU Falsified Medicines Directive and the US Drug Supply Chain Security Act.
Not every company needs every layer. A distributor does not need an MES or SCADA; a small plant may run warehouse and manufacturing functions inside its ERP. The map is for deciding where a boundary is, not for buying all the boxes.
4. One batch, seen by six systems
The same lot appears in many systems. Here is an illustrative batch with invented numbers:
| System | What it knows about batch B260001 |
|---|---|
| ERP | Planned quantity 500,000 tablets; materials issued; cost; finished goods in stock, first in quarantine, then released once the quality unit has decided |
| MES / EBR | Ran on press TP01 from 08:00 to 10:30; operator A; good output 496,900; rejects 1,200; in-process and retained samples 400; recorded losses 1,500; yield 99.4%; steps and signatures |
| Historian | Compression force, turret speed and alarms over time |
| LIMS | Samples, tests, results and review status |
| QMS | Deviation DEV-026, if one was raised for this batch |
| WMS | Where the stock is physically held, and (after QA release) which stock may be picked |
Nobody is wrong here. Each system describes the batch from its own responsibility. Note what is missing from the table: no system "releases" the batch on its own. The quality unit reviews the batch record, the test results and any deviations, takes the decision, and the result becomes a batch status that ERP and WMS enforce. A good boundary stops ERP from trying to be a historian, the manufacturing system from keeping its own copy of laboratory results as if it were a LIMS, and the quality system from creating a separate equipment master that drifts away from the asset system.
5. Pairs that are easy to confuse
| Pair | The difference |
|---|---|
| ERP vs MES | ERP: planning, enterprise transactions, costing. MES: shop-floor execution and the actual evidence of what was produced |
| ERP vs WMS | ERP inventory is the business and accounting view; WMS runs the physical warehouse in detail |
| QMS vs LIMS | QMS handles quality processes and events; LIMS handles samples, tests and results |
| LIMS vs CDS | LIMS manages laboratory workflow; the chromatography data system acquires and processes chromatography data |
| SCADA vs historian vs MES | SCADA supervises the process now; the historian keeps its time-series; MES adds batch and operational context |
| CMMS vs EAM | CMMS is maintenance work management; EAM covers the whole asset life cycle |
| PLM vs PIM vs ERP item master | PLM: engineering life cycle. PIM: product content for sales channels. ERP item: the transactional material or product record |
| eDMS vs distributor DMS | Same acronym, unrelated systems: document management versus distributor-network management. Never use "DMS" without context |
The SCADA example from the table, in a sentence: SCADA knows that the machine is running at 60 °C; the manufacturing system knows that this machine is running batch B260001, drying step, operator A.
6. Who typically owns what
There is no single correct table, because ownership depends on how a company designs its landscape. This is a typical starting point for a pharmaceutical manufacturer:
| Data | System of record (typical) |
|---|---|
| General ledger, payables, receivables | ERP / finance |
| Official purchase and sales orders | ERP (or an order management system, depending on the model) |
| Item and material master (transactional) | ERP, often with master data management |
| Product engineering definition | PLM |
| Regulatory dossier and registration status | RIM |
| Master batch record / master recipe | MES / EBR (approved under change control; the ERP bill of materials must match its version) |
| Detailed production execution and electronic batch record | MES / EBR |
| Process telemetry | Historian |
| Machine control | PLC / DCS |
| Laboratory sample, test, result | LIMS |
| Specification limits used for testing | LIMS (configured from the approved specification in the eDMS) |
| Chromatography raw data | CDS / scientific data management |
| Deviation, CAPA, change control, complaints | QMS |
| Batch disposition decision | Quality unit, using MES, LIMS and QMS evidence |
| Material and batch quality status (quarantine, released, rejected, blocked) | ERP or WMS, changed only as a result of the quality unit's decision |
| Controlled SOPs and specifications | eDMS |
| Personnel training and qualification | LMS (or HCM, if designed that way) |
| Warehouse tasks and bin locations | WMS |
| Employee master | HCM / HRIS |
| Maintenance work | CMMS / EAM |
| Technical life cycle of equipment | EAM |
| Equipment qualification and calibration status | Validation / calibration management (often within CMMS or EAM) |
| Serial numbers and packaging aggregation | Site serialisation system (L3), reported to the enterprise repository (L4) |
| Financial value of assets | ERP |
| Identity and accounts | Identity and access management |
| Enterprise analytical data | Data warehouse / lakehouse |
The principle matters more than the table: an important type of data should have one clearly designated official source, not five systems that each act as master. Other systems may keep an identifier, a cache or a projection, but a change to the official record returns to the owner.
Two pairs cause most of the trouble in practice:
- Specifications. The approved specification is a controlled document in the eDMS, while the limits that results are checked against are configured in the LIMS. If they are changed separately, a result can pass in the LIMS against a limit that is no longer approved. Link them through one change-control record.
- Master batch record and bill of materials. The master recipe lives in the MES, the bill of materials in ERP. When one is revised without the other, the shop floor and the planners work from different versions. Treat them as one change.
The same applies to batch status. If ERP and WMS each keep their own quality status, stock that is still in quarantine in one system can be picked in the other. One system holds the status; the other reads it.
7. Keep data in its own layer
A frequent design mistake is to send everything to ERP. A press sampling every 100 milliseconds should not write into the ERP database. A better flow is:
PLC → edge / SCADA → historian → aggregation and context → MES → ERP
For a tablet press that looks like this:
- Historian: compression force every 100 ms, turret speed every second, alarm timeline.
- MES: batch, start and end time, machine, quantity produced, rejects, yield.
- ERP: production order completed, finished goods quantity, material consumption, inventory movement and cost.
A caveat for regulated plants: if historian data is used to support a GMP decision, for example whether a critical process parameter stayed within its range, the historian becomes a GxP system and must be validated as one. Excursions and summaries of critical parameters should also appear in the electronic batch record, so the batch review does not depend on someone browsing raw trends.
The same logic applies to investigations. A deviation record in the QMS should hold references (historian record reference, batch ID, equipment ID, time range) rather than a copy of millions of sensor readings. A reference only works if the data behind it stays complete, unchanged and retrievable for as long as the batch record must be kept, which in both the EU and the US runs at least until one year after the batch's expiry date.
Four kinds of data have different owners and life cycles, so keep them apart: master data (item, customer, supplier, equipment), transactions (orders, invoices, goods receipts), events (batch started, test completed) and telemetry and documents (time-series; SOPs, reports, certificates of analysis).
8. How the systems talk to each other
Three patterns cover most cases:
- Point-to-point: each pair of systems has its own connection. Quick to start, hard to grow.
- Hub (integration platform): systems connect through a central layer that transforms and routes.
- Event-driven: the source publishes an event such as batch released, and every interested system subscribes. The source does not need to know all its consumers.
Use APIs for request-and-response (get an item, create an order, check stock) and events for things that have happened (order created, batch started, test completed, batch released).
At the equipment level, one established option is OPC UA, which the OPC Foundation describes as a platform-independent interoperability standard for the secure and reliable exchange of data in industrial automation [2]. Above the plant floor, APIs and events are the usual choice.
On security, do not let ERP or a user laptop connect directly to a PLC. Keep corporate IT, an OT demilitarised zone or edge layer, and the plant network as separate zones with controlled crossing points. The ISA/IEC 62443 series, which addresses the security of industrial automation and control systems, is the usual reference for this zoning [6].
Where the standards sit. The ISA-95 / IEC 62264 standard describes five levels, from the physical process (level 0) through sensing and control, supervision and manufacturing operations management (level 3) up to business planning and logistics (level 4, which includes ERP). The standard primarily deals with the interface between levels 3 and 4 [1]. It is useful less as a rigid pyramid than as a way to decide where responsibility changes hands. In batch plants, ISA-88 (batch control) is often used alongside it to structure recipes and equipment inside the manufacturing layer.
9. When to split a system out of ERP
Do not create a separate system just because it has its own acronym. A separate system is justified when one or more of these is true:
- the domain needs depth beyond what the ERP module offers;
- the data frequency is very high;
- the work is real-time;
- it involves physical control;
- transaction volume is very large;
- there is a specialised workflow;
- regulation or validation treats it separately;
- it must work offline or locally;
- it needs equipment integration;
- it needs specialised optimisation.
Two examples. A small manual warehouse may be well served by ERP inventory; a large distribution centre with barcode scanning, wave picking and automation justifies a WMS. A plant with simple production may do fine with ERP manufacturing; a plant that needs electronic batch records, equipment state, electronic signatures, batch genealogy and review by exception will justify a dedicated MES/EBR.
The counterweight: every split creates at least one interface that has to be specified, validated, monitored and owned, and in a regulated plant each of those is a change-controlled item. Count that cost before adding a system, not after.
10. The regulated-industry angle
In pharmaceuticals, several of these systems hold records that are subject to good manufacturing practice, so they fall under the expectations for computerised systems and electronic records. Two commonly cited sources are the US rule on electronic records and electronic signatures, 21 CFR Part 11 [3], and Annex 11 (computerised systems) of the EU GMP guidelines [4]; ISPE's GAMP 5 is the widely used guide to a risk-based approach to computerised system validation [5].
The EU rules are being updated: a revised Annex 11, together with a revised Chapter 4 (documentation) and a new Annex 22 on artificial intelligence, was published as a draft for consultation in 2025. Check the current status before relying on either version.
The idea that links "system of record" to regulation is data integrity: records should be attributable, legible, contemporaneous, original and accurate (often summarised as ALCOA, or ALCOA+ with complete, consistent, enduring and available). FDA's guidance Data Integrity and Compliance With Drug CGMP [7] and the PIC/S guidance PI 041 on data management and integrity [8] are the usual references.
Where ownership sits has practical consequences: it determines which system's access controls, audit trail and validation cover a given record, and where a change must be made and approved. Annex 11 also expects named people behind each system: a process owner, responsible for the business process, and a system owner, responsible for the system's availability, support and maintenance. A record with two "masters" usually also has two sets of owners, and neither is clearly accountable. This guide does not describe what any specific regulation requires of a specific system; check the current text and your own quality system.
On AI and analytics: a data platform or AI tool is a consumer of governed data from the systems of record, not a source of truth for transactional records.
11. Build your own ownership matrix
A simple exercise that surfaces most boundary problems:
- List the data types that matter to you (batch record, lab result, equipment master, supplier status, training status, stock status, controlled document).
- For each, name the owner: the one system allowed to hold the official record, and the people behind it, a process owner for the business process and a system owner for the system.
- Mark the consumers: who reads, references or caches it, and how (API, event, file).
- Write down the conflict rule: if two systems disagree, which wins?
- Check the integration: is each consumer updated from the owner by a controlled mechanism, not by retyping?
- Review when you add a system, change a vendor, or merge sites.
Companies that do this often find a few records with more than one owner, typically specifications, batch status or equipment data, and fixing those can remove more day-to-day friction than adding another module.
Once you know which system should own which record, the next step is choosing and qualifying it; see How to Select and Qualify Software for a Pharmaceutical Plant for requirements, vendor assessment and contract terms.
If you are planning new equipment or process technology as part of this, you can browse the Manufacturing Equipment & Utilities categories on PharmaCountry. For a wider view of where automation fits across the whole supply chain, see Automation & AI Across the Pharmaceutical Supply Chain: An End-to-End Map, and for the supplier side of the quality system, How to Qualify a Pharmaceutical Supplier.
FAQ
Is an ERP enough for a small pharmaceutical manufacturer? Sometimes ERP covers purchasing, inventory, simple production and finance well. The need for separate manufacturing, laboratory or quality systems depends on the complexity of your operations and on regulatory expectations. Section 9 gives the tests.
What is the difference between MES and EBR? An electronic batch record captures the steps, actual values, signatures, materials and deviations of a batch. It can be a module of an MES or a separate system.
Where is batch release recorded? The decision belongs to the quality unit (in the EU, a Qualified Person certifies the batch). It is based on the batch record, test results and deviations, and is usually recorded as a batch status in ERP or WMS, sometimes in the MES. What matters is that one system holds that status and the others read it.
Can two systems share one database and still have separate ownership? Yes. Ownership is about who may create, change and approve a record, enforced by permissions, interfaces and process, not about physical separation.
Does the quality system store lab results? It can show them to support an investigation, but the official result remains in the laboratory system; the quality system references it.
Sources
[1] ISA. ISA-95 / IEC 62264, Enterprise-Control System Integration. https://www.isa.org/standards-and-publications/isa-standards/isa-95-standard
[2] OPC Foundation. What is OPC? https://opcfoundation.org/about/what-is-opc/
[3] US eCFR. 21 CFR Part 11 — Electronic Records; Electronic Signatures. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-A/part-11
[4] European Commission. EudraLex Volume 4 — EU GMP guidelines (including Annex 11, Computerised Systems). https://health.ec.europa.eu/medicinal-products/eudralex/eudralex-volume-4_en
[5] ISPE. GAMP 5: A Risk-Based Approach to Compliant GxP Computerized Systems, Second Edition (2022). https://ispe.org/publications/guidance-documents/gamp-5-guide-2nd-edition
[6] ISA. ISA/IEC 62443 Series of Standards. https://www.isa.org/standards-and-publications/isa-standards/isa-iec-62443-series-of-standards
[7] US FDA. Data Integrity and Compliance With Drug CGMP: Questions and Answers (December 2018). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/data-integrity-and-compliance-drug-cgmp-questions-and-answers
[8] PIC/S. PI 041-1, Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments (2021). https://picscheme.org/docview/4234
This guide is general information, not regulatory or legal advice. Examples, batch numbers and quantities are illustrative. How systems are divided varies between companies and should follow your own quality system and risk assessment.

Leave your rating