Build a Manufacturing Data Model Multi-Site APAC Teams Can Trust
A regional production report can assign the same night shift to different dates depending on which system calculated it. In Deloitte’s 2025 Smart Manufacturing and Operations Survey of large manufacturers with US headquarters or operations, 54% of respondents reported using a data standard through a unified data model [1]. For APAC manufacturers, the reporting problem is concrete. Plants need shared rules for identifying equipment, assigning shifts and counting production.
A manufacturing data model multi-site teams can use is a shared definition of how factory records identify assets, products, production runs and shifts, and how those records relate across plants.
Key Takeaways
Give assets, products and production runs stable identities. Store event times in UTC and assign shifts using each plant’s time zone and approved calendar. Publish shared definitions for units, production states and KPIs. Before expanding, reconcile the same metrics against source records at plants with different schedules and equipment hierarchies.
1. Define the decisions a manufacturing data model multi-site teams share must support
Start with an operating decision and specify the records needed to make it. If regional operations wants to compare first-pass yield, define what counts as a first attempt, how rework is recorded and which production date owns the result.
Use a real example. A unit fails inspection before midnight, passes after rework on the same shift and ships the next day. First-pass yield should reflect the original inspection. Final acceptance and shipment each answer a different question. Ask quality and production teams to agree on the expected results before building the report.
For each metric, document the population, exclusions, reporting period and source of record. Keep the numerator and denominator available so a regional report can combine plant totals correctly. Agree on a reconciliation tolerance and nominate a plant owner to explain discrepancies. These decisions define the data the integration must capture.
2. Create stable identifiers and relationships
Give sites, equipment, products and production runs stable IDs, then map each plant’s source identifiers to them. Use ISA-95’s role-based equipment hierarchy as a reference for site, area and work-center relationships [2].
Keep an asset’s identity separate from its location. If a test station moves to another line, retain its asset ID and record when its parent assignment changes. Identify source records by site, source system and local ID so equipment labels reused at different plants stay distinct.
| Dimension | Source records | Shared manufacturing model |
|---|---|---|
| Asset identity | Local tag or equipment number | Stable asset ID with source aliases |
| Equipment location | Current line field | Parent assignment with effective dates |
| Product | Local item code | Product ID and revision |
| Production run | Order number reused across operations | Run ID linked to order and operation |
| Material traceability | Lot references in local systems | Lot IDs linked to input and output runs |
| Shift | Crew label or calendar date | Site-specific shift instance and production date |
Join production facts to the relationships in effect when the event occurred. This keeps historical output attached to the line that produced it.
3. Represent time and production context
Store event instants in UTC and assign them to shifts using the plant’s named time zone and approved shift calendar. A UTC offset describes the offset at a particular instant. A named IANA time zone supplies the rules needed to interpret local time, including daylight saving changes where applicable [3].
Record event time and ingestion time separately. If a gateway reconnects after a network outage, a late event should still join the shift in which it occurred. Keep the original time and offset when available, and flag timestamps whose source or offset is uncertain.
Represent each shift as a dated instance with explicit start and end instants. For example, a plant may assign a shift crossing midnight to the production date on which it starts. Include an event at the shift start and assign an event at the exact end to the next shift. Test holidays, overtime, gaps between shifts and repeated local clock times. Where a run spans shifts, allocate timestamped production events to the applicable shift.
4. Standardize meaning and quality
Standardize signal units, production states and KPI rules while retaining source values for troubleshooting. A common field name helps only when plants agree on what the field measures and when it can be used.
For example, one controller reports a cumulative unit count and another sends a quantity per cycle. The cumulative signal needs a difference calculation with rules for resets and rollover. The per-cycle signal needs event deduplication before summing. Record both mappings and verify their output against known production totals.
Carry data quality and provenance alongside each measurement. OPC UA’s DataValue structure distinguishes the value, status, source timestamp and server timestamp [4]. Preserve those fields when available. For a source that supplies a value alone, record the acquisition time and explicitly identify it as such.
Use shared reason codes with local-language labels. Flag missing windows, unmapped codes and inconsistent values for review. Publish coverage with each KPI so a plant with incomplete collection can explain an apparent change in performance.
5. Govern changes across sites
Give regional data owners responsibility for shared definitions and plant stewards responsibility for local mappings. Version schemas, calendars and KPI rules, and record their effective dates so published results can be reproduced after changes.
Technical deployment pipeline
- Inventory sources. Use asset registers, product masters and sample exports to establish identifiers and data ownership.
- Map records. Use a versioned data dictionary and mapping register to resolve local codes into shared entities and units.
- Validate boundaries. Use test fixtures for overnight shifts, late events and asset moves to confirm historical attribution.
- Reconcile a pilot. Use plant source totals and exception logs to prove results across sites with different schedules.
- Release and monitor. Use schema checks, role-based access and change records to control rollout and detect broken mappings.
TMA’s company history records its achievement of CMMI Level 5 and ISO/IEC 27001 [5]. Apply those disciplines by requiring peer review for mapping changes, restricting exports of sensitive production data and testing restoration of a prior approved configuration.
Lessons Learned from the Field
In an anonymized industrial robotics engagement, TMA Innovation describes limited factory-wide visibility caused by fragmented integration across robots, machines, sensors and enterprise systems. TMA brought those sources together on an on-premises platform and enabled live monitoring, dashboards and alerts [6]. The published case gives qualitative outcomes, so a buyer would still need local reconciliation evidence to quantify the value.
For a multi-site rollout, require the delivery team to reproduce a published KPI from its source records. When ownership is unclear, changes made in a dashboard or site script can escape the mapping register. A witnessed replay through a shift change and an equipment reassignment reveals whether plant staff can explain the result after handover.
FAQ
Should every plant use the same equipment hierarchy?
Use shared relationship types and keep local levels where they serve an operational purpose. Map each plant to the levels needed for regional reporting, and retain dated assignments for assets that move.
Should a shift report use UTC or local time?
Store event instants in UTC and classify them using the plant’s approved shift calendar and named time zone. Report the agreed production date alongside the local shift boundaries.
How do we handle late events after a shift closes?
Retain event and ingestion times. Define when the report becomes final, who can reopen it and how corrected totals are published. Link each correction to its reason and affected shift.
Can two plants compare a KPI when their source systems differ?
Yes, when they use the same metric definition, population and exclusions. Reconcile each source mapping against plant records and disclose differences in coverage before combining results.
How do we evaluate a manufacturing data integration vendor?
Ask the vendor to reconcile a real shift, explain a discrepancy and replay the result after a mapping change. Confirm who owns the schema, source mappings, access reviews and post-release support.
Conclusion
Pilot one regional KPI across plants with different shift patterns. Trace the result to source records and verify that it can be reproduced after an asset move or a late event. Share sample exports, shift calendars and your target KPI with TMA Innovation to scope a multi-site data model pilot [6].
References
[1] Deloitte Insights, 2025 Smart Manufacturing and Operations Survey: Navigating challenges to implementation, 2025. Survey
[2] International Society of Automation, ANSI/ISA-95.00.01-2025 (IEC 62264-1 Mod), Enterprise-Control System Integration, Part 1: Models and Terminology, 2025. Standard contents
[3] RFC Editor, Date and Time on the Internet: Timestamps with Additional Information, RFC 9557, 2024. Specification
[4] OPC Foundation, OPC Unified Architecture, Part 4: Services, DataValue, undated. Specification
[5] TMA Tech Group, Giới thiệu TMA Tech Group, undated. Company history
[6] TMA Innovation, T-Box, undated. Manufacturing case study and contact form


