A designated system, data set, or process that is treated as the authoritative reference for a specific type of information.
In industrial and manufacturing contexts, a **source of truth** is the specific system, database, document set, or process that an organization designates as the **authoritative reference** for a well-defined category of information.
It answers the question: *“When systems disagree, which one do we treat as authoritative for this data element?”*
A source of truth is always scoped. It applies to a defined domain such as:
– Material master data (e.g., ERP or PLM)
– Production execution data (e.g., MES or SCADA historian)
– Quality results (e.g., LIMS or QMS)
– Inventory balances (e.g., ERP or WMS by location/type)
The designation is usually formalized through governance, procedures, and integration rules, not by the technology alone.
In regulated and complex manufacturing environments, the term commonly refers to:
– **Master data ownership**: Identifying which system owns product definitions, BOMs, routings, recipes, specifications, or equipment hierarchies.
– **Transactional ownership**: Defining which application is authoritative for production orders, batch records, material movements, and inventory positions at different stages.
– **Regulatory records**: Determining which record set (for example, electronic batch records, deviation reports, or calibration records) is treated as the official record for compliance and audits.
– **Reporting and analytics**: Establishing which data store or model is the reference for KPIs such as OEE, yield, scrap, or on-time delivery so reports are consistent.
These designations drive how interfaces are built, how discrepancies are triaged, and how investigations are documented.
A source of truth:
– **Is authoritative only within its defined scope.** A system may be the source of truth for one data type (e.g., inventory valuation) and not for another (e.g., real-time tank level).
– **Does not guarantee correctness at all times.** It is the reference point for decisions and corrections, not a claim that the data is error-free.
– **Is more than a data copy.** Downstream replicas, data warehouses, and reports that consume data are not sources of truth unless explicitly governed as such.
It is distinct from:
– **Single database storage**: Having data physically in one place does not automatically make it the authoritative reference.
– **Canonical data model**: A canonical model defines a common structure and semantics; a source of truth defines which system is authoritative for specific elements.
The term “source of truth” is sometimes used loosely to mean:
– “The system we prefer to look at,” even when other systems remain authoritative for specific data points.
– “The system that is most accurate in practice,” without formal governance or procedures to support that claim.
More precise use in operations and IT architectures:
– **Single source of truth (SSOT)**: Often used for enterprise master data, though in practice most manufacturing environments operate with **multiple sources of truth**, each clearly scoped.
– **System of record**: Closely related; often used interchangeably. In some organizations, a system of record is the application that creates and maintains the data, while the source of truth is any designated authoritative reference, which might be the system of record itself or a governed consolidation layer.
Where MES and ERP balances or transactions do not match, organizations typically:
– Define **source-of-truth rules by balance or transaction type** (for example, ERP as source of truth for financial inventory value, MES as source of truth for in-process quantities on specific equipment).
– Use these rules to drive **reconciliation and triage**: identifying which system must be corrected and how adjustments are recorded.
– Document these designations under **change control and data governance**, so that investigations, CAPAs, and audits refer back to the same agreed authority for each data category.
In this context, “source of truth” is a governance decision that shapes how discrepancies are handled, not a property that either MES or ERP inherently possesses in all cases.