The bottom line
Data governance in manufacturing isn't a compliance exercise - it's what determines whether your analytics are trusted or ignored. If plant managers won't act on a dashboard because they don't trust the data behind it, the governance problem is costing you more than the technology problem.
In This Article
Why OT Data Has No Natural Owner
In a typical manufacturing organisation, IT owns the enterprise systems - ERP, WMS, and the corporate network. The engineering team owns the OT systems - PLCs, SCADA, MES, and the plant-floor network. Operations management owns the production KPIs. Quality management owns the quality data. Maintenance owns the CMMS.
None of these functions owns the data that flows between them. OEE is a good example: it is calculated from availability data (operations), performance data (engineering), and quality data (quality management). Each function has a defensible claim to ownership of one component. Nobody owns the composite metric - which is why OEE figures routinely differ by 5–10 percentage points depending on who ran the calculation and which data source they used.
This is not an organisational failure. It is the natural consequence of a domain where operational responsibility is distributed across functions and operational technology was designed for control, not for data governance. The answer is not to reorganise - it is to establish governance conventions that work within the existing structure.
When OEE figures differ by 8 points depending on who calculated them, the problem is not the formula - it is the absence of an agreed data owner for each component.
The Minimum Viable Governance Framework
Manufacturing analytics does not require a full enterprise data governance programme to get started. It requires three things: agreed KPI definitions, a single equipment hierarchy, and named data stewards per operational domain.
Agreed KPI definitions means writing down - and getting sign-off from operations, engineering, and quality - exactly how OEE, MTBF, first-time-right rate, and the five to ten other metrics that drive decisions are calculated. Which downtimes are included in availability? What counts as a speed loss versus a planned rate? These are political questions as much as technical ones, and they must be resolved before any analytics programme produces numbers that people trust.
A single equipment hierarchy means one agreed list of assets, with consistent naming, consistent grouping (line, cell, plant), and consistent maintenance codes. This is the spine of manufacturing analytics. Every downtime event, every maintenance work order, every quality reject needs to attach to an asset in this hierarchy - otherwise cross-asset analysis is impossible and trend data is meaningless.
The $12.9M Annual Cost of Getting It Wrong
Gartner estimates that poor data quality costs organisations an average of $12.9 million per year. (Source: Gartner Data Quality Market Survey, 2023) In manufacturing, the direct cost is higher because operational decisions are made continuously and at scale - a wrong inventory number drives a wrong replenishment decision that drives a stockout that drives a lost sale that drives a margin hit. The chain is short and the compounding is fast.
The indirect cost is harder to measure but arguably more significant: analytics programmes that produce numbers nobody trusts get abandoned. Once a dashboard is labelled as "unreliable" by the operations team, the investment is effectively lost - regardless of the quality of the underlying platform.
Governance is the investment that makes all other analytics investments pay off. A well-governed dataset with clear ownership, agreed definitions, and named stewards produces numbers that operations teams trust and act on. Without that foundation, the sophistication of the analytics platform is irrelevant.
I've seen organisations with exceptional technology and unusable analytics because nobody agreed on what 'OEE' meant across sites. And I've seen organisations with modest tools and disciplined data governance producing analytics that actually drive decisions. The governance comes first. Everything else follows.