If a plant cannot yet support a required global KPI definition, the answer is not to pretend that it can.
A KPI should only be reported as globally comparable when the plant can produce it from the required source data, business rules, and calculation logic with enough consistency to make cross-site comparison credible. If those conditions are not met, the metric should be flagged as not yet conformant to the global definition.
In practice, most organizations use a staged approach:
-
Document the exact gap. Is the issue missing source data, different event definitions, manual workarounds, poor timestamp quality, weak master data, or incomplete integration between MES, ERP, QMS, historian, or spreadsheets?
-
Classify the plant’s reporting status. For example: fully aligned, partially aligned, proxy only, or not reportable.
-
If the business still needs visibility, allow a controlled local proxy metric, but label it clearly as non-standard and not directly comparable to the global KPI.
-
Put the definition, mapping logic, owners, assumptions, and exceptions under change control.
-
Create a remediation plan with data, process, and system actions required to reach the global definition.
This is usually a governance problem as much as a technical one. A plant may be operationally capable but still unable to support a KPI because event capture is inconsistent, production states are interpreted differently, scrap is booked late, rework is handled outside the system, or quality dispositions sit in separate workflows. In regulated environments, those differences matter because traceability and evidence quality matter.
What not to do
-
Do not force local teams to backfill numbers into a global template without documenting method and limitations.
-
Do not merge proxy values with standard values and present them as one clean benchmark set.
-
Do not treat dashboard adoption as proof that the KPI is valid.
-
Do not launch a full system replacement just to satisfy one KPI unless the broader qualification, validation, downtime, and integration case is actually justified.
That last point matters in brownfield plants. Full replacement strategies often fail because the real problem is not one application but years of local process variation, legacy interfaces, qualification constraints, and long equipment lifecycles. Replacing MES, ERP, QMS, or plant data collection to standardize one metric can create more risk than value if the plant cannot absorb the change.
What a defensible interim state looks like
A defensible interim state is usually possible if the organization is explicit about limitations. That typically includes:
-
a published global KPI definition and calculation rule
-
a site-by-site mapping of which inputs are available and trusted
-
a marked local proxy where the standard KPI is not yet achievable
-
metadata showing source systems, refresh timing, and manual intervention points
-
review and approval of the interim method by the responsible business and data owners
This does not make the proxy equivalent to the global KPI. It only makes the limitation visible and controlled.
Tradeoffs
The tradeoff is straightforward. If you wait for perfect standardization, leadership may lack needed visibility for too long. If you standardize too aggressively on weak data, you get false comparability and bad decisions. Most organizations need a middle path: partial harmonization now, with explicit confidence levels and a roadmap to full alignment.
Whether that works depends on process maturity, data readiness, integration quality, and the discipline to maintain a business glossary and canonical logic over time. Without that, the same KPI name can hide different operational realities across plants.