How do we stop EPMS and BMS reporting different energy values?

When the EPMS says a chiller drew 280 kW and the BMS historian says 312 kW, the problem is rarely measurement accuracy — it is source-of-truth assignment. Without a single document declaring which system is canonical for each measurement, two systems with two near-correct readings produce one operationally untrusted dataset.

Fix it at the data model

• For each measurement type — real power, apparent power, kWh, demand — name one canonical source.
• Every other system mirrors from the canonical source via the integration layer, on a published refresh interval.
• Reported values everywhere else carry a source attribution in the metadata — the operator can always trace back to the canonical reading.

This is a 1-2 page document. The mistake is treating it as an integration deliverable rather than a procurement-stage governance artefact. The earlier it lands, the cheaper the convergence is.