A practical route from operating evidence to a defensible decision.
This guide turns the subject into four connected questions. Use it to structure a team discussion, challenge assumptions and agree what must be verified after action.
✓Relate demand to weather, occupancy, schedules and asset condition.
✓Record exclusions, data gaps and normalisation logic.
✓Refresh the reference when the operating boundary changes materially.
Why a baseline needs operating context
A simple historical average can mistake weather, occupancy or production change for waste or savings. A defensible baseline records which conditions shaped demand and which evidence was measured, derived or modelled.
Minimum evidence set
Begin with interval energy, operating schedules, weather, occupancy or production proxies, tariffs and material asset changes. Record gaps explicitly instead of filling them with unexplained certainty.
Keep comparison periods comparable
Document normalisation logic, exclusions and confidence. Revisit the baseline when the estate, operating schedule or principal energy systems change materially.
Use it as a decision reference
The baseline should support investigation, scenario comparison and later verification. It is an operating reference, not a one-off reporting artifact.
A minimum baseline evidence set
Office baseline after a hybrid-working change
A raw year-on-year comparison suggested a 14% reduction. After accounting for lower occupancy, a milder cooling season and a shortened operating week, the comparable improvement was 6–9%. The adjusted range was used for investigation rather than claiming the raw difference as savings.
Before this decision moves forward
01Confirm the meter and estate boundary.
02Name the operating variables that materially affect demand.
03Document missing periods and substitutions.
04Test residuals and unusual periods.
05Assign an owner and refresh trigger.
Carry the evidence through the full decision.
Agree the reference position and name material gaps.
Compare options under consistent conditions and visible assumptions.
Measure what happened and return the outcome to the operating model.