The short answer
Look at consumption when the building is empty. Plant left running overnight and at weekends is the most common and most fixable waste, and half-hourly meter data shows it immediately.
That data is usually available from the supplier at no cost and is rarely examined.
What the profile reveals
- Baseload that should be near zero and is not
- Plant starting hours before occupancy
- Equipment that never switches off at weekends
- Consumption unrelated to occupancy or weather
- A step change that marks when something was left on
The step change is worth watching for. Consumption that rose on a particular date and never came down usually means a setting was changed or a control failed, and it can persist for years.
Normalise before comparing
| Compare | Adjusted for |
|---|---|
| Site against site | Floor area and occupancy |
| Month against month | Weather and trading days |
| Year against year | Occupancy changes and scope |
| Against target | What the target assumed |
Unadjusted comparisons produce arguments. A cold month looks like a failure and a mild one like a success, when neither had anything to do with management.
Sequence the work
- Get half-hourly data for your largest sites.
- Examine the empty-building profile.
- Fix the scheduling and controls that account for it.
- Re-measure, and keep monitoring so it does not drift back.
- Then consider sub-metering where the cause is still unclear.
Steps one to three cost time rather than capital and usually capture the largest share of the available saving.
Drift is the recurring problem
Settings get changed for a reason and never changed back. An override for a weekend event stays on for a year. Monitoring exists to catch that, and it only works if someone reviews it regularly.
A short monthly review of the empty-building profile per site keeps the saving rather than capturing it once.