How savings are measured
The method behind savings tracking: the calibrated model run with each year’s actual weather and the bills’ periods, compared with the bills, the costs used to price the savings, and when a year or its months so far can be measured.
For: Everyone with access to the buildingWhere: Analysis › Savings Tracking › How it's measured
On this page
The method
Savings tracking follows IPMVP Option D, calibrated simulation. For each tracking year, InverseAudit runs the building’s calibrated model twice, both times with that year’s actual weather:
| Model | Answers | |
|---|---|---|
| Baseline | The calibrated model of the building before the retrofit | What the building would have used without the measures |
| Expected | The same model with the installed measures | What it should use with them |
Then it compares both with the year’s bills, month by month:
- Savings achieved = baseline - actual bills
- Expected savings = baseline - expected
- Achieved vs expected = savings achieved ÷ expected savings
Running the baseline with the year’s own weather is what makes the comparison fair: a hot summer raises the baseline’s cooling too, so it doesn’t hide the savings, and a mild winter doesn’t create savings that the measures didn’t make.
Energy saved is the same comparison in energy instead of dollars, for each fuel: (baseline - actual bills) ÷ baseline.
The baseline model
The baseline is the calibrated model as it was when tracking started, with its data year and its fit (NMBE and CV(RMSE)) shown under How it’s measured. It’s kept as it is: new baseline bills or a recalibration later don’t change the tracking, so every year is measured against the same building.
The expected model adds only the measures ticked under Measures installed. Measures that weren’t installed don’t count, so a measure left out doesn’t make the results look worse than they are.
The weather
Each year uses the building’s actual weather for that year, from the same location as its calibration, prepared the same way as the calibration’s weather. The weather is downloaded once: a later run uses what’s already there and adds only new days.
A month’s weather is used only from the 10th of the next month. In the first days after a month ends, the weather data for its last days can still be forecasts or averages instead of measurements. This rule applies to the monthly estimates, to Update now, to the savings so far and to calculating a year.
The billing periods
When the bills have their last days of billing period, the year is run over the bills’ own periods. Each month’s model use then covers exactly the days of that month’s bill: the year’s weather starts the day after the last bill ended a year earlier, and the model’s months end on the bills’ last days.
Without the last days, the year is run over calendar months. Bills that end mid-month then compare with a month that’s partly different, which matters most in the seasons when the weather changes fastest.
Costs
The savings are priced in dollars so that the years can be compared and added up:
- Monthly bills - Every kWh and therm, of the baseline, the expected model and the bills, is priced at the model’s average cost per unit when tracking started, shown under Costs. Using one price for all three means the savings measure energy, not rate changes.
- Interval data - The baseline and the expected model are priced with the model’s utility rates, demand charges included. The bills’ use is priced at the baseline’s average cost per unit that month.
The bills’ own costs are used to check the price. When a year’s bills average more than 2% more or less per unit than the model, the page says so, and Rerun the baseline prices the measured years at the bills’ average cost per unit. See When the bills cost more or less than the model.
When a year can be calculated
A year can be calculated when:
- All 12 months have their bills, for electricity and, when the building uses it, natural gas.
- The weather for its last bill is final: 10 days after the last bill ends, when each utility has all 12 last days of billing period; otherwise from the 10th of the month after the year’s last month.
For example, a year whose last bills end on February 14 can be calculated from February 24. The Bills dialog says what’s still missing, or from which day.
Savings so far
Before a year can be calculated, its months so far can be measured already. They’re the months that have bills, one after the other from the year’s first month, whose weather is final (as above, month by month).
They’re measured exactly like a full year, with the weather of the 12 months that end with the last of them, and shown as Preliminary. They aren’t scaled up to a year: a part of a year isn’t a year, because gas savings come mostly in winter and electricity savings more in summer. Preliminary results stay out of the tiles, the total, the PDF and the emails.
What it doesn’t adjust for
The measured savings include everything that changed the building’s energy use since the baseline, not only the measures: more occupants, longer hours, new equipment or a vacancy. InverseAudit doesn’t adjust the baseline for these. Note them under Changes this year in the Bills dialog: they’re shown with the year’s results, so whoever reads the results knows what else changed.
The monthly estimate before tracking
Before the measures are installed, the same two models run each month with the last 12 months’ actual weather. The difference is what the selected measures would have saved. See Before the measures are installed.