Measuring Savings with Option C
Whole-building savings analysis — baselines, regression models, ASHRAE-compliant results, adjustments, and exportable M&V reports.
Option C is the whole-building approach to measurement and verification. Rather than metering a single piece of equipment, it uses your utility bills for the whole building to answer one question: after the improvements, is the building using less than it would have, and how much less?
Huckleberry builds a statistical model of how the building consumed energy before the work, projects that model forward across the period after the work, and reports the difference as savings — documented well enough to support sustainability claims and incentive programs.
Find it under Analytics → Measurement & Verification → M&V Option-C. The page is organized as four sections you work through in order — Configuration, Data Review, Non-Routine Adjustments and Savings Results — and choosing a model happens in between, once you’ve run the analysis from Data Review.
Use twelve continuous months of baseline bills if the result has to stand up to scrutiny. Huckleberry will fit a model on as little as six to eight months, which is useful for a first look, but ASHRAE’s whole-building compliance path requires an unbroken twelve-month baseline — and a full year is what captures both a heating and a cooling season. See how a model is judged.
Configuration
Choose what you’re analyzing
Pick the site, the service (electric or natural gas), and the meters to include. If a site has several meters, you can analyze them together or narrow to the ones affected by the work.
Bring in the data
Pull from invoices uses the utility bills already in Huckleberry — this is the normal path, and it keeps the analysis consistent with the rest of your dashboards. You can also upload a file if you’re working from data Huckleberry doesn’t hold.
Pick your independent variable
The model needs something to explain consumption against.
- Weather — outdoor temperature, the standard choice for buildings whose load is driven by heating and cooling.
- Your own monthly data — any monthly figure you track as a KPI, such as production volume, occupancy, or operating hours. Useful when weather alone doesn’t explain the building.
- Both — Huckleberry fits weather models, custom-variable models, and blended models, then compares them all.
Record the measures (ECMs)
An energy conservation measure is a piece of work you did: a lighting retrofit, a controls upgrade, new rooftop units. For each one you record the type, the completion and effective dates, a description, and the expected reduction.
You can verify a package of measures in one project — attach several ECMs and mark one as the primary measure. Measures you’ve saved before appear under Saved ECMs, so a measure recorded once can be attached to another project later rather than re-entered.
The effective date is what the analysis uses to divide before from after. A measure completed in March but not fully in service until May should carry May as its effective date.
Set the periods
- Comparison period — the stretch of bills before the work. This is the baseline the model is built from.
- Performance period — the stretch after the work, which the model is projected across.
Huckleberry suggests both windows from the invoice data and the measure’s effective date. You can accept the suggestion or set the dates yourself; if you edit them, your dates are kept.
Data Review
Before any modeling happens, review what came in. The meter breakdown shows each meter’s usage and cost for the periods you chose, with a total, so you can catch a missing month or a meter that doesn’t belong before it distorts the result.
Two settings live here:
- Confidence level — how much statistical certainty the reported savings range is quoted at.
- Weather normalization — analyze against actual weather (what the building really experienced) or against typical weather (a long-run normal year). Typical-year normalization is what you want when you need a number that isn’t distorted by an unusually mild or harsh season, and it’s what makes one year’s result comparable to another.
Then press Run Analysis. Huckleberry fits a family of regression models to the baseline and scores each one.
Choosing a model
The model comparison table lists every model that was fitted, with its fit statistics. Huckleberry marks one as Recommended, grades the others, and flags any that are Invalid. You can accept the recommendation or select a different model, and your choice is recorded with the project.
How a model is judged
Huckleberry checks one thing automatically: the baseline CV(RMSE) limit set out in ASHRAE Guideline 14, section 4.3.2.1(e) for whole-building analysis. The limit depends on how much performance-period data the savings are computed from:
| Performance period | CV(RMSE) limit (energy) | CV(RMSE) limit (demand) |
|---|---|---|
| Under 12 months | 20% | 30% |
| 12–60 months | 25% | 35% |
| Beyond 60 months | 30% | 40% |
R² and NMBE are shown alongside for information — they are useful context, but the criterion in 4.3.2.1(e) is CV(RMSE).
A model marked ASHRAE-valid has passed the CV(RMSE) test, not the whole compliance path. If you are claiming savings under ASHRAE Guideline 14’s whole-building prescriptive path, that path also requires a baseline spanning a continuous twelve months with no gaps in energy or independent-variable data, at least nine valid baseline data points, no baseline points dropped, expected savings above 10% of whole-building use, and a savings algorithm that passes the net-determination-bias test. Those conditions are yours to satisfy — Huckleberry doesn’t verify them for you. Check them before you put a number in an incentive application.
If no model meets the criterion
The page explains which of the usual three causes applies:
- A weak weather–energy relationship — the building’s use doesn’t track temperature closely. Common with heavy plug loads, data centers, or steady internal heat gain. Try a custom variable instead of weather.
- Too few data points — six to eight months is the least the tool will work with and often isn’t enough to find a pattern. Extend the baseline to twelve or more continuous months, which is what the compliance path requires anyway.
- Operational changes during the baseline — if occupancy, schedules, or equipment changed while the baseline was being measured, no consistent pattern exists to model. Move the window, or record the change as a non-routine adjustment.
Model drift
Once a project is saved, Huckleberry watches its model over time and raises a drift alert when the relationship stops holding — CV(RMSE) degrading, a worsening trend, or the savings signal shifting. Drift usually means something real changed in the building, and it’s a prompt to revisit the baseline rather than to distrust the tool. Drift monitoring can be turned on or off per project, and alerts can be acknowledged as you work through them.
Non-routine adjustments
Sometimes the building changes for reasons that have nothing to do with the energy work — a server room is added, a floor is taken out of service, a tenant moves in. Left alone, those changes get counted as savings or as losses.
A non-routine adjustment corrects for them. Add one with a name (for example, “Server room addition”), an optional description, and a justification or source. Documenting why the adjustment was made is what keeps the analysis defensible when someone reviews it later.
Savings results
The results section reports:
- Annual normalized savings — the headline figure, expressed for a full year under typical conditions.
- Normalized savings % — savings as a share of what the building would otherwise have used.
- Typical-year baseline and performance totals — the two figures the savings are the difference between.
Charts show the fitted model against the actual data, the residuals, and savings over time. A glossary is available on-screen for each statistic, so you don’t have to leave the page to remember what CV(RMSE) means.
Export as PDF produces the full report — configuration, periods, measures, model and its statistics, adjustments, charts and results — as one document you can attach to an incentive application or hand to a client.
Saving and comparing projects
Save Project stores everything: scope, periods, measures, the selected model, adjustments and results. Saved projects can be reloaded, updated in place, or saved as a new project to fork an alternative analysis. New Project clears the page for a fresh start.
Project comparison puts two saved projects side by side — their periods, service type, selected model, CV(RMSE) and whether each meets the ASHRAE criterion — which is the quickest way to justify one approach over another, or to show a result holding steady across two reporting cycles.
Related
- Analytics & Measurement & Verification — how M&V fits with report cards and benchmarking.
- Utility Billing & Data — where the bills behind an Option C analysis come from.