This article describes Measurabl's Efficiency Percentile, our building performance benchmarking technology.
Measurabl's Efficiency Percentiles compare buildings based on how efficient they are and scores them from 0-100 on energy consumption, water consumption, and carbon emissions. There are numerous ways of comparing buildings against each other. At Measurabl, we have developed a proprietary efficiency metric based on each building’s consumption compared to a baseline. This baseline is generated using machine learning models trained on Measurabl’s database of real estate utility data covering more than 13B square feet. The resulting Efficiency Percentiles represent the expectation of how much energy and/or water a building of certain use type and size, in a certain weather zone, should consume at a certain time of the year, and how much carbon it should emit.
Where can I find my portfolio's/building's Efficiency Percentile?
Your portfolio's and buildings' Efficiency Percentiles are displayed on the Portfolio Overview. Individual building's Efficiency Percentiles are displayed on their Site Overview page.
For customers who have access to Cohort Insights, our tool that enables creation of custom cohorts of Measurabl buildings to compare your portfolio against, Efficiency Percentiles are also displayed inside Cohort Insights analyses.
What are the requirements for my building to receive an Efficiency Percentile?
In order to receive an Efficiency Percentile benchmark for a given metric (i.e. energy, water), buildings must have at least 1 meter reading within the measurement period. The default analysis period used on the Portfolio Overview and Site Overview widgets is the 12-month window ending two months prior to the current month. Why does Measurabl offset the "Last 12 Months" analysis by 2 months?
How are Efficiency Percentiles determined?
Measurabl’s database includes time series usage records for tens of thousands of buildings. Because we have so much utility data, we are able to accurately predict what a building should consume based on all of the buildings in our database (across multiple owners, operators, locations...).
Expected usages for these buildings are calculated through machine learning models. These are models trained to recognize patterns in utility consumption. The patterns in our data emerge across different building types, sizes, and climate zones.
The output of these machine learning models is the expected usage for a building which we compare to its actual, tracked utility consumption. The ratio of these two metrics is how we determine the efficiency of a building, and this is ultimately how we rank buildings against one another and assign them an Efficiency Percentile.
What factors are used to determine a building's Efficiency Percentile?
The building factors used by our machine learning model when generating expected consumption values for energy, water, and carbon are: primary use type, floor area, heating and cooling days, meter breakout, month, and year.
We selected factors that are universally available across all use types and locations and, at the same time, are helpful in explaining the variance in consumption patterns. The goal of our machine learning model is to effectively “equalize” all building data so that we have a larger number of comparable buildings, which we can evaluate based solely on their consumption metrics.
Which buildings from Measurabl's database are my buildings compared against for determining their Efficiency Percentiles?
Measurabl compares your buildings' performance against every other comparable building in Measurabl.
Customers with access to Cohort Insights can filter down the universe of Measurabl building data to generate Efficiency Percentiles for their buildings relative to more targeted cohorts, refined by geography, use type, and/or data completeness.
How are data outliers handled?
Outliers are excluded from the data used to generate Efficiency Percentiles. This prevents outliers from having an impact on the baseline usage. Since there is no general way to tell the difference between an outlier whose atypical usage is correct versus one whose data is incorrect, all outliers are excluded indiscriminately.
Why might my building have a particularly high/low percentile?
A site may receive a particularly high or low percentile for one of the following reasons:
The site’s data in Measurabl is incorrect. Specifically, please verify the following are correct: meter unit types, meter allocation, space square footage, and building location.
The site may have consumed significantly more or less energy, water, or carbon than expected by our machine learning models. This could be the case if the use type of the site is not well represented in Measurabl or if the building is more or less efficient due to factors intentionally not taken into account by our models such as retrofitting projects.
The site’s percentile might be exceptionally low due to poor data coverage.
Can I customize the set of buildings that my buildings will be compared against?
Yes, our Cohort Insights tool enables customers to filter down the universe of Measurabl building data to generate Efficiency Percentiles for their buildings relative to more targeted cohorts, refined by geography, use type, and/or data completeness.
How do I access Cohort Insights
To learn more about Cohort Insights, please reach out to your Measurabl Account Manager or Customer Success Manager.
How often do the Efficiency Percentiles update?
Efficiency Percentiles update monthly. Each month the default 12-month analysis period shifts forward 1 month.
Which members of my Measurabl portfolio can see Efficiency Percentiles?
The portfolio-level Efficiency Percentile widget on the Portfolio Overview page is visible to Portfolio Managers, Portfolio Members, Subgroup Managers, and Subgroup Members.
The site-level Efficiency Percentile widget on the Site Overview page is visible to all users in the portfolio that have access to that site, including Site Managers.
Are details about my buildings revealed to other Measurabl customers through Efficiency Percentiles?
No. Your data is not ever revealed to other Measurabl users.
How does Data Coverage impact the calculation of Measurabl’s Efficiency Percentiles?
The Measurabl Data Coverage combines floor area coverage and meter data completeness within a single metric by taking into account the space-floor area-weighted data completeness contributions of each meter within a building. In other words, the total meter completeness of a space is the sum of the completeness values of all meters assigned to the space. Measurabl’s Efficiency Percentiles are considerate of this Data Coverage metric and users will see that improving their floor area coverage and meter data completeness will, in most cases, improve their Efficiency Percentile.
Why does Measurabl offset the "Last 12 Months" analysis by 2 months?
On average, more than half of utility data comes in at least 6 months after the end usage date. About one third of data is in after 2 months just slightly more after 3 months. There is a built-in 2-month lag to account for this data entry behavior. There would not be sufficient data for meaningful comparisons within the first 2 months of data entry because the comparison data set would only include less than 1/3 of the Measurabl database.
Are percentiles a common methodology used for comparative analytics?
Yes. Leveraging percentiles to represent one’s rank within a dataset is common. However, the means for which they are ranked, in our case expected usage, is highly proprietary to Measurabl.
Why aren't Efficiency Percentiles based on usage intensities?
Usage intensity is usage divided by floor area. It is correlated with efficiency but does not tell the whole story. There are other variables which contribute to usage efficiency besides floor area—weather, property use type, seasonality, who pays the bill, to name a few—all of which are included in Measurabl's Efficiency Percentiles.
How is the expected usage calculated when there are mixed-use sites? How does it roll up to the property level?
As energy is often reported at the space level, we calculate expected monthly energy usage for each space in Measurabl. This allows us to calculate expected energy usage with a higher accuracy for buildings of mixed use. The space-level expected usages are rolled up to calculate a building-level expected monthly energy usage.