Under the Hood of the WegoScore (Part II)
With our large, rich dataset of multifamily properties from across the country, we were tempted to create a complex formula to distill a score from many building characteristics and auxiliary data. But after much investigation, we decided on a simple, intuitive approach.
Peer Groups Make Comparisons Simple and Fair
A building's WegoScore is computed relative to similar buildings' performances. We define 'similar' based on just two characteristics: climate zone of the building and construction type (high-rise, low-rise, and such). So you won't see a three-story California apartment complex go head-to-head with a Chicago high-rise.
NOTE: We even worked out the math to define peer groups with variable numbers of characteristics, drilling down to heating and cooling configurations, income level, building age, and so on until the group size became too small. But, simplicity prevailed.
Peer groups determine the constraints building owners and managers play within. High-rises must have elevators, and buildings in Boston must have heating. Beyond these constraints, owners have the prerogative to upgrade old buildings, add swimming pools, crank the air conditioning, or install high efficiency washers (all of which will affect their WegoScore).
Once we've defined the peer groups, we let the usage data speak for itself. We make no effort to adjust the empirical distributions by drawing trend lines (regressions) through the various axes.
Carbon Levels the Playing Field
It would be unfair to score a building based on on-site energy use alone. Electric heaters have the advantage of being nearly 100% efficient, but that energy was made via a less efficient process at the source, wherever the region's generators are.
A good way to account for these differences is to keep track of the pounds of CO2 each fuel type emits. For electricity, that number will depend on the regional mix of natural gas, coal, nuclear, hydroelectric, etc. We recently added this finer-grained information into WegoWise, and take advantage of it for the WegoScore, too.
Taking carbon into account levels the playing field, as the plot of cold-weather low-rise buildings below demonstrates.
Gas-heated buildings stick out as the secondary peak on the lefthand plot, at about 100 kBTU/sqft/year. However, once carbon dioxide is taken into account, these two peaks nearly merge into one, as seen on the righthand side. That's because electricity generated in New England (where most of our cold-weather buildings reside) has roughly twice the carbon intensity of natural gas burned on site.
In some regions, such as New England, where the carbon impact for electricity is lower, electric heat presents a remarkably efficient and environmentally responsible option. This realization surprised us, as we had internalized that electric heat was a 'dirty' choice due to the higher national average electric carbon coefficient. So the WegoScore taught even its creators some actionable insights about building efficiency.
More Bonus Q&A
It's tough to get tenant data for a whole-building score. Is there a common-area-only option?
Unfortunately, no. We encountered too many metering configurations shared between common space and apartments which led to biased or uncertain usage estimates.
It's great that you take climate zone into account, but why not something more granular like weather normalization?
We investigated the possibility of incorporating heating or cooling degree days (DD) into the analysis, and are certainly open to using DD in the future. But for now, the method raises some major questions.
First, there's no obvious way to take DD into account. For example, simply normalizing total use by DD also divides temperature-independent base load by DD, which doesn't square with reality. If instead, just the (inferred) heating use is divided by heating DD, there's no natural way to combine that value with the base load to yield a score, since the two quantities have different units.
Moreover, we can't in good conscience use DD relative to a fixed value, like 65 Fahrenheit for heating. We've seen stark evidence of people setting their thermostats to different set points. In the WegoWise app, we fit individual set points to each building to disaggregate heating, cooling and base load. If we performed this analysis for every WegoScore, we would be hiding many finely tuned parameters in our otherwise simple analysis.