← Blog Home

The Science Behind the Tendril ESM Platform: Part 2 of 3: Data Science

In this age of Big Data there’s no shortage of information about utility customers and their energy use. We can track customers’ behavior by demographic, by program enrollment, by location … really by whatever we want. But to use this data wisely, we need to apply scientific principles to our aggregation and clustering processes. This science of analyzing data is aptly named Data Science, and we rely on it heavily here at Tendril.

While behavioral science helps us tailor our offerings and their delivery to certain customers, data science helps us pull the information we need from our data sets to guide us in our application of behavioral science principles.

Make sense? I’ll say it more simply: data science lets us filter the information we gather so that we can use it to optimize our product design, our customer interaction, and even identify and target customers with personalized recommendations for energy-related products, services and programs.

Here are some of the types of data science we put to work on a regular basis:

Grouping: One of the grouping algorithms we use is clustering, which leverages the Canopy and K-Means modules within the Mahout library to create subsets within larger collections of data. For instance, to fine-tune Home Energy Reports, we use neighborhood cohort grouping. We can show customers tailored comparisons of their energy consumption and that of the people living closest to them, which increases the relevance and value to the customers’ lives. 

  • Categorizing: Though it also helps break down large data sets into manageable sections, categorizing works differently than grouping. Here, classification algorithms help assign data points to sets of pre-existing categories, For instance, with one of our customers – Seattle City Light – we analyze bill and weather data to determine which customers use electric versus gas heating. This allows us to deliver recommendations that are targeted to the unique needs of the user.
  • Estimation/Prediction: This set of algorithms uses techniques that predict a home’s energy consumption based on a number of input parameters. We use it to predict energy consumption when little or no other data are available.
  • Recommendation: This tool uses analysis of a customer’s past behavior to come up with customized recommendations for future or additional product offerings. For example, this capability is useful when recommending new savings actions when analyzing a customer’s savings plan.

Because of data science, we at Tendril can make calculated decisions about how to help our customers increase their energy efficiency. We can also continue to meet customers on their term, which makes them more likely to take an interest in their energy usage. 

There’s one more science that informs the way we do things here. Up next, some insight into how we apply Building Science to our work. Coming soon!




  • Continuous Demand Management
  • Customer Ops & CSAT
  • DERs
  • DSM
  • Data Analytics
  • Demand Response
  • Disruption
  • Energy Efficiency
  • HERs
  • High Bill Alerts
  • Privacy & Security
  • Smart Home
  • Solar

    Invalid Email Address

    Subscribe to This Feed

    Thank you for subscribing to Behind the Meter Blog.