Driving Value in Energy With Collaborative Filtering

Energy Marketing’s New Data Science Technique

Collaborative filtering, a data science technique new to energy marketing, is an innovative substitute for traditional market segmentation, which relies on the collection and analysis of demographic customer data. Collaborative filtering, or CF, is “the method and process used to match data of one user with data for similar users, based on purchase patterns” (Webopedia). It enables providers to target deals to customers based on actions customers did or did not take in the past, either in the natural course of an engagement or in response to market or social stimuli, removing factors like age, gender, income, and location from the marketing equation. As a result, CF can more reliably yield positive business outcomes.

A fixture in ecommerce for years, CF uses the ‘people who bought x also bought y’ algorithm made famous by Netflix and Amazon. It enables companies to search a large, broad customer base and to find smaller groups within that base containing people who have similar spending habits and preferences. Before collaborative filtering emerged in energy, providers used traditional segmentation to better understand customers: They aimed to predict how customers would respond to certain offers based on demographic data, as people with similar interests are likely to respond in the same way to comparable ads and offers.

Better Than Traditional Segmentation

But there are two big limitations to using traditional segmentation and demographic data to build customer groups. For starters, demographic data may give too broad a view of segments to enable targeted marketing: Not all people who live in a certain area, for example, are going to respond to similar deals in a way that drives revenue for providers. Secondly, such data is less deeply researched in newer markets. While energy is an established sector—not a “newer market”—it has only recently started researching customers, as it only recently needed to: Deregulation enabled customers to switch providers, which inspired companies to better understand the people they serve.

CF avoids the pitfalls of classic segmentation because it makes fewer assumptions about what customers will do based on their demographics. It suggests future customer behavior strictly according to past customer behavior, which is a more unbiased method for forecasting market response. CF reveals probable customer actions based on habits, choices, and preferences customers have demonstrated, not based on inferences attached to people’s ages, neighborhoods, and tax brackets. The difference between classic segment work and CF is this: The former targets x offer to female customers living in New Jersey simply because they are female customers living in New Jersey, and the latter targets x offer to all customers who typically respond well to x offers.

More Revenue, Customer Satisfaction, and Renewals

For providers, collaborative filtering provides new opportunities to increase revenue and improve customer satisfaction. For example, if an energy supplier knows that customers who buy x bundle often purchase y service, it may consider creating a new offer that combines elements of x and y, advertise a discount on y for customers who buy x, or reduce transactional barriers between x and y.

The technique can also give suppliers unique insight into customer churn and energy usage. Say, for instance, a provider learns its retail customers with highly variable energy needs tend to cancel service right before the start of high-usage periods. It is possible, if not probable, that many customers who left took similar actions beforehand—perhaps a sizable portion asked about contract extensions or a specific package. A provider that discovers the actions common to defecting customers can better prepare for churn, intervening in the engagement of dissatisfied customers earlier and thereby increasing renewals.

CF may be the future of energy marketing. Classic segmentation predicts customer behavior using demographic data, which often paints too generic a picture for targeted marketing and may be poorly researched in sectors subject to recent dramatic change (e.g., energy since deregulation). Conversely, CF forecasts customer actions according to their past actions, empowering providers to develop more inclusive deals, offers, and service initiatives.

To learn the next steps for implementing data science techniques like CF at an energy provider, download our white paper “Increasing Customer Renewal in Energy: A 5-Step Plan for Improving Client Experience and Decreasing Churn.”