Predictive Analytics in the Real World: Utilities and the Pandemic
Using predictive analytics can help utilities safeguard revenues and protect at-risk customers.
In the age of coronavirus, when millions of Americans have lost income due to business closures and can’t pay their bills, utility companies are worried. They have always worried about their most vulnerable customers, but now they are also worrying about a new group of people: middle-income earners who have lost their jobs, are suddenly unable to pay their energy bills, and have begun carrying high balances.
Analysis of one mid-Atlantic gas company’s most recent customer data shows that customers will owe well over $20M this year — revenue the utility may never see.
This means that when state-issued shut-off moratoriums are lifted, more customers than ever before could be sent to collections and potentially have their power or water disconnected. Shut-offs are obviously bad for customers; research shows that energy insecurity and power outages lead to poor mental and physical health outcomes, sending people deeper into risk.
Shut-offs are bad for utilities, too. They are expensive, laborious, and unpleasant. If company budgets fall short, utilities may not be equipped to deliver the same robust level of service to the hundreds of thousands of customers who depend on them. According to some experts, lost revenues could cause utility workers to lose their jobs and energy rates around the country to spike.
Luckily, utilities have a way out of this potential bill-payment catastrophe: predictive analytics, which is being deployed to protect utility revenues and build safeguards around vulnerable customers.
BlastPoint — the company I founded — uses household-level historical payment data to establish a benchmark that compares past customer payment data to current customer payment behavior. This enables us to distinguish customers at the household level who have suddenly bumped up against a wall of financial hardship from those who have been falling off a financial cliff for some time.
We use AI, machine learning, and data algorithms to help us quantify territory-specific dollar thresholds which reveal those customers most likely to continue paying in full and on time each month; those who will be able to pay in part but carry a balance from one month to the next, and those who are unlikely to be able to recover from a growing stack of overdue bills and late charges.
We call these specified amount-due thresholds customer balance risk zones. Depending on a utility’s unique zone limits, Green indicates a customer is financially safe, i.e., in a position to pay current balances even if they are a little behind. Yellow signals that a customer has entered a period of financial hardship and needs attention now. Red denotes that a customer is in financial danger.
Image (c) BlastPoint, Inc. 2020. Used with permission.
With Customer Balance Risk Zones, utilities have clear indicators that can help them bring as many customers as possible to safety. However, getting there requires a new way of thinking.
Descriptive sorting, or customer segmentation, has been used by businesses of all types since the early days of the 20th century. Typically, segmentation uses company data to place customers into different buckets, and companies can handle those buckets differently depending on the situation.
