The conference was treated to an exclusive peek at analytical modelling recently conducted for a UK utilities portfolio.
Presenting the model, Panayis Fourniotis Pavlatos, director of intelligence decisions for Qualco, argued that automated analytic modelling offers collections teams unprecedented insight into every part of their day to day operations.
“Automated analytics can impact every business decision, not just a few strategic ones,” he said.
Emphasising the need to identify decision points and ensure quality of data, Pavlatos highlighted three type of analytics: descriptive, predictive and prescriptive. Descriptive models explore the data and sort it for humans to use, predictive analytics identify patterns and suggest what may occur, while prescriptive models decide that action to take.
“Analytics on their own cannot determine causality,” he said. “They work hand in hand with your business.”
Large scale automation can eliminate manual processes, maintain and enforce best analytics practices and encode and enforce business best practices. It reduces costs and allows businesses to use analytics everywhere.
“That is what is revolutionary about modern analytics. Having ten average models to support ten decision points is better than having one perfect model to support one of ten decision points,” said Pavlatos. “That will make a huge difference in your day to day operations.”
He urged collectors to identify decision points and questions that can improve those decisions. Once the question is stated in a quantifiable manner and the analytics needed are identified, the model can work its magic. “Close the loop by using the outputs in day to day operations, taking action based on the outcomes,” he said.
Pavlatos also covered ways of navigating the contradictions analytics throw up. “Analytics can only tell you what you already know but don’t know you know,” he said. “They assume the future is like the past but seek to change the future.”
While analytics offer no crystal balls and may challenge common sense, their potential to improve collections is significant.
In the demonstration, the model broke down the utilities portfolio into segments, continuing its interrogation of the data until it identified useful segments. It used account characteristics as well as the performance of individual DCAs to determine the likely outcomes for accounts.
“It means you can focus on the best performing customers and consider alternative approaches for the worst performing ones.”