B2B Energy – Behavioural Predictive Model

Application Track:

Code:

Domain:

Proposed by:

Entity Logo:

Summary of the entity:

EDP Group operates in different markets, in this challenge we are focused on the liberalized energy retail market, through the sale of energy and energy efficiency services to B2B customers. 

In this business area, customers are segmented according to their energy consumption, being divided into different categories. The SMEs management segment is responsible for the management of the 3 bottom segments of the B2B pyramid.

Summary of the challenge:

The challenge’s main goal is to build an analytical model which predicts the willingness to buy energy or energy efficiency services, clustering the portfolio customers, comparing and predicting behaviours such as churn. Through the model’s clusters we would be able to have a dynamic segmentation through which we would have the ability to define the best course of action based on the 2 main predicted behaviours:

• Willingness to celebrate an energy or services contract
• Churn Probability

Description:

EDP Commercial different business areas gather customer information in diverse fields, such as complaints, information and transaction requests, billing, debt, acquisition and switching out, energy and services opportunities, proposals and contracts.

Although all this data is in different systems, we gather it all in Celonis and other databases. We have a lot of information about our customers, but at the moment we don’t use it in an aggregated way to get useful inputs and insights to our operation.

The objective would be to create a predictive model for acquisition of energy & services and also a churn probability. One of the hypotheses would be to create customer clusters and when a customer fits into a cluster we would be able to predict some future decisions considering the typical behaviour of that cluster. With this clustering of our client base we would be able to perform a dynamic segmentation and get insights based on the predictive model to decide the future next best actions to prevent or promote a certain behaviour.

Data:

MarketInfo – Information about all electricity consumption points in Portugal

Contracts – Detailed information about EDP Commercial contracts

Billing – Detailed information of all EDP Commercial customer bills

Debt – Detailed information about all EDP Commercial debt. Both historic and current debt

Requests (Information and Transaction) and Complaints – Detailed information about all customer requests and complaints

Churn – Historical Data about switching out (churn) requests

Energy & Services Opportunities – Detailed information about all opportunities (customer approach/lead)

Energy & Services Proposals – Detailed information about all proposals presented to our customers and their decision

Pricing – Market price historical evolution

Expected outcomes:

• To create dynamic segmentation
• To give place to an Energy and Services predictive acquisition model
• To devise Churn Predictive Model