How Data Pilot improved an energy company’s customer experience
Duration
3 Months Dec’22 to Feb’23
Industry
Government
Services
Advanced Analytics & BI, AI/ML, Data Engineering, Custom Data Solutions
Tools and Technologies
Company Background
K-Electric is a public listed company incorporated in Pakistan in 1913. Privatized in 2005, it has been energizing Karachi for over 100 years. It is the only power utility in Pakistan that handles all aspects of energy management, from generation to transmission and distribution, providing a seamless energy supply to its customers.
The company supplies power to over 3.4 million customers across a network spanning 6,500 square kilometers. This includes all residential, commercial, industrial, and agricultural areas in Karachi, Dhabeji, and Gharo in Sindh, as well as Uthal, Vinder, and Bela in Balochistan.
Challenges
Unable to segment customers
The client’s approach of segmenting its customers based on their payment behaviors was ineffective as it overlooked the diverse circumstances and behaviors of individual customers. For example, it couldn’t differentiate between a consistently punctual bill payer and someone experiencing temporary financial difficulties.
Difficulty in allocating its services to customers
Because they were unable to segment their customers based on their bills, the client could not effectively allocate its electricity services to its customers. This led to discrepancies, such as cutting off the supply of electricity to customers who had already paid or giving more electricity to those who had not paid at all.
Unable to forecast late payments
They were also unable to predict the actions of their customers e.g. who is likely to complete their payment on time, who will delay it, or who will not pay it. As a result, the client faced delayed payments, inefficiencies in targeting at-risk customers, and a lack of proactive resource allocation for collecting debt.
Solution
Data Pilot built a predictive model that provided detailed insights into customer payment behaviors across a dataset of 102.1k customers with a two-year payment history. This model would also forecast the likelihood of customers paying their bills on time by virtue of metrics like dates and segments.
- ETL for consolidating data on customer payments
- Predictive analytics using historical payment data
- Visualization tools for mapping customer payment behaviors
The platform consolidated data from over 102.1k customers and applied advanced ETL (Extract, Transform, Load) technologies to manage the large datasets. After consolidating the data, new columns were added to add more depth to the analyses: ‘Days’ calculated the number of days a payment was made before or after the due date, ‘Paid Type’ categorized payments as “Timely, Late, or Not Paid”, and ‘Min/Max Late Days’ recorded the minimum and maximum number of days late for each payment within the customer segments.
This dataset was used to train the predictive model. The model was then integrated within the company’s internal systems through the Rest API. Through predictive analytics, the model could segment customers by their payment patterns so that the company can identify which customers are more likely to pay on time, pay late, or default. This helped the company implement targeted strategies for effectively collecting revenue and enable an effective means of allocating resources for timely follow-ups and personalized reminders.
Visualization tools were integrated with the model to provide a holistic view of customer behavior, enabling enhanced resource allocation for timely follow-ups and personalized reminders. This helped the model forecast payment behaviors using key metrics (days late, paid type, and customer segments) so that the company can proactively manage customer engagement and improve satisfaction.
Through analytics, 4 customer segments were created:
Stars
Accounts with 90% or more of their payments made on time (e.g., at least 22 out of 24 payments made before the due date).
Potential Stars
Accounts with 80% to 90% timely payments.
Irregular
Accounts with 50% to 80% timely payments.
Defaulters
Accounts with less than 50% timely payments.
The Impact
Significant decrease in collection costs per loan
The solution improved the collection of revenue by significantly decreasing collection costs per loan. It helped provide accurate forecasts of late and default payments, enabling the company to target at-risk customers effectively.
Improved customer communication engagement
By providing segmentation and behavior insights, the model significantly enhanced the client’s customer communication engagement.
Enhanced recovery rates in collecting payments
The solution optimized resource allocation by identifying high-risk customers and allocating follow-up resources accordingly. This led to a more proactive approach in addressing payment delays and defaults, ultimately improving recovery rates and ensuring more effective debt collection.
Industry Applications
Personalize Repayment Plans
The solution can analyze the payment history and financial behavior of its customers. It can predict which loan customers are at risk of not paying back their owed money. The solution can also allow banks to offer personalized repayment plans or alternative payment plans before customers default.
Reduce Customer Churn for Subscription Businesses
Companies offering subscription plans to their customers can employ this solution to maintain a stronger relationship with their subscribers and reduce revenue loss from unpaid fees. By using insight gathered from the solution, it can send timely reminders or offer flexible payment options to prevent customers from canceling their subscription.
Ready for your transformation? Let’s connect
Your transformation starts here.