Data scientist with exceptional banking industry big data experience and consumer side analysis
Technological Institute of the Philippines
Current role has an emphasis on handling projects related to consumer finance, retail banking and customer experience with responsibility to identify business challenges and opportunities through analytics. Implements solutions to drive an enhanced customer experience.
Improving Salary Loan Product using Data Science:
For years, the salary loan product of my company was never been highlighted as a major source of revenue. Simply because we are targeting the wrong customers.
When we formed the data science team, I was assigned to help improve the salary loans product. We have millions of customers all over the Philippines, we know how much money they have or how much they earn, where they live, and how do they spend their money, etc. And we asked ourselves, among these people, who needs the product most?
Using simple statistical methods, it was obvious that a specific segment of our customers needs the product. So we re-launched the product designed for this segment. To be specific, the segment is consist of payroll customers of the bank.
We started with almost nothing. So to test the product, we developed a basic targeting criteria based on previous experiences and bank policies. As we get more bookings, we have learned a lot of things from our customers simply by just using the data that we’ve collected. For example, with the help of the data, we found out drivers of delinquency and one interesting that I saw was that customers who work in the BPO and Fast Food Industry are usually delinquent in payments. With the data that we have, we were able to improve the targeting criteria by excluding these people so as to make sure that we are offering the product to the right customers who will pay for their loans.
After 6 months, the business was not happy because we are producing a very small number of leads for them. I started to develop the attrition model. With this model, we were able to predict the probability of a customer leaving their job in the next 12 months. The business were happy with the validation results and so we started implementing this to help improve our targeting criteria to get more leads. It made our targeting criteria even simpler because we were able to remove certain parameters because of it. For example, prior to the implementation of the model, we only offer the product to people whose age is between 21 and 65 years old. Also we exclude BPO employees and people who works in the food industry. With the help of the attrition model, we removed these parameters and instead we used the attrition score (as proxy for risk score) for risk-based pricing therefore extending the offer to more customers but with the right price based on their attrition score. In short, the model was used for targeting and risk-based pricing.
Now we have so much leads. However, we have problems of engaging that huge number of people so that they know that they have an active offer from us. One reason for that is we can’t reach them using their contact information. Another is, using channels like SMS and Tele-Sales to reach them are very expensive. So I developed a propensity model for this. With this model, I can predict the likelihood of a customer getting a salary loan. The business was happy with the validation results so we implemented this. Now we are using the propensity score to prioritise the leads. Tele-sales can prioritise customers who are more likely to get or need the product. We now only send SMS to those who will more likely to get a loan.
However, we want everyone to know that they have an offer for them. So we wanted to maximise other cheap or free channels like email notifications, app notifications, and ATM. The latter made the business curious when I said to them the idea that, these are payroll customers and they use the machines regularly. Therefore it’s the best channel to communicate with them. But currently, the bank is working on migrating to a new switch so we cannot request for software enhancements for the moment. We don’t want to wait. So using the ATM data that we have, I analysed where do our qualified leads usually use go to transact with the machines. Using visualisation, I mapped out where exactly in the Philippines our customers go for ATM transactions. Using this methodology, the business were able to focus their marketing strategy on specific hot locations by just looking at the visualisation tool that I developed for them.
With the rich data that we have for this product, we constantly update our models and can easily monitor drivers of delinquency therefore maintaining our delinquency rate at the minimum and booking more customers bringing more money for the company.