Data Science Navigating Finance

Kirill Odintsov, Head of Data Science, Home Credit Indonesia

Kirill Odintsov, Head of Data Science, Home Credit Indonesia

Kirill Odintsov is the Head of Risk Underwriting and Data analytics at Home Credit, Indonesia. With over five years of experience in managing and building data science teams, Mr Odinstov is a prodigy in understanding the implementation of data science across the domain. Currently, Mr. Odinstov is adapting to his new role as Credit Risk management through which he will be able to walk in the client’s shoes.

1. What are the primary challenges that replete APACs financial services industry?

Big challenge is getting the right data about customer. We want to make our approval process seamless and quick without asking customer to fill in too many data fields or provide us with too many documents, but at the same time we need to understand our clients’. We need to know who they are, what do they need and their ability to repay a loan. We want to provide services relevant for our customers and ensure we are not over indebting them. Getting the right data becomes especially challenging for underbanked population about whom much less standard information is available.

2. How can financial service providers leverage data science and machine learning to navigate the dynamic market scenarios and serve their customers better?

By utilizing non-standard data and advance Machine learning models we can understand client much better without inconvenience for client of having to go through long and manual processes. Understanding the customers better allows us to better estimate their risk profile and offer fair pricing (lower risk – lower price).

Data Science helps us to understand which clients are currently interested in our service and which are not, so we don’t bother those who are currently not interested with communication. It can help us understand what product to offer, using what communication medium (call, email, SMS) and what communication style suits the client the most (details, to point communication, ….). We receive many free text written feedback from the customers, using NLP and Text Mining models like Bert, Word2Vec, LDA we can go through all this data quickly and adjust our services based on the customer feedback or we can react to emails by urgency of its’ content. Working with the alternative data and the free text feedback also helps us understand what our competitors are doing and properly react.

3. You have had hands-on experience on a wide range of analytical and modelling projects. It is also critical to note that you employ the Pareto principle while observing the latest trends in data science and analytics. How does Home Credit Indonesia benefit from your expertise?

Personally I supervised and even build myself models on many markets (China, India, Indonesia, Russia, Vietnam , USA,…), on many kinds of data (Standard banking data, Telco, Ecommerce, Mobile app behavior, …) and for many departments (RISK, CRM, Marketing, Custex, HR, …). All this experience helps me to find opportunities for new projects and to find weak points in Data Science projects to look out for. It helped me build strong Data Science team with experts who don’t simply run scripts in Python but also understand the business benefit of the models and focus on implementable realistic models.

Working with the alternative data and the free text feedback also helps us understand the industry and how we can improve to deliver better products & services amid the competition

4. What is the one piece of advice you would give to market players and budding entrepreneurs?

Data Science is not a magical wand. It can’t solve everything. You need to have good quality data and huge internal support from other teams to create truly impactful data science projects. Good data science team is not the one that uses the most advanced methods it is the one that uses the simplest possible methods to solve all the tasks and goals set by the business. For those starting with data science in their organization I would recommend starting small by collecting ‘low hanging fruits’ with simple interpretable methods. Only after this foundation is built business and data science teams gets mature enough to move to more advanced “cool” methods. In my experience people sometimes tend to mask misunderstanding of business or data by using overcomplicated methods.