Expanding financial inclusion with AI analytics

The challenge of lending in opaque markets

In an increasingly customer-centric world, the ability to capture and use consumer insights to shape products, solutions, and the buying experience as a whole is critically important. Even more so with credit scoring:  drawing on diverse sources of consumer behaviour expands financial inclusion to the unbanked, triggering downstream impact in numerous ways.

Take MPower, for example, a European clean energy and fintech startup that provides connectivity and affordable, reliable solar products to off-grid and under-electrified households and SMEs in emerging markets in Africa. The company partners with local entrepreneurs and provides them with sourcing support, state-of-the-art software, and micro-financing solutions. 

However the relatively high up-front costs for these products and software were an obstacle for the investment. Microfinancing can bridge the gap, but any form of efficient financing requires a dependable credit model, which in turn relies on access to financial data including borrowing and repayment history. Yet MPower’s customers are typically unbanked with no recorded credit history and therefore no reliable financial data.  

Sophia Bieri, data scientist at MPower, says that "It is really important to allow people to be able to obtain financing and repay a loan in order to participate in the economy, to build up their wealth on their own."

At this level of financial exclusion there’s simply no traditional way to verify creditworthiness, and for MPower’s customers, it is exclusion that is compounded. The lack of access to power means traders cannot do business without solar-powered tools, but they’re also excluded from getting the financing to buy these tools.

Until recently MPower was unable to provide financing because the data that feeds classic credit scoring models simply does not exist for their target market. The FICO Score for instance, which is the most widely used credit score globally, cannot even be generated for individuals without financial or credit records. MPower, therefore, had to find a way to build a level of customer insight that enables it to lend widely and responsibly. 

That meant constructing a novel credit scoring system, with the use of a wide variety of raw datasets,finding entirely new, meaningful, innovative credit scoring features by searching for signals in complex data. It also needed to do so within a relatively short period and with a compact data science team.

A comprehensive redesign of credit scoring

For MPower, it was clear that they needed to look beyond classic labour-intensive approaches to data science to quickly develop a reliable credit scoring model. It quickly emerged that SparkBeyond Discovery offered key advantages across the data science workflow – including the platform’s ability to join disparate data sets. The modelling capabilities of the platform were also an advantage, and SparkBeyond made it far easier to draw insights and communicate these to stakeholders. According to Sophia Bieri, SparkBeyond offers a “simple and fast way to do all of the steps in data science.”

Using the new Discovery platform, MPower managed to extract unconventional features from a complex data set – combining free text sources interpreted by advanced data analytics, with existing resources such as OpenStreetMap. This increased understanding of its customer base enabled MPower to make lending decisions even where it had no access to traditional financial data.

What’s more, SparkBeyond’s embedded methodology built into the platform allowed MPower to refine the credit application questionnaire and gain more meaningful data for credit decisions. This process improved MPower’s data science metrics – including the area under the curve (AUC), which indicated that the company’s credit scoring model was becoming increasingly accurate.

It means that MPower can now offer financing to a much broader range of applicants while reducing its risk exposure, and by consequence reducing exposure to lending losses.

Mining for patterns buried deep in raw data

SparkBeyond’s platform assisted MPower in feature engineering by harnessing text mining to find patterns in written submissions, such as crowdfunding loan listings  – searching for features in raw text data.

For example, MPower asked its credit applicants to describe what their line of business is. Using natural language processing, SparkBeyond Discovery was able to automatically cluster applicants into groups dependent on the sector and type of their business. Distinguishing between applicants that produce goods and applicants that resell goods, for example.

The platform’s text mining capabilities also delivered powerful insights based on the applicant’s personal, as well as social networking characteristics – building a behavioral and reputation profile, allowing for an assessment of the individual’s willingness to repay,  rather than solely the ability to repay ascertained from financial data. This 360-degree, holistic view of borrowers is one of the cornerstones of fair, smart credit scoring. Using the Discovery platform, MPower redefined the way it viewed the risk profile of applicants where little hard data was available. That enabled MPower to increase its addressable market because the company was now able to understand risk, and forecast risk accurately.

Improved customer understanding, expanded inclusion

Through the remainder of 2021, MPower intends to roll out a wider range of solar-powered products, with increasingly accessible borrowing facilities.

MPower’s improved understanding of its customers now means that it is in a much better position to evaluate prospective customers for creditworthiness. 

In turn, access to credit supports wealth creation for a population who were previously excluded from access to financing. In the long run, MPower’s novel credit scoring model ensures that a growing number of people can participate in the local economy.

Features

No items found.
No items found.

It was easier in this project since we used this outpout

Business Insights

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Predictive Models

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Micro-Segments

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Features For
External Models

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Business Automation
Rules

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Root-Cause
Analysis

Apply key dataset transformations through no/low-code workflows to clean, prep, and scope your datasets as needed for analysis

Join our event about this topic today.

Learn all about the SparkBeyond mission and product vision.

RVSP
Arrow