By Maria Veikhman – Head Of Risk at IT Smart Finance
‘The non–banking financial market is growing rapidly and we know that many institutions have managed to position themselves in the digital finance sector. But if we are aware of anything at IT Smart Finance and our consumer investment platform Nibble.finance, it is that as the market grows, so do the risks,’ argues Maria Veikham.
One of the main risks is loan repayment. A high quality score helps prevent and combat this, as the purpose is to technologically evaluate the borrower. The scoring system is designed to determine if the customer will return a loan with certain parameters of profitability and delay. And the more precisely the external and internal scoring model is configured, the greater the efficiency of the entire business model of a microfinance company and the better the returns and security for investors.
Taking into account the realities of today focusing on the low financial education of the population, possible new measures are being developed to reduce the rate of PDL loans, the high quality rating becomes more relevant than ever and two tasks stand out mutually excluding: the most detailed and accurate analysis of the borrower, and instant decision making.
Modern scoring technologies allow solving the problem of the effective evaluation of a microfinance institution borrower. When constructing a scoring model, more than 10,000 different variables can be taken into account, obtained from a wide variety of sources, which should have the greatest predictable power. There are so-called factors in the scoring model, that is, unprocessed data that must be converted into variables, and there are also complex variables or complex factors that are obtained by calculating or adding several simple variables.
Western scoring systems analyze information from a variety of sources available through APIs, geographic services data, delivery services, statistical web services (such as AppStore, GooglePlay, Google Analytics, Flurry, Localytics, etc.), mobile operators, official data from tax authorities, cloud accounting files (Intuit, Sage, Debitoor, FreeAgent, Xero), social media pages, and this list is far from complete. Of course, no one uses all this data within a model, but it is their combinations that constitute the knowledge of each particular system. The more successful this combination is, the more accurate the borrower’s portrait is, resulting in a lower percentage of defaults.
For example, the German Bintbond platform, when evaluating a borrower, uses the personal account data of its online services: before a loan application is accepted, the borrower must give access to at least two of their profiles: PayPal, eBay, Amazon, MercadoLiber, Google Analytics, Debitoor (online accounting) or in your personal online bank account. So does the British Funding Circle.
Sometimes revolutionary scoring decisions are extremely unsuccessful. For example, the Quakle platform, founded in 2010, decided to use the solution developed on eBay to evaluate the borrower, a personal rating based on the comments of the counterparts. The model turned out to be wrong: a year later, Quakle ended its existence with 100% default.
To date, the Applications Approval Rate is increasing in the best accurately way, aiming to decent market figures helping us to receive a respectable percentage of the total number of people who wish to obtain a loan online. We operate using our own scoring technologies and those of third parties. Our own score is based on the data of our customers that they provide.
This is a very complex verification algorithm that is constantly being improved. Our software allows us to quickly and efficiently integrate new verification methods into the borrower’s evaluation algorithm, which are based on social network data, payment aggregators and geolocations. Therefore, we obtain a complex and quite high quality sociodemographic model. But this is not a 100% factor in deciding whether to grant a loan.
Internal scoring is complemented by external scoring technologies. For many years, our partner has been Scorista, which improves the quality of the credit risk assessment of Joy Money and participates in the generation of profitability and delay forecasts. The online service chosen by us as a partner specializes in evaluating borrowers for Micro Financial Institutions, has the highest predictive power in the market and helps to make loan decisions in less than 1 minute, directly helping us solve the most important factors. determinants at the time of granting a loan.
As part of the IT Smart Finance group, we have created the Nibble.finance platform to give consumers access to this great scoring technology and a way to invest in Joy Money loans.
The unique developments in the field of scoring technologies allowed Scorista to develop individual predictive models for Joy Money that analyze borrowers according to three key criteria: probability of delay, probability of fraud and degree of loyalty. In addition, the company offers recommendations to the subscribers of said platform for the verification of the borrowers. Nibble.finance is an extension of this technology and allows investors to benefit from the enhanced security and protection offered by this scoring system.
The constancy of the high commercial indicators of Joy Money is achieved through continuous monitoring and analysis of the application flow, both by the Scorista system and by internal scoring technologies. The online service monitors any abnormality in automatic mode and immediately notices it. The response time to the flow changes is a record <59 minutes.
Such a combination of internal and external scoring technologies gives Joy Money and Nibble significant advantages: it allows you to keep the risks controlled within the framework and at the same time receive high income: the level of timely repayment of loans by customers is now more than 62% The level of collection of overdue debt is also consistently high: more than 75% of overdue debt is actually collected monthly, which guarantees high profitability of the business.