Mutual Finance Operation | How to build a big data risk control system from 0 to 1

Risk control is the heart of finance, and data is the blood of risk control. In the past, we mainly relied on experience and macroeconomic forms to implement risk control. Later, we conducted risk control through data and scoring. Now we are using big data to do risk control.

Since 2014, risk events have begun to erupt on a large scale in the Internet finance industry, and many financial institutions have begun to question the traditional risk control model. At this time, big data technology is developing rapidly and is gaining favor in the Internet finance industry with its reliable risk control technology. According to the "2018 China Big Data Risk Control Research Report", the scale of China's big data risk control market reached 14 billion yuan in 2017.

At present, major Internet financial companies have adopted the technical means of big data risk control. Ant Financial, Rong360, Paipaidai, Dianrong.com and others have developed independent big data risk control systems. Big data is an inevitable trend in Internet finance and even traditional financial risk control. Its development will bring great benefits to the financial field.

So what exactly is big data risk control? How to build big data risk control? Below, Box Fungus will answer them one by one:

1. What is Big Data Risk Control?

Big data risk control refers to establishing a risk model through big data core algorithms. Based on the collection of data from various dimensions, it combines Internet-based scoring and credit management models to extract data that is useful to the enterprise, and then conducts analysis and judgment to ultimately achieve the purpose of risk control and risk warning.

Big data risk control is a new idea for Internet finance platforms to innovate credit management and risk management. Compared with traditional risk control, big data risk control has no essential difference in modeling principles and methodology. It just utilizes the characteristics of the Internet era. Currently, the leading big data risk control still uses small data. It focuses on customer information and evaluates the customer's credit level from multiple dimensions such as property, safety, contract compliance, consumption, and social interaction. It also establishes customer credit data for them, thereby reducing the source of risk.

2. Why use big data risk control?

According to statistics, the traditional risk control model of banks is effective for 70% of customers in the market, but for the other 30% of customers, the effectiveness of the risk control model will be greatly reduced. The advent of the big data era has enriched the data dimensions of traditional risk control, using multi-dimensional data to identify borrower risks, including social, credit, consumption, interests, etc. The more customer data there is, the more fully the credit risk will be revealed and the more objective the credit score will be. The data dimension in risk control in big data can serve as an effective supplement to the other 30% of customer risk control

The role of big data risk control is to find qualified customers from the originally rejected customers and to identify high-risk customers and fraudulent customers who have passed the review. It can greatly improve the efficiency and risk control capabilities of the Internet finance industry, effectively control the bad debt rate, and thus make companies profitable. Big data risk control is a technological means that must be integrated into the development of the financial industry.

3. Application Scenarios of Big Data Risk Control

The application scenarios of big data risk control models are very broad. As long as the industry involves Internet finance, big data risk control is indispensable. From a financial perspective, the risk control model is designed to assess customers' ability and willingness to repay, combat fraud and cheating, prevent customers from taking advantage of the platform, and ensure platform security.

From an industry perspective, it mainly includes consumer finance, supply chain finance, credit lending, P2P, big data credit reporting, third-party payment (fourth-party aggregate payment) and other sub-sectors. It can also be used by "traditional" Internet companies such as e-commerce, games, and social networking. It can be said that any Internet company needs risk control.

4. Three Steps to Building a Big Data Risk Control Model

Big data risk control covers everything from customer acquisition, approval, maintenance during loan process, enhancement of customer value, reuse, in-depth exploration, as well as customer retention, collection and exit. Financial risk control is no longer simply about lending and collecting money, but about completely maintaining a customer’s life cycle.

If we want to form a complete closed loop, we need to divide it into three steps: pre-loan, mid-loan, and post-loan.

1. Before the loan: Look at its face to draw its shape

Pre-loan mainly includes the formulation and division of access credit rules.

(1) Access

In the pre-loan stage, customer data needs to be collected, cleaned, analyzed, and applied. This is a very long chain, and using traditional risk control is time-consuming and labor-intensive. But now we have big data technology that can accurately mine multi-dimensional information about applicants, including demographic information, social information, historical consumption records, consumption patterns, interests and hobbies, social preferences and other related dimensional information. This information is combined to form a user portrait to judge the customer's loan qualifications, repayment willingness and repayment ability, and assist in review and decision-making. Unqualified customers are directly blocked at this stage, which not only prevents "accidental killing" in the later stage, but also ensures the customer quality of the platform and achieves twice the result with half the effort.

(2) Credit

Credit is granted based on the consumer demand of a platform. We can use consumer demand as a base to establish effective credit models and scoring rules. By utilizing flexible and open data import technology, multi-dimensional credit strength and weakness relationship scoring items, and professional rating models, we conduct a more in-depth and comprehensive "dissection and analysis" of customers' repayment ability and willingness, and make an overall rating for the platform's credit decision. Customers with different ratings: First, the risk factor is adjusted differently; second, there must be an upper and lower limit on the credit limit for each customer group with each rating.

2. Loan: From the outside to the inside, prescribe the right medicine

The loan is divided into two parts, one is anti-fraud and the other is credit limit adjustment.

(1) Anti-fraud

Anti-fraud may be used by many people more often before taking out a loan. But in fact, anti-fraud runs through the entire customer life cycle. Anti-fraud protection is required not only in the credit link, but also in the account login and registration links. The current fraud methods mainly include fraud under someone else's name, intentional fraud by the individual, and unscrupulous intermediaries deceiving others to indirectly commit credit fraud.

Anti-fraud requires two things: information verification and behavioral analysis. In the big data risk control system, there are many advanced technologies to support it, so we don’t need to worry too much about this aspect. In the process of behavioral analysis, we rely on risk control experience, customer information verification, and partial behavioral data for predictive analysis. Based on customer behavior, we identify the risk levels of different customer groups through labeling.

(2) Credit limit adjustment

At this stage, most customers have made at least one repayment, so the platform needs to consider how to adjust the customer's credit limit level and interest rates to ensure that high-quality customers get lower interest rates and higher credit limits, while customers with poor data performance need to use higher interest rates to cover risks.

However, it is meaningless to blindly pursue high returns without considering the risks, or to pursue low risks without seeking returns. The focus of credit limit adjustment is on reasonable estimation of customer needs and risks. In fact, it can be seen as an allocation of funds among different risk returns, so that the overall risk return can be maximized under a certain risk.

3. Post-credit: Listen to the five tones to identify the disease

Post-loan work mainly involves bill collection and post-loan monitoring.

(1) Bill collection

When the platform releases funds, it must ensure that they can be recovered, so it is necessary to track the funds. Once an overdue payment occurs, the collection team will be activated to assist in completing the overdue processing and asset recovery work. There are also certain strategies for debt collection. First of all, differentiated debt collection measures should be formulated for different risk segments of customer groups. Secondly, grasp the timing of debt collection. Because the resources for debt collection are limited, we need to allocate debt collection resources according to certain allocation rules.

(2) Post-loan monitoring

Finally, we enter the post-loan monitoring stage. In the credit process, even if risk control is in place in the early and middle stages, it does not mean that the credit transaction is foolproof. Changes in the borrower's environment, changes in repayment ability, and changes in repayment willingness often occur. By utilizing big data technology, we can track and monitor multi-dimensional dynamic events and market information of borrowers, quickly detect and discover abnormal data of borrowers after the loan, and issue timely post-loan warnings, effectively preventing the lender from running away and the occurrence of bad debts and dead debts of credit institutions.

V. Conclusion

The big data risk control system that monitors the entire process from pre-loan, mid-loan and post-loan can effectively control financial risks. However, I would like to remind everyone here that the construction of the risk control system must be based on one's own business in order to play its real role.

Source: Activity Box