How do Internet financial products use big data for risk control?

How do Internet financial products use big data for risk control?

There are too many Internet financial products. Here we take P2P online lending as an example and try to discuss the relevant aspects involved in risk control.

Sales link:
Understanding the customer's application intention and the authenticity of the application information is applicable to the loan officer model. The key points of risk control are:
Meet the applicant in person, see the applicant's ID, see the applicant's signature, and meet the applicant's employer in person. Approval process:

 The verification of basic credit policies mainly focuses on verifying application information, certificate information, and whether the application is fake.
The system will review and eliminate customers who do not meet the basic credit policy requirements, such as those with serious bad collection records, internal default records, or those who have been included in the associated blacklist due to high risks in the recent period, and those who do not meet the regulatory policy requirements. After basic review, different applicants will be automatically distributed to different credit processes based on the classification of customer information. These different processes are generally designed based on factors such as customer classification, application amount, whether the customer is a new customer, whether the customer is an existing customer, etc., and then enter the specific review stage. The review process will consist of two interactive parts: system review and manual verification. Only those who fail the review, have questions, or pass can enter the subsequent stages, including rejection, return for additional investigation, return for additional information, pass, conditional pass, etc.
Applicable to the credit factory model, key risk control points:
1. Logical verification of information submitted by customers.
The information filled in by the customer includes the declaration information filled in on the application form and the information in the qualification certification documents provided. Since fraudulent customers fabricate all or part of the information, there is a high possibility that the relevant information they self-report may contain unreasonable situations. By utilizing the location service provided by Internet big data, the address information filled in by the customer can be located as the address location coordinates and compared with the customer's commonly used logistics address location coordinates. If it is found that the customer has provided an address that is too far away, there is a possibility that the address information is false. For mobile channels, such as PAD, the specific application location of the Internet customer can be located and compared with the address information or occupational information filled in the application information. Behavioral information of the customer's application process can also be collected, such as how long it took to fill in the form, how many times it was modified, and what content was modified. These information items can become variables in the application fraud model or important rules of the application fraud strategy.
2. Logical verification of customer reported information and the company’s own existing customer information.
For example, if the unit phone number filled in on multiple applications is the same, but the corresponding unit name and address are different, the possibility of batch counterfeit applications is very high.
3. Comparative verification of external information.
Malicious applications often conceal facts that are unfavorable to them, such as debts, problems in company operations , court enforcement information awaiting processing, etc. By crawling the applicant's business information and court enforcement information on the Internet, the applicant's true qualifications can be verified. Credit granting process:
Customers who enter the scoring rule engine will go to different segmentation modules according to their types to adapt to different segmentation models, including different products, different industries, and different customer groups, such as car loans, consumer loans, mortgage loans, personal business loans, etc.
Key points of risk control:
Different types of loan applications call for different credit scoring rule engines.
Data automatically captured based on user authorization:
Personal information, captures data from multiple dimensions such as e-commerce purchase data, search engine data, social data (Weibo/Renren, etc.), credit card bill email information, and China Higher Education Student Information and Career Center information left by users on the Internet, and obtains personal information about personal character, consumption preferences, intentions, academic qualifications, etc.
Merchant information: Capture merchants' e-commerce transaction data (logistics, cash flow, information flow data), e-commerce business data (such as visitor volume, transaction volume, user reviews, logistics information, etc.).
Finally, it is converted into personal credit scoring data and merchant credit scoring data through a specific model. Appendix: Big data credit data source map
Post-loan stock customer management link:

 Credit adjustment for existing customers is an important part of existing customer management. Various business methods will eventually involve the adjustment of credit customers. Failure to pay attention to the management of credit limits is likely to cause a rapid increase in risks, and it is also possible to turn "good customers" on the introduction end into "bad customers" on the existing end.
Key points of risk control:
1. Observe the default situation, such as whether there is early overdue payment, failure to pay the arrears for multiple consecutive periods, invalid contact information, etc.
2. Check the information association, such as whether there are any existing customers that match the newly added blacklist or graylist data.
In the process of stock risk management of small and micro merchants, merchant transaction flow information can be obtained from data partners, and their transaction flows can be monitored and warned. Warnings can be triggered for sudden inflows and outflows of funds, declines in transaction flows that do not comply with business rules, and large transactions in normal operations. Through real-time monitoring of big data, once serious negative information about customers, public security law violation information, court execution information, tax payment information, important industry news, major negative situations in the borrower's social network, the borrower's online browsing behavior, fund payment and settlement conditions, etc. are discovered during the external data monitoring process, warnings can be triggered in a timely manner. Post-loan overdue customer management links:

 Poor willingness to repay and insufficient repayment ability are the main reasons for customer overdue payments. This link mainly involves overdue customer management and lost customer management.
Key points of risk control:
1. Optimization of collection models and strategies. The different responses of different customers to different collection methods can be mined through big data. For example, for a customer who has almost no Internet history, sending emails for debt collection generally fails to achieve the desired effect, and voice reminders may be more effective; for a customer who is addicted to Weibo or Zhihu, the same debt collection and repayment reminder text will be more effective when sent via Weibo or Zhihu private messages rather than via mobile phone text messages.
2. Identify lost customers and restore their information. For example, by cross-checking the customer's logistics information with external e-commerce companies, it is found that the contact information provided by the customer when applying for a loan does not match the contact information used in recent online shopping. This may mean that the customer has updated his contact information. At this time, you can proactively initiate communication and contact with the customer to avoid the occurrence of customer loss of contact. For lost customers, the large amount of related information accumulated on the Internet can help to understand the customer's work, life, and social network. Fund liquidity management
Liquidity risk is the main risk of P2P online lending platforms . An important reason for the disappearance of P2P online lending platforms is bank runs. Liquidity management under big data is actually an application of real-time BI. Traditional BI data is T+1, while big data is real-time BI.
Key points of risk control:
Integrate all the data from both the borrowing and investment ends of the platform and analyze from the following two dimensions
1. Funding Dimension
2. Business dimension For more details, see a previous answer
What is the most important indicator of liquidity on a P2P platform? Loan issuance

 The loan issuance process is a key step in preventing account takeover and misappropriation of funds.
Key points of risk control:
Designated account fund transfer and targeted payment. For example, if a customer borrows money to pay for tuition fees for training or further studies, the customer will be required to provide the school's relevant account number in advance during the application process. summary
P2P big data risk control model construction path
-END-

Mobile application product promotion service: APP promotion service Qinggua Media information flow

The author of this article @刘渝民is compiled and published by (APP Top Promotion). Please indicate the author information and source when reprinting!

<<:  Content information mini program qualifications, what basic functions are needed to develop a news information mini program?

>>:  Clothing wholesale WeChat applet, how to develop clothing WeChat applet?

Recommend

Product Operation: Is your product suitable for fission?

Regarding fission, there have been several questi...

[Li Jiaoshou] Why does it get worse the more we learn marketing from big brands?

"A very strange phenomenon is that we like t...

Share Bilibili’s regular money-making project, and monetize by uploading videos in batches!

I have shared with you several projects of Bilibi...

How much does it cost to develop a Nantong flower mini program? Nantong flower applet development price inquiry

How much does it cost to develop the Nantong Flow...

What are the tips for maintaining a Douyin account? The correct way to maintain a Douyin account

This article mainly introduces the skills of main...

How much does it cost to develop a Puyang fitness mini program? What is the price quote for Puyang fitness mini program development?

There is no doubt that the topic of mini programs...

How to make users addicted to your product?

Why do users keep using some products (including ...

Be careful when reading long and informative articles! ASO optimization experience: Tips for APP keyword optimization

For your product, choosing the right keywords for...

How can community operations be converted from 0 to thousands of users?

Why is no one joining my community? Why can't...

How to plan a fission event promotion?

Hello everyone, I believe that every operator who...

An analysis of the entire process of Keep’s customer acquisition: How to gain 140 million users in 3 years

In October 2017, Keep surpassed Yuedongquan Runni...

Is the effect of bidding promotion declining? You must analyze these 8 factors!

Nowadays, most of our SEM promotions revolve arou...

Three steps to teach you how to improve personal and team competitiveness!

In the past, human resources, especially human re...

After reading 300 fission cases, I want to share with you the traffic acquisition thinking

This article will help you change your taste. It ...

Useful Information | Learn Pinduoduo’s 5 high-conversion marketing strategies!

Pinduoduo, a new internet celebrity and a three-y...