The low penetration rate of auto finance, the younger car buyers and the popularity of the concept of advanced consumption, coupled with the favor of capital and the blessing of the Internet, have made new auto retail and new finance quickly become new outlets in the field of Internet finance.

Then the question is coming! How can auto finance enterprises or platforms be in an invincible position under the new retail and new financial model, and how can they use big data technology to establish their core competitiveness? This paper mainly aims at three modes and three main risks of new automobile retail, and puts forward how to solve the risks through big data.

? 0 1 data-driven new car retail model?

Whether the business model is good or not and its competitiveness is not strong depends on whether it can effectively solve the pain points of users. Then, in the scene of buying a car, users' most concerned issues can be summarized into two: one is whether they can borrow money to buy a car; One is whether you can provide a suitable product with fair interest rate when you can borrow money. In order to solve the above problems, auto finance companies or platforms can use big data to make efforts in the following aspects.

1.？ Establish a pure online automatic credit system

Integrate business data and Internet big data, and use advanced machine learning technology to create a pure online automated auto finance big data risk control system, thus changing the way that traditional auto finance relies on manual review of offline submission materials and giving users a better car purchase experience. At the same time, focusing on credit, you can also lock in a potential car purchase user in advance, because a user who is willing to submit information online for credit is definitely more clear than a user who just browses the models on the APP.

2.？ User-centered credit granting

User-centered credit is equivalent to giving users a credit card in the automotive field on your platform. As long as it does not exceed the relevant quota and service life, you can buy any car on the platform. In other words, credit weakens the interference to factors such as vehicle models and car prices, thus greatly improving the user experience. Imagine how bad the experience would be if you had to go through the approval process again every time you changed models or changed car prices.

3.？ Dynamic risk pricing, providing flexible financial support programs.

Internet is to eliminate information asymmetry. If users are not professional enough, using information asymmetry to match products that are expensive for users or beneficial to the platform will kill the goose to get the egg sooner or later, which will not pay off. According to the actual situation of users, using data to carry out dynamic risk pricing strategy and provide financial supporting schemes suitable for users will help improve users' loyalty and stickiness to the platform and be more conducive to the long-term development of enterprises.

? What are the main risks of new auto finance?

In China's auto finance business, there are mainly four participants: commercial banks, auto finance companies, financial leasing companies and internet auto finance platforms. Among them, banks and auto financing companies are undoubtedly the most mainstream participants. Banks have capital advantages, and auto financing companies are often auto manufacturers or dealers with channel advantages, which have occupied 95% market share.

As representatives of new auto finance, financial leasing companies and internet auto finance platforms have chosen the road of differentiated competition: the channels are sinking, aiming at the people that banks and auto finance companies can't cover, and the products and services are deepening. At present, all enterprises are basically aimed at young people who are below the second or third line or even lack funds to buy cars in rural areas. They have advanced consumption consciousness, high acceptance of financial products and are familiar with the Internet.

At the same time, the penetration of auto finance business in low-tier cities, the customer quality is still significantly lower than that of banks and auto finance companies, so the risks are obvious. While expanding business scale, strong risk control ability will become the decisive point for emerging auto finance institutions in this round of competition.

First of all, let's analyze the main risks of auto financing leasing business (here we mainly discuss direct leasing), which can be divided into the following three categories:

1.？ Credit default risk

Credit default, that is, we usually say that the default is caused by insufficient repayment ability, and these people who default are real car buyers. This kind of risk is mainly due to false materials when applying, such as high income, or later for some reasons, such as unemployment.

2.？ Vehicle cash-out risk

This risk is mainly because the lessee chooses to cash out the car for the need of cash flow. In the process of renting a car, because the ownership of the vehicle does not belong to the lessee, the cost of realizing it in this way is too high. Such people are generally unable to borrow money through formal channels (including peer-to-peer lending platforms) or are already credit blacklist customers.

3. Fraud risk

Automobile financial fraud can be divided into personal fraud and gang fraud (intermediary fraud). It can be considered that the probability of personal fraud is very small. Suppose a car with a price of 65,438+10,000 yuan needs a down payment of 1% plus insurance, which costs almost 20,000 yuan less. It can be bought in the normal second-hand market for about 70,000-80,000 yuan. However, because the rented car lessee has no ownership, it can only be resold on the black market, about 30,000-40,000 yuan. Therefore, the amount defrauded in this way is not high, but the procedures are extremely cumbersome and the price is high. People who are willing to take this risk can actually fall into the second risk group mentioned above.

? 03 Auto New Financial Big Data Risk Control?

For the first risk, on the one hand, establish a multi-channel cross-validation mechanism to prevent user information from being tampered with; On the other hand, integrate business data and internet big data, and use advanced machine learning technology to comprehensively evaluate users' performance ability; In addition, it is necessary to establish a loan supervision mechanism. If users are found to be abnormal (such as abnormal operation of the work unit and recent multi-platform borrowing behavior), they can take the initiative to intervene in advance to minimize the losses of enterprises.

For the second kind of risk, a black-gray list model can be established. Blacklist data has been favored by mutual gold companies, and almost everyone has come. However, due to data pollution and other problems, the quality of blacklists in the market is uneven, and the overall quality has a downward trend. Therefore, if the blacklist is still blacklisted, it is definitely inappropriate to reject this strong rule logic, which will shut out many customers who are essentially high quality.

Here, we can use Adaboost algorithm to better tap the value of the blacklist and gather the strengths of everyone. According to the principle of this algorithm, each blacklist can be regarded as a weak classifier. With more and more data sources accessing external blacklists, each blacklist is given a certain weight according to its performance, and finally a strong classifier is formed. Divide the black and gray list according to the score of the final model, so as to take measures such as refusing or increasing the down payment or reducing the credit line.

Adaboost algorithm structure

For the third kind of risk, because the fraud in auto finance is different from 3C products or online loans, the procedures are extremely complicated, and all of them need a professional industrial chain team to operate, which generally lasts for a long time and involves a long link. It is precisely because of this fraud feature that the establishment of relevant knowledge maps through big data, combined with offline manual audit, can effectively prevent gang fraud.

Author | First Consumer Finance? Gan Hualai?

Article Source | Financial Technology Security