Three horse enter personal credit, who can play what pattern?

Since the central bank issued "on the preparation of personal credit Service notice" (hereinafter referred to as "notice"), personal credit market, the above-mentioned "notice" requires sesame credit, Tencent Levy letter, Shenzhen Hai Zhengxin, Peng Yuan Levy letter, in good faith levy, in the wisdom wanted letter, Lakara credit, Beijing Hua Dao Levy letter Eight institutions to do a good job of personal credit service preparation, the preparation time is 6 months.

Apart from the established institutions represented by Peng Yuan, these eight commercial organizations have their own backgrounds, sesame, Tencent, the former sea behind the "three horses" again brought about the hot market, in the past two years, "three horses" in the Internet financial field, competition is accompanied by cooperation, and for personal credit this piece of new business, What kind of tricks can three play?

Recently, the "First financial daily" reporter interviewed three credit service related responsible person to interpret their business model, obviously, "three horses" have a focus on, who in the future can become a leader, to market trials.

Dominant data source decision methodology

After the announcement of the announcement, three families began a new round of busy, "This does not mean that the plate issued, the central bank gave 6 months of preparation period, and thereafter to carry out a round of checks to see if the conditions are met." Chu Han, a former general manager of Shenzhen Hai Zhengxin, said in an interview with the first financial daily.

Chu Han Speculation, the next central bank is still concerned about personal information security and personal privacy protection, "whether it is the establishment of internal control system, System Information security mechanism or authorization regulations, is certainly the focus of the inspection direction, these are the preparatory direction." "she said.

And just before the announcement of the announcement, three have also been based on their existing data to carry out a large number of internal testing, ping ' an in 2012 on the credit service, and the central bank also had an intention to communicate until the Credit Management Ordinance and the Credit Institution management measures, respectively, in March 2013 and December, It is the legal basis for commercial organizations to apply for the establishment of credit agencies.

Ping An has 80 million customers, as financial institutions, safe customers are basically real-name, and transaction behavior is more of a financial act, and personal credit required data source highly correlated.

"We actually did some internal tests, more is based on the needs of internal subsidiaries, do not enter the credit system, after the license, we will talk with the group of professional firms, but Ping an group of financial industry is very complex, the face of the regulation is not the same, so still need case processing. "Chu Han said.

In her view, peace has been dealing with risk, different products have different forms of risk, on the basis of integration into the new data sources, you can refer to the traditional experience of a new height.

"For any person's future risk assessment is based on historical verification, the evaluation model without historical data for long time verification is roots, no matter how far the credit development, historical data is always very important." "she said.

and Tencent General Manager (CHIP) Thant in the "First financial daily" reporter interviewed, said Tencent credit data is based on Tencent's own years of data accumulation on the basis of "Tencent's internal data is very rich, including social, payment, games, virtual behavior, network behavior, etc. These data are areas rarely touched by traditional companies.

"In the past some time we did a lot of exploring. Every day in the discovery of new variables, new patterns and new laws, but also in the verification, to observe whether it is suitable for credit scoring, after scientific modeling, testing, will put it into the letter, as a product to provide external. "he said.

It is understood that Tencent currently covers 800 million QQ accounts, 500 million micro-credit accounts and 300 million pay users, for the outside world to question Tencent lack of directly related to the financial data, how to play the credit? Thant said that Tencent has done some credit products, this product is not only for customers with credit cards, there are many attempts to target more grassroots users, such as just enter the Society of college students, the loan amount may only 500 yuan, 1000 yuan.

"We are accumulating this information, perhaps financial institutions or banks think that Tencent does not have so much direct performance data, in fact, this thing we have been doing, now has accumulated a considerable amount of data, enough to let us make a credit model, enough to allow us to a certain degree of user credit judgment, We can now make a credit score for hundreds of millions of users. "Thant said.

Compared to the top two, in the field of personal credit, Ali has been placed to play more tricks, sesame Credit is subordinate to the Ant Financial services Group, the central bank issued the "notice" before the ant under the ant under the micro-loan to the joint Taobao, the cat jointly launched a "flower chant" consumer loan services.

"Flower Bai or a trial of the scene, only open to a part of active users, whether it is a flower bai or micro-loan, behind a lot of models and sesame credit is common." Ant gold Suit a person in charge to this reporter said.

"In fact, personal credit within Ali has been advancing for many years, Ali small loan is to use the data on the line to judge the seller's credit, this is the letter data, Ali small loan credit data can even give the bank for reference, at present, sesame Credit internal testing model has been more than tens of thousands. "he said.

It is understood that the current ant gold Clothing has 300 million real-name users, "the outside world has been saying, we only have the electricity quotient data, but in fact, the electric quotient data accompanies a series of behavior: registers, receive goods, the logistics, the attestation, the money based social tool, the finance, the Reservation hotel, the booking flight, the electricity electricity and water coal expense, the electric business is one The things that diverge around this point are very rich. "said the ant in charge."

Modeling capability determines effectiveness

Although three think that they have accumulated the most abundant data in the industry, but only data sources are not enough, behind it is the data processing capacity and research, how to establish a large number of data to judge the model of personal credit, and verified to be effective, is the key.

"The so-called Big data analysis, in fact, is to find the correlation between different data, find causal relationship, there are many times, these causal relationships in our common sense is difficult to think of, in the traditional area of credit, judge a person's credibility of the dimension is very simple, such as the income level, status, there is no room, there is no car, But in the Internet world, a dimension that you never pay attention to may be related to or inverse of credit. "said the ant in charge."

So, how to set up the correlation between this dimension and the dimension becomes the first problem before the three credit agencies.

"Credit is complicated, like two people bought the same phone today, but does not mean that the two people are the same, maybe their consumption concept is not the same, the former person is more economical, so bought this mobile phone, but he himself very rich, after a person more wasteful, after buying mobile phone pocket empty, two people's credit affirmation is not the same, So use historical data as a reference to verify that this rule is valid. "Chu Han said.

"The model is not very special, but these models are ultimately effective, whether it is suitable for users, it is necessary to verify that the model is only a presentation, it is important to modeling the ability to analyze the ability." "Chu Han said.

She believes that people are very complex, alone from any direction to evaluate is not enough, to the various dimensions in tandem, especially with financial-related dimensions, financial dimension is more important.

Thant also said that it is very difficult to establish an effective credit decision model, to try one in each variable, "tens of thousands of variables have to be validated," Thant said, Tencent's large data on the Internet to use the vast number of information on social networks, such as online, property, consumption, social and other situations, Create a credit report based on Internet information for users.

Specifically, the credit system will use its large data platform Tdbank, in different data sources, collecting and processing related behavior and basic image data, and the use of statistics, traditional machine learning methods, to obtain the user credits score, the formation of personal information reporting.

In the eyes of everyone, social software is often lack of authenticity, if taken as a reference to the credit data, is it effective? To this, Thant explained, the effectiveness of social data is a cutting-edge topic both at home and abroad, and there are many different voices in the industry, "but for us, social data is certainly useful, and the explorations and models we've done prove it, and now we're working on how social data works and how it works, How to make it more stable in ratings. ”

For example, for example, a QQ account is a virtual identity, although we may not know who is behind the user, but can not say that this account is worthless, a scarce QQ number on the black market price may be tens of thousands of or a hundred thousand of, the account itself has produced value.

"If the person behind this account says to the bank one day, I will give you this account, can I borrow 500 yuan?" We think it can be done, and that's the value of virtual property. "Thant said.

"Sesame Credit construction model more than tens of thousands, these models are also core competitiveness, no one company will disclose the composition of the model, the model once leaked, one involves commercial leaks, and two may cause credit arbitrage." "The Ant's Gold suit insiders say.

"For example, we found that a consumer's donation behavior is rich, credit is usually very good, and the ratio of both is unchanged." We will talk to the market, if often love, credit will be good, but I will not tell you one months to offer how many times of love, amount of money, calculate credit good. "he said.

"We have an open, distributed data Platform system, the data to do some sorting, calculation, will be drawn to the user's portrait, including his financial relations and so on." After testing, it was found that both the crowd coverage and the business scene credit to the degree of distinction between a good performance. "The Ant's Gold suit insiders say.

The person explained, Sesame Credit can refer to the dimension includes Taobao, Alipay wallet, credit card, payment, payment, consumption, age, sex, occupation, family, status, registration information, whether real name certification, interest preferences, the grade of goods, there are no blacklist records, funds, payment channels, financial contacts, Wait a minute. "There are a lot of relationships, we can chat, add a friend, but the money is the most accurate, those who may have a financial relationship with you will be more than ordinary people who chat with you more closely." "he said.

In addition, Sesame credit to these personal credit evaluation system developed, but also with the external data two-way flow, so that the feedback data back to sesame credit, so as to amend or accumulate this person's credit situation.