Data Xuanxiaohua of China Institute: Practice of large bank data

2014 Zhongguancun Large Data day on December 11, 2014 in Zhongguancun, the General Assembly to "aggregate data assets, promote industrial innovation" as the theme, to explore data asset management and transformation, large data depth technology and industry data application innovation and ecological system construction and so on key issues. The Conference also carries on the question of the demand and practice of the departments in charge of the government, the finance, the operators and so on to realize the path of transformation and industry innovation through the management and operation of data assets.

In the afternoon of the financial @big Data Forum, the Chinese Yuan chairman Xuanxiaohua for the "Bank of large data practice" keynote speech, sharing how banks use large data for data analysis.

The following is his speech to where to ask:

Xuanxiaohua: Thank you, Huanhuan, for the invitation of the General Assembly. The topic of the General Assembly is the practice of large data in the banking industry, frankly speaking, the practice of large data banks in China is relatively small, accounting for about 10% of the proportion. This does not mean that we do not attach great importance to bank data, but I have always thought that banking finance is the highest, or one of the highest, in the Big Data index. Big data in the banking industry we are very concerned about. 2002, when the founding of the Chinese Academy when the first industry we want to do is the bank data mining. Because the quality of the bank's own data is very high, and overseas, the bank's data market, data analysis and mining application market is the largest, at that time, because the Internet has not how to get up, the electrical business has not how to get up. When we started to do it we found that we did the banking business and did a great job.

Banks in China at this stage, the previous bank is very small, just one. Banks have more than 10 20 banks by 2000, and the banks ' own earnings are very late in China with foreign banks, especially with foreign retail banks. Like Bank of communications, retail banks earn less than 10% and have a small share of the overall income. And the banks are so far so easy to make money, as long as the deposit inhaled, can be relatively safe, higher interest rates to the enterprise, the spread of 2% to 3%, so most of the bank's profit is very good, so in this sense, really push the whole bank of large data practice or real application, I think the main one is the interest rate marketization, reduce spreads, spreads narrowed, the entire bank's operation must have the big promotion, the big promotion in the data and the big data can produce the function to be very big. Including the marketing, the entire retail bank is also the retail industry, so the customer management is very important.

The second is the nature of the bank, risk management. Especially in the risk management of personal business, data based decision-making is indeed considered a very good way, and probably the only way. The bank to individual decision-making, everyone according to the account manager, the management of a very strong manager is not good, through technical means, is the model, which is based on data decision-making and risk pricing. So I just spoke of the banking data itself, just ahead of the construction bank's Liu always said, so I do not expand, large data in the bank can be in the customer demand matching, marketing, the entire customer management, personalization and risk areas can have great potential. The bank itself has a very good foundation, has a lot of data, has a strong internal capacity, and many banks themselves have a lot of data-related talents.

Just now I talked about Bank two very important one is customer management, customer management is very important, the bank from the first day into his customers, the next no matter whether the provision of personalized products, whether it is money, whether it is related to risk products are also good, etc., are required a lot of automated marketing management methods, Because the bank has hundreds of thousands of users, I can't do it according to the account manager, account managers tend to do big, big customers have to automate the way, when customers need what to produce, and when to recommend, which requires customer lifecycle management, including predicting when such a customer will change to another bank, This process must be alerted by the data alerts, so this is the customer management needs a very strong data and data model capabilities. It's automated software, so banks have to do the same things when there's no big data, and they've developed many of these models. Around large data, you should be more convenient to understand the nature of the customer, or more means, so personalization becomes easier, so large data is to provide more data resources and better insight into the user's ability.

The second is our very important event in the bank is the breakdown, or the customer segmentation. Banking services to a large number of customers are not subdivided, so the work itself is to subdivide, we have to compare a few equivalent? Can I divide my clients into different groups, if this thing can be done first, the actual is the basis of all customer management you have, this basic management is to the big crowd into small people, and the second is that everyone has a strong view. A very important way for a bank is to do it based on its behavior, because the bank records all the payments, the loans, including the day, all the records, all the records can be used as the basis for subdivision. The algorithm I do not say, is the bank all customers into 10 groups, each group of users need the characteristics of the bank is very similar. The characteristics of different groups are very dissimilar. But how to subdivide this is to use a lot of fields, this bank has, some fields can also be extended. For example, the average number of dollars per month, this is also through the bottom of the record is very easy to produce, which produced a customer description. Based on this subdivision can do a lot of recommendations, this recommendation is for example, when the customer has come to the bank, see the counter all day for customers to recommend financial products are often not targeted, is to ask the financial products to do not, this is not accurate recommendations, or some customers call the bank, How can you clearly determine what this customer is probably looking for, this job is a prediction, a relational model or a precision recommendation. It's not that hard to do, so the big data practice of many banks is not the technology to limit it, nor the big data itself limits it, the most important is to be able to customer-centric ideas and business practices to limit the current majority of banks can not be in the application of large data quickly or a lot.

Like when a client comes in or calls in, my manager is not able to accurately recommend a reasonable product and solution to the customer, this job is large data can make judgments, large data to do the way of judgment is also very simple, the equivalent of what we have said about buying a book model is the same, based on product relevance, or based on human similarity to recommend the right product to customers, this is not technically difficult to do, and can do well, this is the second example.

The third example can also give you a little bit of talk about how banks look at value customers. When we say the bank will find that the bank to judge a value of the customer is the way to deposit, that is, 5 million of the deposit is very large customers, 500,000 to 5 million of the people are also, 500,000 less important, this is a simple judgment. But not every user's money exists in a bank, especially if you run a bank, and even if I have 10,000 banks in this bank, I may be a very high asset client. So how does this judgment work? Actually it's easy to do with data, why? This is also our model, is an asset prediction model. If you can analyze all of its withdrawals, every time he went to the ATM, although he has only tens of thousands of yuan, before the limit of our withdrawals is 2000 yuan, he each withdrawals are 2000 dollars, or every time the transfer is very large, then he is also a very important customer, such a customer for the bank is actually very good. This customer is not in your bank now, but you can tell by these actions that this is a very valuable customer, this valuable customer you can use a good way, whether it is a good product or good service, to attract him to his home bank, this is a very good development of the user's channels and ways, And the way to develop the user is very cheap, based on the customer's behavior in the bank, can infer the value of this asset, this is very important. Including director Wang just said, although in the investment deposit is not much, but also gave him Golden Sunflower card, this is through the data to judge the wealth or influence of the person. This is another service that can help the bank to do the customer well through the data.

There is also a growth value, if we can analyze the bank, many customers are very young, is the former old customer's law, found that a class of customer bearing value is high. For example, he is not paid much, wealth is not very big, but his path to develop the so-called growth of the user model and the crowd, this kind of crowd is also very important to the bank. Because this kind of people although the wealth is not very big now, but it has very strong wealth increase ability, this situation is also through the data model to do, is also a model, is the customer growth model.

So I'm talking about a few models that can be very simple for banks to do, many IT companies, lots of big data analytics companies can do. As long as it is the business and IT departments of the bank to treat this matter as a business, as a customer-centric concept, if in the bank itself has become a real sense of the assessment, to do it is very very fast, this is what I said around the customer.

Finally, I'd like to mention our credit score. The other is wind control, credit scoring follow the control, big enterprises wind control or through the people and team to do, they have to spend a lot of time to understand the business of large enterprises, this relationship with the data is not so big. The second type is the small, small micro itself can be related to the data, gradually there will be bigger and more small loans are done through the data, but it is the most difficult, in China is even more difficult, because the financial data itself can not be a very important symbol, does not mean that can not be done. Like Ali the small loan may have exploited more Internet transactions in his data here, or supply chain, because know this person's partner and business situation, so microfinance based on data can do, and there is great potential, small micro in China is too big, so there are very many banks. The bank I speak of is not necessarily a traditional bank, it can be the current Internet finance up these, can be based on the data to the small micro-finance work first start. But I think the most mature of the big data is the personal credit score, and it is the easiest to do. Also based on this, the Chinese courtyard has also set up a company not long ago, in fact, we are through data to make personal loans easier. The company positioning itself is to help banks or financial institutions, can be the banking industry can be internet companies, can really carry out personal risk measurement. The traditional way of measuring personal risk is in the United States (English), whether it is loans or whatever, is the standard score, of course, many banks have their own ratings, combined with the transaction with the bank, to make a personalized rating. Whole to Internet finance before actually (English became a very standard score, in fact, the U.S. Business Credit Bureau has also launched another personal credit score, the development of internet finance in recent years, especially in the United States (English), essentially in the development of peer-to-peer loans, so now mention another way to call (English), It is thought to be based on large data, based on more variables, the (English) is also based on more than the traditional financial and financial information used, (English) also used to score, and the financial score is perpendicular, is a better data. But (English may apply a more generalized score, with tens of thousands of variables, I have not communicated with it myself and I am not clear.) But what we're doing, the ratings that are expected to be created in the country are indeed based on a broader range of data, which includes the traditional, recent data collected by our center, including internet business data, can be Weibo, can be a distributor, and do so on a licensed basis. This process we are also doing in the process of each of the industries we have done before, such as operators we have done, the previous operators are paid and forward charges, the phone interrupted, since your phone cost is over, this is a very bad experience. We need to help China Mobile to reopen this thing. You can continue to play after the call, but not everyone can play, but based on the credit score to decideWho can continue to call, some can play 5 yuan, some can play 25, some can play 50 yuan, based on the so-called score. So each industry we do score, the telecommunications industry to do the grading, telecommunications industry we are very familiar, because we serve more than 60% of China's network, 60% of the network operators with our analysis, we developed a data analysis and marketing platform, marketing automation tools.

So we did this score, but where is the difficulty? We want to be a relatively unified score, and there is no good way to integrate data into the People's Bank of China Credit Center, consumers do not want the data can be opened, so we can do is to each industry has a rating, this consumer acceptable. But consumers are unwilling to accept that operators can release the data to a company that makes loans, and that it needs a good rating. And this score is to use a comprehensive rating of various industries, it should be said that has not been very good, but now there are companies to try such a rating, only to try to do this score well. We are also allowed to have bad debts, do not have bad debts must be patted the head, no corresponding to the probability of bad debts. So we must have certain bad debts, bad debts and those who have bad debts, who did not produce, need two kinds of people, can continue to optimize the score. There are some Internet financial enterprises that specialize in making very small loans, the biggest use of personal credit scoring is in small loans, thousands of blocks of loans and a lot of consumer finance, consumer finance is very big potential in China. Now the Internet companies have financial white stripes, like Beijing East has Jingdong white. It is through credit and marketing risk into marketing leverage that you seem to be running a risk, in fact, in marketing, consumers can be earlier. Like Haier, it is a car financial development is more, when buying a car can use a lot of loans, loans to buy a car, now see relatively little is to buy a refrigerator, I buy an air-conditioning can be done by stages, this I think is a large number of needs.

This is a lot of places where banks and so-called big data really can produce value, since last year, the data from the Chinese Academy began to pay much attention to the financial industry, mainly because interest rates are open, so I believe that the profitability of banks will be reduced, so personal retail banking will be greatly developed, so this brings a very very big opportunity. Second, internet finance, although business is very weak, but it does bring a huge impact, ideas and ideas of the impact. The large number of banks that are now being developed by small lenders or internet finance companies is very large in demand and potential for new modern banks. Thank you, I mainly talked about the bank. The Chinese Academy specializes in data mining, but the process also has a deep incubation of many companies, including telecommunications, specializing in carrier servers, which we have done a lot. At the same time we have to do electrical business, there is mobile marketing. Also do personal credit management, mainly to help all industries and fields to achieve data.

(Responsible editor: Mengyishan)