He Hongling: The practice of China Mobile in large data application and platform

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 forum, China Mobile business director He Hongling a keynote speech to share the practice of China Mobile in the application and platform of large data. He Hongling is China Mobile Business division, presided over China Mobile's normative framework and basic research work.

The following is the full text of He Hongling speech:

He Hongling: Hello, everyone. I thank the host, every time Yang always feel like a big, I am in China Mobile is responsible for architectural planning and basic research work, today very Jong came to such an open day and you share and report about our China Mobile in the application of large data and platform practice, I hope to share and discuss with you.

Just saw fan general introduction, our three major operators are similar, and this kind of thinking is invariably, have done an open platform of the way. Let me introduce you more comprehensively.

The first is the application level, because China Mobile's users are mainly mobile phones on the one hand, we look at the entire mobile phone above the traffic growth is very fast, this quickly reflected in it is a non-linear growth, may be different from our usual perception, it is not the same layout. We see that the voice level is very small, and data mobile phone generated by the volume of data growth is very fast, we now feel that the mobile internet has penetrated into our lives in all aspects of the process. This has enriched our lives for us, reduced the cost of communication and exchange, and provided a lot of sharing. On the other hand, it also precipitates a lot of data assets on the backend of operators. Let's take a look at the typical left to right look at the data we put into the data asset, the first aspect is the business production aspect, is the business produces the data, these data are the earliest to include the data analysis system, mainly includes the user information, the bill, the statement and so on information, this information forms our traditional data information main link, At this level we have done information management, accurate recommendation, accurate marketing methods.

We have a new order in the back of the network, the network New order, call New Order, these data in fact it is close to the user data, can profoundly insight into the user, understand the user, so it constitutes our most valuable data now, we are now a lot of internal and external data has been around this unfolding. Another thing that can not be ignored is that the operators of the three curves, voice traffic and value-added services are mobile internet products, the three major operators have done a lot of work, I will be specific to the following point. In this category of products to the Internet products also enrich a lot of data, because it is closer to the user, so close to the user's original behavior, so it's more valuable. The data is not large enough for our traditional carrier platforms, but this is a good addition to helping us improve the level.

It's a vital part of the big data, and we see now that as Moore's law continues to evolve, and as the closed walls are constantly being opened, the shift from scarce economics to the now-soaring economics is a big one. The ancients said that it was difficult to increase from the reduction, but in fact, we from the comparative scarcity of the times to the rich era many things are facing subversion. Now a lot of things have become free, operators are not excepted. Our operators are likely to be subversive, but the time has not changed, we only 24 hours a day, the time to stay awake is more than 10 hours, how to make full use of the time to collect users, this is the most valuable foundation. Large data for operators in the external and internal two levels, internally we have been trying to achieve the data to achieve profit growth and efficiency of the operation, this is the major data is the main data. Through this data we can arrange to have hundreds of billions of online investment each year, tens of billions of of the annual marketing cost of how to quantify, and it also produces a lot of value, which is the main position and main focus. But what is not to be overlooked is the external one, because the big data everyone knows that there is a very important feature is externality, just said data is not for external applications, are to provide better management, provide better network continuity, but the data saved for many other business models to provide better. This is the data externality application, data externality application is a growth stage, it is not only mobile internet mobile communication industry, but for all enterprises can use. I speak a few internal cases, the first is certainly monitoring the enterprise in the past what happened, more important is to monitor what is happening in the enterprise now, this is what everyone will have, but we through the development of large data, become more real-time become more sophisticated, accurate.

The second one we have to know is where our customers are and what the customer features are. So customer identification is very important. I give an example here, the first example is for campus customer identification, how do we identify campus customers? We may have been the first to decorate and appear in the location of the customers, and then we add time, the user's time and space characteristics, so that some teachers and janitors excluded, or other people on campus outside the campus exclusion. Besides, Unicom users and telecom users are not in our area. There are also truancy and class level he is in the dormitory, or he often run out to play users, this is not accurate enough. We connect through the whole network. Over the years we have found that many students are very different in using the Internet scene, and we can see that the recognition accuracy is very high.

We China Mobile in the past broadband is the successor, these users he handles broadband, his holiday and night and daytime there is a big difference, we have broadband users identified. Also based on the characteristics of recognition and the user a variety of characteristics, still can identify which is the courier, because the courier and other users have a great difference, identify the courier after the identification of which users have done online shopping.

The second one is all about long-term user accumulation, and in fact, we are paying more attention to his short-term changes when big data is booming, because short-term concerns, especially those concerned with change, mean changes in short-term behavioral patterns. This may be a very important time, and then a step further, according to this feature to form his current situation, this is very good. For example, someone in the cinema is not convenient to answer the phone, in commuting, driving or talking about the situation at the meeting, actually through the data we just said to insight into these several things, we have not developed this level, but in fact, we can do in minutes to identify user characteristics, Just a minute to hang up the phone, then I can not not call him, which improves the efficiency of our business.

We combine long-term and real-time insights, which makes BMP, a very sophisticated platform for data insights, through which we can do more of the following applications. The main application is to do marketing, such as 2g3g4g different packages, different business, I have to do a lot of business, these things need to pass the precise way. Just talked about our customer insights, do marketing also need to do a good job of customer and product matching, that is, I actually want to form a personalized recommendation engine. Here to give a terminal example, how do we identify the end of the machine, how to identify the user's preference for changing machines. In the process of changing the machine also has a lot to do cross-selling, such as the user for 4G mobile phone, I may want to give him the recommended 4G package, the previous download can not use the use of 4G end. So we need to do a lot of personalized recommendations, including follow-up we have to do some network optimization or some of the sales work. Through this, we should be last year 150 million terminals, this year need to set up 200 million terminals, this is a very large amount.

The second data quotient three embodiment, for example, flying a letter, this is an Internet business, now its monthly landing number 180 million, or a lot of users, which provides a lot of marketing resources can be used. There are now mobile phone reading, UV 30 million, PV volume 19.5 billion, mobile phone animation and mobile video also has a considerable number of UV, this requires us to provide a lot of recommendations, with customers to do contact information. The second is based on user and user recommendations, as well as content and content recommendations. But we have a lot of user data, it's easier for us to make recommendations for users and users, carriers and contacts, and social networks, so I can know what your friends like. Customer Service staff Contact, we now want to contact the customer service staff to make full use of it. Why do you do that? Because now the active contact with the user with the policy specification this piece already very scarce, but we passive contact user means, we hope each contact user is more accurate, this enhances the efficiency. We are now through with the provincial means to establish services, to one months can probably create millions other recommendations, by texting the billion-level, if through the app is Bai other opportunities. This is just fan always said we are at the provincial level, because we invested in hkust voice recognition, we through voice analysis, customer complaints, recording quality through recording, so we very quickly through the hotspot to collect user insights.

3rd, using network planning, such as our network of large data, priority in where to build, priority adjustment in which, this can be used to plan out large data. And also our entire terminal, as well as the entire switch, the various types of switches its joint situation also requires very real time insight. This is the root of a network.

Next is the network Big Data external service, I now is the big data uses individual customer and the group customer to play the value in four quadrants, the first quadrant is the second quadrant, namely utilizes the social resources and the group value. Third individual customers, the fourth group of customers did not develop. We now have a social index, based on our group of clients, we have some social insights, such as the level of communication between the 31 provinces, the picture on the left, the right and the bottom are very typical features, it is different from foreign maps. I can simply say that the left side of the area calls more, of course, we are not experts, and later to do some dedicated community and scientists to work together, maybe we have some obscure data in their place very great value. Tianjin, for example, has the largest number of calls in Japan, but the largest number of calls from abroad is the United States, Jilin is very high level of communication, answering the phone 57% is from South Korea, 60% from South Korea call. This may be interpreted in the context of the local economy, which, of course, requires a more professional person.

The second is social science research, I combine my own professional, I think if someone can do on our platform analysis, can be more effective more plump. This is the model that our own in-house employees have shaped their influence including the direction of influence and a value, this basically through the data and our implementation of the value is basically one by one connected, and we are within three degrees, what is the situation within six degrees, what the situation, in what class to achieve transmission, this is very interesting point.

Then there is the characteristics of tourism, we based on location recognition of the province as a unit of innovation, made a call intelligent tourism platform, it can provide some of the tourism industry standardized products, can also be combined with large data analysis reports, through partners to provide analysis reports, You can also provide API permissions to partners and governments through an open API. This is the data for the overall customer, that is, its IP is the main object, not the child, and it complies with the KN rule.

This is the screenshot, this is the tourism data index of Beijing, the right is Jiangsu's information products. This is the model after 51 last year, by identifying features, as well as user behavior, including user access to the app, we can know how users travel, how they travel, and what their business goals are, and how they choose to travel, together with data enhancements for some members, and foreign exchange enhancements, This is what the airline does. In the field of traffic information is same, traffic information in the field between the city rarely cloth probes, the cost is very high, a probe cost will be 10,000 yuan, but through the group sounding, he can achieve through the mobile phone. There are cities in the use of the situation, although not as accurate as the lbs map, but through the group behavior, we can analyze the group behavior analysis, through this analysis can better for the city's development, public transport planning to provide some reference. This is the development field of traffic informatization. Also in the retail sector, the Spanish agents, meanwhile we are also trying to do the same thing, and we ourselves on the marketing category to some places, with Wanda's cooperation to do joint marketing. In addition to the financial sector to provide a joint verification of financial, through his information on the financial site, this financial site through the input of the verification code, through the main contacts, as well as consumption through our consumption form to verify. This is the exact marketing we're trying to talk about just now. Precision Marketing In addition to just deep customer insights, as well as personalized recommendation engine, and through the channel to push, this is very important, this is our big data outside the use of a very important resource promotion channel.

Include the following two, one is DNS navigation, you can not find the page back to the navigation page, as well as the page certification page.

We have just talked about the application of the situation, our company application personnel and the provincial company's personnel to try to do the application, external application is based on the specific external platform for cooperation, for me is to be able to on this platform low-cost fast long out, this is our large data platform for China Mobile. We China Mobile's big data platform and Unicom have differences, our natural is two level of large data platform, provincial level is responsible for foreign large data exploration. The group converged on the full web of data, our daily 8TB things system data, 400TB log data, we have a professional company, has headquarters, as well as the provinces of data exchange and data enhancement, the data issued by the work we through the service data, as well as deepening and large data exploration of external applications, including data value, Data exchange and industry applications, the future of this architecture may be adjusted. Because in fact, the province's platform and group platform to form a certain degree of competition, this competition is relatively benign, so that everyone to reference each other for reference, each other for the purpose of the experiment. Why do you do that? Everyone may have a little bit of our technology people may feel profoundly, the infrastructure is already cheap, and it's going to be cheap along the curve, and I might be willing to try something when it becomes less expensive, because I want big data to be fully developed, And did not hinder the play out.

Follow-up is through the two-level large data platform to do cost-effectiveness. Technically similar, everyone is similar to several major operators, traditional enterprises are similar, we have stored a large number of structured data in relation to the Internet, structured data needs to be probed and probed, we will have a BMP data, there are Hadoop, we also have a mashup system, with fan general said that the same, We also through the EQR to unify the management, the Unification service, forms an overall data open platform. For the platform, how can I use this platform to give full play to the value of large data, the application of large data is a long tail of the characteristics of the long tail big head is important, but we must not ignore the tail of the long tail, how we are combined with large data platform is very large, but in fact a small point, to play it out. This is a very important question. The traditional times I serve my internal system, I am very professional, I know the situation of the market very much. But the internal network Planning network optimization, as well as I want to service some of the audit, as well as the service of financial information management, we are lack of expertise in these areas. In addition, especially in the external, our new professional record, the lack of external resources and industry resources, whether external value or economic value will face this problem.

Our method is consistent with what we have just said, which is to build an open platform put the platform a lot of quick and flexible application, fast to try and error, quick try, as he is to understand you in the field of application, you have this background, and to gradually fast, is because fast, failure is very fast. The second is to do good management, you need to know who used this resource. For data acquisition, processing, and display storage of these things, like water and electricity, cloud computing resources to provide.

We also have a reference architecture for a large data platform, in the Reference Architecture we formed a data PAAs platform, there are storage life table, as well as documents, as well as the acquisition capacity has been done gradually, and both control, through their own service can be applied, whether external or internal are such applications, We can build a sdore such a concept to exchange. At the same time as a data exchange, we have in this platform also has the infrastructure of data exchange, the whole inside is to do the platform, do a good job of measurement, do a good job, we can know that Amazon platform mainly for business operations system, our large data system is mainly targeted at special systems, of course, the

Finally, I have a little thought, big data inside we have a point to say, why do we in the individual customer try not much? Because we now feel that there is no solution to the problem. That means I'm doing big data. How do I achieve a win with the user when this value is realized? Because now many large data application patterns are likely to ignore the user, just from the point of view of advertisers and media, this is actually a departure from the current Internet age users occupy the dominant role of the consumer chain, the sovereignty of the user, more demonstrated this logic, I think this logic is 0 choice of circumstances, of course, we have to win. That's what we're going to solve, and it's a difficult solution at the moment, but we're evolving now. We have two ways, one way is personalized recommendation engine, but the individual customer is fully owned by the customer itself, but also he can choose to be hosted on the data platform and behind the application device. This is the right to use the entire data to the user's hands, through such a way to reverse such a data faucet, so that users enjoy more rights, so that we do any operation we can achieve with the user win, this is our idea of large data applications.

I think the big data age, especially the digital economy era, there is a very important point is that we need to share more, and we will not lose anything by sharing, for example, after we communicate today, you will not lose anything, I will not lose anything, but we have increased social value, This is the purpose of our communication today in this forum.

The next step for large data is the most important is open cooperation, the most important cooperation of large data is the external application, China Mobile is a late, very lack of industry support, we in the financial industry, credit industry, in the retail industry, there is no industry knowledge, but also lack of industry experience, So we are eager and willing to have application partners to explore with us how to make society better through data, and make business models of industries and enterprises more efficient. So I hope to be able to have more cooperation, through the contact way, we realize and the user data demand side, the data producer, the data provider forms the win-WINS environment. That's all I'm saying, thank you!

(Responsible editor: Mengyishan)