Session: Data and technology two-way drive electric business marketing development

Session: Data and technology two-way drive electric business marketing development

News and Science and technology 2012 Fourth annual effect of the integrated Marketing Forum held in Beijing December 18, the theme of this forum for both internal and external marketing broken Bureau, Dispatch as the exclusive financial media to this meeting to carry out the full report.

The graph is the topic "Data + technology, two-way drive electric business marketing development" of the peak dialogue, the moderator for the site analysis in China founder Sing, CMO Xu, Tian Product Network CEO Bear, the public comments chief traffic Officer Lu Xin, 58 with the city Vice President Zhangkuang, Google ADX responsible Judy Zhou, Tencent SNS advertising director Zhao, Lohong vice president of the data and technical issues discussed.

The following is a transcript of the dialogue:

Sing: Microphone to me, please other participants in the forum to take the stage. Again, I hope everyone will give a warm applause to the guests at least 6 times times as warmly as I have just asked. First of all, I am very grateful to the million MA can invite me and distinguished guests to participate in such a forum, the previous speech is very exciting, the topic will be more in-depth, in-depth people want to know the details of things. I met some of the guests here and we had some exchanges. I have a request to them, this request hope everybody more to say the detail, more say dry goods, I these friends all agree.

You can expect something in the next one hours. Before we begin, we'll introduce the general issues of our forum. First of all, I would like to ask each of you to introduce yourself, and then I will, on behalf of our friends here, question each and every one of you. After that, please friends, if time permitting, according to the number of times, please ask a few friends under the table a few questions. You may be interested in asking a particular guest or ask a question directly. Hope everyone is not too nervous, the next friend may ask some of the more sharp questions they want to know. Please ask the King first.

Wang: We are May, I try to introduce, I am a big bear. is currently doing flash website.

Xu: I am Xu.

Zhao: I believe that there are some of our partners have more understanding, we and the billion MA is very close partner, but also hope that the future, we are as a service provider, hope that the future to provide more effective and accurate, better marketing services.

Judy Zhou: I'm from Google Adx.

Lohong: The DSP business of Yili I am in charge. is closely related to the subject.

Sing: From the introduction of the guests know today for everyone to answer the question of a strong lineup. Start with the Big Bear. Day Product NET is also the website of the Flash purchase, I learned before, the Big Bear team is very strong. For me, for E-commerce, a very good team mature logo is that we can use the data to do their own operation and management to do a good job marketing feeling, for many electrical dealers in the past do not pay special attention, but now I have talked with Xu, we feel that this piece is very necessary to strengthen the field. First throw to bear, what experience do your company have to use data to do operations management, there are some references to use data to do brand management, what is the experience?

Bear: First of all, I have to say that it is too much pressure to talk about data in front of China's top data experts. So I have to ask the song teacher, we Do data, to tell the truth in the past Taobao experience, we have their own system, this we are self-built. We and some of the mainstream domestic advertising companies have some cooperation, including the Yi Ma side is also our important partner. Data this piece I share, we do data some experience, first we have a complete set of data systems, is our bi this is very important, I hope that every company if the value of data or to have such a system. Secondly, we use data to guide the operation, because I want to get the user this data marketing, then another expert will speak. I'm going to say something on the back end. How data is used in operational management. Here are some things from the previous experience in Taobao.

Day Product network is a Flash purchase site, we see more of our brand-oriented, in the home page from top to bottom things, here there is learning. How does the data play a role in it? For example, if we don't take into account every day, why this brand more than that brand before, first, we are monitoring the home page also similar to the location of the brand CTI Ctr, this is the first dimension, if the two position is similar, a brand click rate than the other high, we have a low click-through brand adjustment. The second dimension is the behavior after the click, click to a few PV open new session inside we will do conversion rate monitoring, very simple. The 1th is to appear in our home page all brands we will monitor the CTR, the second monitor the entire conversion rate, the third we will monitor the speed of sales and our inventory between the degree of Mach, we do snapping, realistic discounts such. A lot of times we have big cards, like Nike will snap up very badly we are in a relatively short time, a lot of merchandise rob. At this time very important application is the sales speed and inventory depth to set up an inventory alert line. We will move the fast-selling brand down, the stock conversion rate is good to the top. A system is not a hot meat implementation, we use a data tool to guide the operation. This is a very small share of our data in the operation of this piece.

In fact, this is all because before Taobao do double 11, double 11 with this principle to do replacement. If a location hits very well, the inventory depth is not enough to affect the conversion.

2nd I say a topic. From the internal discussion of our company. For example, we have more than 20 people buying teams talking about brands all day, talk about some good goods all day, come back some controversy, some people say that this brand is very cool, how, some people say this brand I have not heard, can sell good sell bad, there is controversy, buy their own argument. We believe in buying hands to judge. I give some advice on whether to use some other platform to sell or something to monitor this thing. Later, we agreed, we choose the brand or selection dimension, not all dimensions, is the data support is three pieces.

The first is to see the brand Baidu Index, which everyone will see. The second look at this brand Taobao search index, the third look at Taobao turnover index, these three data is not the whole brand, at least a dimension of supporting is not suitable for online sales, this for us, for us to buy the hands of 5 million or 20 million goods or not, this is very important. This is a small application of our three data selection brands. This still has a data support, which analyzes the role of the entire data.

Sing: You first mentioned my very sensitive words bi, a lot of e-commerce, whether Taobao or their own independent brand E-commerce, BI is not open the topic, direct decision in the operation of this efficiency is very significant. Before you Taobao work, including the current work bi this is very experienced, can you share the experience of building bi?

Big Bear: This can be said to be very simple, we have a number of BI main modules, the first is the teacher song is always good at the field WA, the entire Web page traffic analysis, this is the first link. Complete and flow-related. Each distribution, which is the first module. The second module is what we call a Product management module, which means that all of our products, including which category, how many category attributes, attribute values, including the style index, are also in our system. The third block is about inventory and supply chain links, that is, how much inventory, and from which suppliers, the data, the fourth is the financial, because a large number of related to us such as margin, net profit, as well as the vendor settlement between the accounting period, the financial module. Then there is a courier-related, such as express arrival, and express company docking, this piece we do is not good.

And it's important, through all the merchandise Management financial Information module, may be our report system, because we want to calculate, a UV produce how much sales, how much profit margins, and even eventually produce how many PV distribution in which category, this user came to the habit of which block, this and the whole chain has relations. We have a comprehensive report that runs through the basics.

Sing: How long will it take to make such a system? How much does it cost?

Bear: It's hard to share if you've seen a lot of data on a big platform, know what the underlying data is, how many of the original data are multiplied together to get the new data method, and there is no problem with a slightly more strong development. Because we did it.

Sing: You didn't count. Thanks Bear, I'm not asking anymore. Please Xu answer the following question. I want Xu to share with you how to use the data to do operations and management?

Xu: First to play a commercial, today we meet for the first time, sing in the current domestic Internet data analysis, I am not good to say one, but a certain first three. A Long online blog I also share my friends and comrades around me, people want to focus on data-driven e-commerce or the Internet must be concerned about what he has written and the future to write the blog.

Simply said I believe that the link, talk about DSP, I can say 80% people do not understand, because they do not know what is DSP. But personally, I am extremely respected. I was in the forward 07 out to rest 1.5, many are now doing DSP company that time does not say DSP, said a lot of nouns, at that time was I veto, I do network advertising 12, the former section of the Iris, said no my name, because I do 12, because above 15 years. No, it's party B, I want to do 12 I think I can.

I was very understanding, but I think it is impossible to do, I think the promotion of Internet marketing development is the two main driving force, a technology a market. 07, 08 time, talk about the Arabian Nights, I feel very good. Many bosses don't understand. Do not simply throw online ads, you should have a deeper understanding of DSP.

For example, we all know that advertising, advertising alliances, the biggest controversy is in portals and vertical Web sites to display ads or client ads, which is the most controversial, because you see the ROI miserable, if you have been unable to see. This time, such as DSP out of time, is not able to take part of the cost to do membership marketing. For example, I work with the service provider, for those who did not produce orders to the shopping cart, this ratio is very high. To the shopping cart link did not generate orders, there are people discard orders, do not pay, this does not say. I'm going to go through some definitions to all the ads when it comes to the shopping cart only, pushing what? Must push once to put the goods inside, this time I was going to advertise, I see, we are doing very rough, because and partners this October test, very rough, before a few dimensions, add up huge data table. To the present form is not implemented, but the most academician DSP way, than I simply to cast a hard wide roi efficiency 4.3 times times.

The second how to evaluate the media, we are based on the user to identify the media value, last year in this time also said, we all know a person to eat steamed stuffed bun may be 3 full, henceforth to the third, the first second does not give. Some need 4, 5 paths, some 20 paths, we judge the media and the boss to judge the last order to judge, but there is no want to in other delivery also come in, out, point in out, no orders, the last order most from this section.

I cast a position in good 123, not my website name, home page on the two you can see. This advertisement position one month order more than 1700 single, because is the display type advertisement, actually within 30 days once only orders to visit this advertisement media 21000 single. You know how to evaluate a media.

You last time to count him, we look at this media how many orders, if it is to his users have never come before, the first time to, point this media orders, how do I evaluate. If the first and last time as the most important case, their proportion is higher, what the media situation, there will be a large dimension evaluation of the media, we put ads a new customer an old guest, when the error, to the market orders, this is a new customer, the main market impact of new customers, because the operation can affect the old customers, new customers difficult to affect. But not to the market orders the most important 99% is a new passenger, I said there is a problem. When I put on advertising, I must meet new guests, the ROI is really low? I think now many websites, everybody does not look that platform is particularly big, do not know oneself new guest cost how much money, calculating way is what? All the marketing costs to go with this phase of the new customers calculate a ballpark, not the most accurate, I see a two-month trend, but not accurate, to accurately monitor the cost of 100% new customers how much money. If it can be counted, all ROI will be good for the new customer. I just give an example.

These examples you can understand, I speak network advertising, no media no value, the key point this time the media is not needed, with data, multidimensional. Now we evaluate too single, the last order to generate media evaluation results, this time will be caught in a misunderstanding.

Sing: Thank Xu very much, Xu always said a lot of views to my heart. I have a very strong resonance, one is how to view the value of the channel, including the Million-ma land a lot of energy to do the box, let us see the last transformation channel before the end of what channels? We've never been concerned about these things in the past, and if we look at these things and find out that we have overestimated some of the media and neglected the value of others, we may have problems with the way we put them. Xu always emphasizes how important it is that I see the value of these channels and how much assists we can make using tools such as the billion-box. Some electric dealers are now aware that we are more comprehensive, similar to a user lifecycle approach, and it is important to re-evaluate channel roles and channel strategies in such a way. Special thanks to Xu General sharing.

Please Lu Xin to talk about the following questions. I'm using a public comment this noon. The public reviews There are too many data, quite is the user behavior data, the user leaves own interest, this kind of data populace reviews how uses these data, the populace reviews how uses the resources to make the big money?

Lu Xin: I had a chat with my friends yesterday, no doubt, the data itself we think is very big gold, see how you dig. The front has an analysis of DSP, in fact, the most core is data mining, a data technology. The data level tells us that we have a lot of advantages, for example, we have a lot of users left very valuable information on our website. Based on our technology how we use, the information better feedback to our users.

Profit for the current company did not think of this point how to use data to do more profit points, the whole company for the direction should we use the data, the user and the merchant services better. At the user level, we see, for example, location-based recommendations, based on interest, we know what restaurants you are interested in and what you are interested in. It's probably based on some location information, knowing where you're going every day, where you work, and what time of day. Based on this location, combine your interests and recommend a suitable scenario for the user to better the experience.

Today to the basic should be very little part of want to listen to the public reviews come here, this is the will of the billion, million Ma professional to do marketing. I share my marketing experience as a marketer for so many years. I have been talking about data marketing in a lot of the meeting, because I have done the products in the last two years, only a few years ago only large companies have professional bi, they will talk about a concept of data products and operations. I have heard more and more recently these two years, the related bi will be many, is the future trend. I've been asked what this career might look like in the future, and I say at least one career is important in the next 10 years, a career called BI, data analyst, and technical data mining.

I am a marketing-born, product operations do not want to speak more. Talk about marketing experience. We cut from the marketing point of view, Marketing Department team division of Labor in accordance with the profession down, based on this we have some different, we are based on marketing perspective, set up a marketing team, we have specifically targeted at a very understanding of the market and products, understand our users, the product selling point is very good packaging out, specializing in product packaging team, There's also a team based on selling points that specialize in copywriting. Media distribution of advertising, all media mix to tell users my story, the public comment on exactly what, why buy my products, but not the data analysis team. Because data analysis is a must for a marketer, no matter what to do, each channel to launch, to all 8 of the value of the delivery must understand, to be able to directly measure the data, including each of the different value of the media evaluation, we ourselves, we may have a model for all of our media have a different dimension, Very comprehensive assessment of his channel value. The whole marketing goal is what I have direct new customer costs, each channel to see a year of customer return is not the same, guest unit price, gross margin are different.

And we hope that every person who spends money doesn't want to spend the first time with the user, and the second time, to see how expensive each user is to turn to another user. This ratio is very large, the media is not the same. In addition, there are assists, marketing such multi-channel like playing basketball, there are strikers shooting a little more, many companies and departments to the media value is very direct, such as visit Baidu, Google, visit good 123, in a good 123 orders purchase. A lot of media value evaluation to 123, in fact, the Million MA, Google has passed.

Sing: Thank Lu Xin, can you introduce how to evaluate, just you said algorithm, I have not asked, you said. We also have a special want to cross-examine the content, a later communication. Zhang Luo always should be technical origin, today's topic is data and technology, for the current marketing, for the current precision marketing no technology is not accurate, I would like to ask two from the technical point of view, how do you look, from your work and technology to see how accurate the impact of technology?

Zhangkuang: I have a reverse problem with this question, to everyone. Really technology to this extent, I do products, I have done products, my own feeling may be data over technology, at the current stage. More future stages may be more important than technology, which is related to the development of the system as a whole. I first to share this piece, if you have the idea of DSP, I 58 to do a DSP, because the market can not find suitable partners. If you can find the million Ma can not do it yourself. This is another idea. As an advertiser to put it.

Sing: Generally in the presiding time rarely inserted, usually after the host often have a very good idea.

Zhangkuang: When we look at the data, we think the data is the current core word. I did the exact marketing myself. For example, in fact, we were the simplest product, last year on-line a product, on-line time uses 80 days from 0 to 2,000,001 days, when doing this product, the definition is very simple, because before Baidu, everybody does any product to know has a product, I to its definition two decides this product the victory and defeat two. The first is not to cut the word. The second is from small and medium start to try, do not start from the big customer. Why did you do that? At present the data is the simplest and easiest, do not cut the word without considering the match effect, BMW 3 is BMW 3, do not consider the matching effect. The accuracy of the second data can be seen within two hours of data far better than 24 hours of data. 24 Hours of data is better than 30 hours, we were 2 hours 24 hours 3 days 7 days, 7 days after the data clicks and ordinary ads, no effect.

All users need to be reflected within 2 hours, he searched BMW 3, two hours later attenuation. 24 hours will also be higher than 50% to 70%, has been attenuated to a sufficiently high level. This person is a tourist search is not effective investment? I'm not sure. I have to catch the fastest match, a no-cut word, this technology is really effective. For other companies 10 minutes, for Baidu for a second, this function is very powerful. Guaranteed delivery effect is very obvious. This is the data.

Second, we can imagine that when we put the precision out, we have to do is a big customer, large customer word volume, but limited coverage, because the need for a system, distributed to the entire system of Baidu Customer system will find words more than tens of millions, billion, we found dozens of days we buy the amount of billion, then check, I'm afraid the coverage won't reach a certain level. Through small and medium-sized quickly sold, because the General Assembly to buy the best beauty Hospital in Beijing, how to find, these words will be bought. When employing methods to solve technical matching problems, it will be found that the accuracy and effectiveness of the best.

Personally I think, I put forward the opposite opinion, the first question is first we put the pure data problem if solve, this is the first question. This data is most important because the last time I joined the Google Conference was to endorse Google first-hand data. In fact, why do I do DSP, because I have completely left the effect of advertising, the effect of marketing circle, as advertisers to tell a story, Lao Yao said this year to cast 200 million in the league, the original did a Baidu alliance, now is not to do 58 league, I do not, cheat, Back-click Evaluation these things were not built in a day or two. But the DSP will not be the same.

We are now working with Google and Tencent, and we have three ads next week, so I don't have to build alliances, and then I make a system of alliances. Most of the traffic has been signed. We just do one thing, I vote directly in the advertisement, you want to find a house in the Asian Games village, want to move in the Asian Games Village, show the application, the alliance system will be successful. What does DSP achieve? As an advertiser to achieve a bit, I actually have no league, because of how difficult it is to do the alliance, I saw a little done. Do the alliance is very difficult, Baidu Alliance in the system 6 years kungfu, too difficult. Because of another giant number. This makes all want to do Alliance enterprises do not have to do alliances.

Other enterprises can make use of DSP. I hope I do not use 58 DSP. If a deeper measurement match is to be achieved by technology. Back to the data and technology, data can solve 65% problems. Think high points must be done by technology. Do not solve 60% problems, first cut the word, solve the content, solve 40% useless. All advertisers and everything on the data level, accumulating 60% has been solved. We may not imagine Baidu's original core revenue, is using data, using technology to solve. Baidu now invest in tens of millions of solutions to solve the problem, is to solve the 40% problem.

The first one is to address the core of the data problem first. The second is to invest in technical strength, the third is I think the data problem, do not ignore the advertisers factor. Advertisers provide accurate words more also represent the data, this data volume millions of enough to accumulate hundreds of billions of discoveries will not be wasted.

Sing: I have a very good harvest, and I have some thoughts, such as Baidu data volume, how to use this data, I agree with Zhang General said first is the data itself, how to look at the data. On this basis, the question of what technology is used to process and apply this data is 40%. There are also some problems that need to be solved quickly, such as RTB and so on, which requires a quick resolution of some requests in milliseconds, which requires technology. The relationship between the two is difficult, and more is the case that everyone is in harmony together.

Zhangkuang: In fact, to separate this, two technology is separate.

Sing: The next one is Tencent's Zhao. is to do SNS marketing, Tencent has a very good SNS products, including Tencent Weibo, as well as Tencent Space, QQ zone, such as some products. These products accumulate a large amount of data, but also include not only dsp,dsp current emphasis on behavioral aspects of data, SNS is not the same, but also a congenital advantage, and innate, they have the user's social attributes, personal properties of individual attributes. How do you Tencent look at the data? How to use this data?

Zhao: Thank you very much, I would simply say, our products, from the end of last year just on line, this year and there are many partners together to achieve very rapid development, summed up in the course of development, in fact, the entire product system based on the algorithm and technology three horizontal one longitudinal system, creating large flow data large effect product loop. Data play a vital role. How to use this data? A guest mentioned that advertising, social advertising and search ads have different points. What is a social advertisement? Our overall advertising data and algorithmic analysis is based on one person, our traditional advertisement may use kooky to do (sound), this person buys what item tomorrow, buys a pair of shoes tomorrow or the day after next buys the shoes, we are very difficult to relate all his behavior to a person, we only know the single behavior, cannot outline this person stereoscopic state. Based on Tencent's large data platform, we have the opportunity to achieve a more comprehensive three-dimensional understanding of a person. Because through the QQ account, in addition to the PC uplink for us to better understand, in fact, in the mobile behavior we can do a better understanding of the same kind.

In the Tencent platform has many product lines, Tencent internal product line from the electricity quotient to the game, from the wealth through account to the Q currency account, we all know this user's interest and the demand all round. We also have Q interconnect products, we know the behavior data in the outside. For example, may be in the beautiful said to browse an object, or to buy a commodity in Jingdong, these data stations inside and outside the combination of our grasp more comprehensive.

From what we do now the whole algorithm level to look at the relatively elementary, there is a user's physical properties, static properties, we can easily understand the region, gender, age and so on. In addition we have some dynamic tracking, from dynamic tracking more effective to know the big recent needs, this is a large data a basic part. This is also based on all SNS large flow based on. In addition to our QQ space, the entire client on the QQ platform will be more and more intrusive platform system. In addition to the physical properties of this piece, there are different series of dynamic properties. We are now doing a relatively junior, such as a friend to recommend what things, or friends play a game, your friend and your relationship level, you have recently communicated, including your friends in other sites or other operational actions of depth, which can help us to achieve higher accuracy.

I said the initial data used, some in-depth data added incrementally, this block based on the overall user attributes, as well as user and user relationship chain impact. We know QQ space, there are a lot of content. For example, the photo stock 360 million, how many users of the iphone, how many users with SLR digital camera upload, we define user consumption ability and circle layer, many ads mainly looking for white-collar users and high paying ability users, we from his supplies, he usually chat keyword we understand his level and positioning.

Just now, which one mentioned very important, from the advertiser feedback from the data, we now the entire advertising platform, the user clicks have transformed to do, did not transform no ROI, in the click to turn and the process of user behavior is divided into many classes, I refer to browsing, browsing several times or under the order. We push ads for all customers who have placed orders. On the whole circuit we are now more than want to be in the overall circuit to make every detail point more detailed, including the user each behavior, the behavior of the value of how high, according to our final deal to improve this efficiency.

This product development relatively elementary, is also a new product, very happy to have this opportunity and everybody communicates. We can communicate if we have a problem. Now many advertisers, our product target this year CTR priority, if the provision of valuable content more clicks, we think is now the most effective. Many advertisers say that no matter how high the CPC is, it will not be consumed. Or how to upgrade, we thank the billion MA professional company to help us, in the next year we will be more inclined to CVR promotion, more attention to conversion rate, pay attention to the final results.

Sing: Because you talked about CTR, I'd like to ask about the pay model.

Zhao: CPC and CPM. That's all I'm going to say. We hope that we have cooperated with partners or no cooperation partners, we hope that more understanding of the real needs of everyone, hope to be able to better serve everyone. Thank you.

Sing: The last two questions, Judy Zhou and I have been friends for many years, Judy Zhou is very understanding of this product, we now talk about DSP this way can not leave the other a bit like the trading parties, there must be a buyer seller, Judy zhougoogle ADX is the seller platform, I would like to ask you how this platform to help you do precision marketing?

Judy Zhou: Just about the DSP situation, in the entire Google ADX inside the buyer and seller, on the same platform, the seller platform like the site will put some do not sell traffic on the platform, and then let everyone in this huge pool to buy traffic. The buyer is a DSP, the buyer has a DSP advertising alliance and so on, are Google ADX buyer in this flow to buy.

Just now Xu always mentioned that in the 07, many companies had this idea and wanted to do so. At that time immature, there is no open platform for everyone to buy. Originally bought on the website, I bought on the website CTM or the parcel section buys is this website traffic, buys is he buys to you the situation. Then there is an advertising alliance, to help you cross the media to buy, for example, Sina Sohu NetEase do not need a home site to see, in an advertising alliance to buy the desired, advertising alliance to help you gather flow resources, through their technology and data can be better to buy some of the desired advertising positions. This time, including the site, or the ad league, buy or based on this web site such as the Alliance traffic things. For example I want to buy this user not in this ad league what to do, there is any way to find these people. There was no way because there was no open platform for you to find them. Each house is a relatively closed platform.

After the acquisition of Google07 and 08, one of the products is an open platform, a series of upgrades and planning to make all traffic open, no longer just an ad league, or a platform based on the alliance. So that all traffic can be based on their own data, want to buy people and traffic to buy the desired data.

We should be earlier in the year before the end of last year, we launched the Google ADX in China, the whole year to this year, 1.5 time domestic DSP growth very fast, we are basically nearly through the certification of more than 20 DSP in operation, the actual we can see, Through the ADX, we refer to the data in the way of purchase, which mainly tells us both. The first one is the first party data, because many of you are electricity dealers, have their own trading data, some users access to data, including the billion, we have been doing the purchase of electric dealers will accumulate a lot of data. This kind of data will be very effective in helping you to do it based on every kooky purchase, and each purchase is based on traffic. Before the day to buy, a lot of traffic is useless. Through RTB to buy a lot, maybe every site as long as buy two can achieve results. Not like to buy a time before. This is a subversive revolution of DSP and data.

In addition, the data in the market does not form a comparative forefront of DMP, another called Third-party data, can actually and some of their own partners, can get their data, according to their data can be in the entire trading platform through DSP to help you buy the desired traffic. New things in the entire ecosystem. We have just a few guests mentioned that the search advertising effect is very good, 1% how much conversion rate, in fact, through the RTB of the purchase method, as well as technology, display ads can bring a better effect of search advertising, key data based on the results of technology.

Sing: So RTB way we want to achieve a few points, a point directly to achieve user interest up. Second can achieve a different, can real-time bidding, advertisers want to buy this position this person can bid. This one, which I think is revolutionary, has just talked about the change in the name of the revolution. Please summarize the last question. You come from Yili, summarizing the help for precision marketing through technology including data.

Lohong: I describe the DSP in a non technical way. DSP is the purchase of the form of change, mainly because when a user browsing online, there is a specific condition. When a user browses online, we buy traffic to see when and where the user is using what equipment, and in which site to see an advertisement. DSP to do accurate time will take into account the user's Internet environment, which is a very critical technology.

Second, we want to know what he has done before, this user's previous data, I must know, must be accurate, and be very timely. The timeliness and timeliness of data are important. This is our current billion. This piece of data is relatively good, all order data, user data collection and so are real-time, is an advantage, talk about real-time.

Another I have so much data, what is good for DSP? For data mining analysis capabilities, it should be very much to find out which of these data is the core data that affects the end-user's intention to purchase. Because this piece of how to find the key point, do data mining is very crucial to the core work. According to the final decision of these things to show this user what ads, show what ads have a very critical problem, that is to show what ads for users, the general traditional promotional activities may be a creative, now to promote from a simple brand to promote the promotion of goods, not participle is what? I need this type of product. This requires a very key technology dynamic creativity, dynamic creativity according to the user has searched content to show him, his specific needs of goods, which we have solved according to a user in different places, times, circumstances, according to his past behavior to show him the most needed goods, so that he will have a stronger desire to purchase.

Sing: Thank you. Finally Xu always give us the finishing touch.

Xu: Do 12 years, we say a lot of people still don't quite understand, I use a metaphor. The first generation of nature to buy traffic, the second generation of network advertising nature to buy orders, the DSP is essentially to buy customers. The starting point is different, the operation idea is different.

Sing: Thank you for sharing. Thank you very much, gentlemen.