Association letter. Network data CTO Roven in an interview with CSDN said that enterprises in the choice of DMP platform, need to consider four factors. He thinks, the first is DMP data platform's data richness, whether can produce the value to oneself, second is whether the DMP platform provides the matching application, may use own data conveniently, third is DMP platform's technical ability, whether has the strong system structure to support the DMP system The third is the service of DMP platform operators, the operator of the DMP platform can give more help to the enterprise in the application of data.
As a third party Internet Data Service provider, with independent research and development of the Super Large network data Service Platform, association letter to the unique panoramic data service model, in the Web site operation effect, network media value assessment, network advertising marketing effectiveness, network PR public opinion, E-commerce and other aspects for various types of websites, brand Enterprises, PR and Advertising agencies, government departments, such as providing detailed professional data monitoring, analysis and consulting services.
At present, the association letter has accumulated more than 400 million continuous analysis of netizen behavior data, daily average data processing capacity of 3 billion, customer base coverage of domestic mainstream media websites, government, automobiles, it and other industries. The Association letter has a great advantage in the underlying data accumulation. Recently, CSDN for the users concerned about some of the problems of the letter. Network data CTO Roven conducted an interview, the following is an interview record.
Beijing Association Letter Internet Data Technology Co., Ltd. CTO Roven
Association Letter Technical TeamCSDN: First introduce yourself and the Letter of association and the technical team behind it?
Roven: I was in 2007 when the letter was founded, at that time in the company is responsible for the development of data statistics products, in 2012 as the CTO, in charge of the Association letter Technical Management work.
Association letter technical team has more than 40 employees, divided into three major parts: product development, system development, system operations, product development is responsible for all the business products of the display, calculation and other work; system development is responsible for data platform research and development, including system architecture construction, data processing, mining and other work The system operation and maintenance is responsible for the stability of the system platform, the hundreds of servers and the network.
CSDN: Can you tell us about the current situation of domestic data analysis and the position of association letter in this field?
Roven: Domestic data analysis field, from our customers so many years of contact, more and more enterprises realize the importance of data, many enterprises from no data to have data, from data to have available data, and then from available to useful enterprise data process. and "available" and "useful", one is data management, one is based on data management data open, but also the current application of large data is the most urgent to promote the link. Business and data disjointed, fast-acting for quick-impact, data island model, these problems often make the enterprise's data process stalled or can not realize the value of data value-added.
Association letter has been through the data should be used to help enterprises to do data management, and the use of data in the ease of doing a lot of work, such as we developed this year's Web site user portrait, web site analysis and other products have received praise from customers.
product composition and userCSDN: What is the current product composition and business direction?
Roven: Our products provide a "one-stop" solution to the site, from the basis of daily traffic statistics to the user's interest mapping analysis, content recommendation, advertising guidance. The Data products of association letter are siterating website Traffic monitor system, adrating network advertisement effect monitoring system, clickrating user Click Statistic System, apprating app data management system, userportrait website user grouping portrait, etc.
CSDN: How about the size and composition of the company's customers at present? What are the heavyweight customers?
Roven: At present, the group of clients covering the domestic mainstream media sites, government industry authorities, large DSP companies, top 4 a agency and automotive, IT, FMCG, home appliances and other industries of the first-line brand enterprises.
Including such as Sina, NetEase, Sohu, Phoenix, China Net, CCTV and other websites, and SAIC, Ren Pharmaceutical, Rui Jie Network and other enterprises and MEIDAV such DSP and we have a cooperative relationship.
Advantages and Technical frameworkCSDN: For the Internet data analysis companies, there are a lot of domestic, compared to other data analysis companies, what are your advantages?
Roven: Several advantages of the letter of association:
's core team members have long Web site work experience prior to joining the letter of association. For the operation of the website, the use of the site data has a deeper understanding; the letter of association was established in 2007 has been in the service for the website, in this 7 years accumulated a large number of data analysis experience; 3 billion PV per day , about 200 million users ' access data has a great advantage in the underlying data accumulation;CSDN: Can you share the technical architecture of your data mining platform? What are the biggest difficulties in the development process and what are some good experiences to share?
Roven: The data Platform architecture diagram of the association letter is as follows:
Data collection: Our way of counting the data is to embed a JavaScript file on the page of the website, when the user visits the page, the Javscript code will count the current page, source page, flash version of the user's visit, And a URL is sent to our data receiving server (the mobile end is embedded in an SDK package on the app). Our receiving server uses Lvs+nginx to do data reception, the data received is saved in Web log format, and data transfer tools we use Flume,flume is a distributed and highly available data collection tool that enables real-time transmission of massive amounts of data through a simple configuration.
Data processing: Through the flume convergence will eventually be saved in Hadoop, first we will clean the data, cleaning the purpose of the log is required to extract the field, do structured processing, the other is to remove dirty data. For the data after cleaning, we will be in accordance with the needs of statistical services to the data calculation, and generate results to provide statistical services for the query, in addition, we will be mining these data, analysis of user preferences, points of interest and other characteristics.
Data applications: Generate front-end applications based on different requirements, get data through backend APIs, and render.
Several difficult points in development:
1. Massive data processing, our data platform will have 3 billion of new data every day, how this data can be in the limited computing resources in time to complete the processing is a very big challenge, which requires us from the system and processing programs to do continuous optimization.
2. Semantic analysis, in order to analyze the user's preferences, we need to user access to the Web page for semantic analysis, through the content of the article to get the user's concern content, this piece we currently have a team in charge of this work, and now has made some progress.
3. Industry knowledge system collation, such as a user like the car, then he is concerned about what level of the car, what brand, what price, buy a car he is more concerned about the structure, appearance or fuel consumption? These need to have a knowledge system to support, we have since 2010 set up a specialized team responsible for the collation of the industry knowledge system.
How to choose DMP Platform for EnterprisesCSDN: Note that there are some arguments about whether the enterprise is a "first party" or "third party" DMP, what do you think?
Roven: The first side of the DMP platform more emphasis on the independence of the enterprise's own data, can be more effective and convenient to manage their own data, but the first party's data may be isolated, one-sided, he can only reflect the company's access to the situation of data. The third party DMP platform will be different channels, different kinds of data through to form a three-dimensional data chain, can be said to produce 1+1 greater than 2 of the value of the data. Data can only be achieved through continuous interconnection, and large data development must avoid island data. In this way of thinking of the formation of the data marketing, so that the user's behavior data on the Internet can carry out the whole process, a full sample of records, and because of the whole network, its present value is real and effective, and with the continuous extension of the data chain, the relationship between the data richer and more perfect, the application effect will be more and more. Of course, from the industry point of view, whether the first party or third party DMP, can have a positive effect on the enterprise.
CSDN: What factors do you think companies need to consider when choosing a DMP platform?
Roven: The first is the DMP data platform data richness, whether can produce the value to oneself; In addition, whether the DMP platform provides corresponding matching application, can use own data conveniently; Thirdly, the technical capability of DMP platform and whether there is a strong system structure to support DMP system The third is the service of DMP platform operators, the operator of the DMP platform can give more help to the enterprise in the application of data.
CSDN: From the user's point of view, what are they most concerned about, and how do you deal with it?
Roven: The problem that users are most concerned about is the value that our products and services can bring to them, that is, whether we can help our customers to make money or save money, our products are designed based on this point.
China Innovation "Pioneer" Enterprise Series report serial number company name establishment time CEO/CTO official Micro-blog company product/direction 1. Yun Yu with 2012 Chen BenStart building with 50+ products and up to 12 months usage for Elastic Compute Service
1 on 1 presale consultation
24/7 Technical Support 6 Free Tickets per Quarter Faster Response
Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.
A comprehensive suite of global cloud computing services to power your business
Payment Methods We Support
