Practical case of enterprise big data

First, the home appliance industry

Take a home appliance enterprise as an example, it not only makes well-known air conditioners, refrigerators and rice cookers, but also makes smart homes with hundreds of products. In its group structure, IT departments and HR, finance and other departments operate in the form of business divisions.

At present, the household appliances and consumer electronics industries are experiencing "internal troubles and foreign invasion", overcapacity, price war and serious homogenization. Internet companies participate in subverting the competitive model. Xiaomi's "fan economy" and LeTV's "platform+content+terminal+application" all focus on operating "users" rather than production. The company hopes to build the ultimate products and personalized services, and recommend the right products to the right customers through the right channels. In the CPC mode, there is only CP matching (product channel) at present, and there is no panoramic support from users, so it is impossible to get through the matching of "CP (customer product)" and "CC (customer channel)".

Based on the above internal and external environment and business drivers, the company hopes to make big data the hub of all business solutions. Take big data DMP as the core of enterprise data, make full use of internal data sources and external data sources, organize enterprise data according to different domains, and form a complete enterprise data asset. Then, use this system to serve all kinds of applications in the whole enterprise value chain.

Then the problem is coming. Company data is scattered in different systems, and more Internet e-commerce data is scattered in major e-commerce platforms, which cannot be effectively utilized. How to solve it? The company's response strategy is: 1) First, start with external Internet data and introduce big data processing technology. On the one hand, it can solve the shortage of external Internet e-commerce data utilization, on the other hand, it can test big data technology. Because there is no problem that a large number of internet data need internal coordination, it is easier to get results quickly; 2) Make DMP a unified data management platform for enterprises, integrate internal and external data, make user portraits and build a panoramic view of users.

Phase I Construction Content: Technically, customized Spark Crawler crawls Internet data every day (mainly user comments from Tmall, JD.COM, Gome, Suning and Taobao). ), and use Hadoop platform for storage and semantic analysis, and finally realize three modules: industry analysis, competing product analysis and single product analysis.

The effect of the first phase construction of big data system in home appliance companies is quickly reflected in market insight, brand diagnosis, product analysis and user feedback.

Phase II construction goal: to build a unified data management platform, integrating the company's internal system data, external Internet data (such as e-commerce data) and third-party data (such as external cooperation and consumer data provided by Tabu).

The biggest value of the company's big data project to the enterprise is to turn the precipitated data assets into productivity. IT department plays the role of agile IT by building a unified data management platform for enterprises, integrating internal and external data of enterprises, quickly supporting new applications; Through the insight into products, brands and industries, the Division assists enterprises to optimize and improve product design, advertising marketing and service optimization, helps enterprises to carry out refined operations, and helps enterprises to create the ultimate service experience for users based on accurate marketing and personalized recommendation of user portraits, thus enhancing customer stickiness and satisfaction; Strategy department, through market and industry analysis, helps enterprises to carry out product layout and strategic deployment.

Second, fast-growing industries.

Taking Procter & Gamble as an example, in the cooperation with Procter & Gamble's China Marketing Department, it is found that it is not necessary to integrate internal and external data before making user portraits and customer insights. P&G captures all the data related to P&G evaluation on mainstream websites, grasps the shopping preferences and habits of different consumer groups through semantic analysis and modeling, and quickly realizes customer insight by only using external public data.

In addition, P&G has also made innovations in channel management. Use Internet user comment data to listen to the community, monitor user comments related to 50 retail stores cooperated with Procter & Gamble, conduct channel/shopper research through online data, and guide channel management optimization.

Implementation process:

1, lock in Internet data sources such as Weibo and Public Comment, and collect P&G shopping-related content talked about by millions of consumers;

2. Using natural language processing technology, the multi-dimensional modeling of user comments is carried out, including shopping environment, service, value, etc. 10 and 50 secondary dimensions, which realizes the quantification of user comments;

3. Continuously monitor 50 retail channels such as Wal-Mart, Watsons and JD.COM, and the results are presented in the form of dashboards and regular analysis reports.

Therefore, P&G can relate the internal data of the enterprise, grasp the overall situation of KA channel more effectively, and even further grasp the key details, advantages and disadvantages of KA channel, guide the adjustment of channel rating system, and help make product promotion plans.

Third, the financial industry.

For consumer finance, the cases of household appliances and FMCG are equally applicable, especially in precision marketing and product recommendation. Here we mainly share the application of credit risk control. Obviously, if Internet finance conducts on-the-spot investigation on micro-loans like banks and invests a lot of manpower for analysis and evaluation, the cost is very high, so there is a batch credit scoring model based on big data. The ultimate goal is to achieve corporate portraits and corporate key figures portraits, and then use data mining and data modeling methods to establish a credit model. CreditEase's pleasant loan and sesame credit are essentially this structure.

In the contact with financial customers, it is found that both banks and financial companies need external data more and more urgently, especially external strong characteristic data, such as records of dishonesty, records of third-party authorization, network behavior and so on.

The above is the relevant content about the actual case of enterprise big data shared by Bian Xiao. For more information, you can pay attention to the global ivy and share more dry goods.