How does big data land in enterprises?

How does big data land in enterprises?

I often hear many concepts and trends of big data, but there are relatively few actual introductions. Based on the practical experience in the field of Internet and data, the author summarizes the application model of data value pyramid in enterprise operation. The model corresponds to different levels of data requirements in enterprise operation, and is introduced layer by layer below.

Database platform layer, the bottom of the pyramid is also the basic layer of the whole pyramid. If the basic layer is not well built, it is difficult for the application layer above to play an effective role in enterprise operation. Without data or high-quality data, all analysis is misleading, and all data mining is wrong guidance.

The goal of this layer is to string all the user (customer) data of the enterprise with a unique ID, including portraits (such as gender, age, etc. ), users (customers) behavior and hobbies, so as to achieve a comprehensive understanding of users (customers). There are three keys to doing well: 1. Enterprises need to determine the unique ID for accessing data. Some enterprises use member registration numbers, some use mobile phone numbers or ID numbers, and so on. 2. Interdepartmental data integration. Enterprises with big data usually have many departments, and the data of various behaviors and hobbies of users (customers) are scattered in different departments, which requires enterprises to consciously carry out compulsory integration; 3. Manage data through technical means and normative means. The problem to be solved here is what the data in the data warehouse means and how to store and calculate it efficiently, which involves data access system, metadata management system and computing task scheduling system.

Business operation monitoring layer. The first thing at this level is to build a key data system for business operations. On this basis, the data products developed by intelligent model can monitor the changes of key data, quickly locate the reasons for data changes and assist operational decision-making. If an enterprise builds the ability of real-time calculation, it can find many problems in its operation in time.

User/customer experience optimization layer. This level mainly monitors and optimizes the user/customer experience through data. Both structured data and unstructured data (such as text) are used to monitor the experience. The former is more realized by using various models or tools of user (customer) experience monitoring, while the latter is more to find negative word-of-mouth by monitoring Weibo, forums and internal customer feedback systems, so as to optimize products or services in time.

The business operation monitoring layer and the user/customer experience optimization layer finally hope to realize the intelligent doctor of enterprise operation. The tools made by these two levels are like thermometers, sphygmomanometers, B-ultrasound, CT and other tools. With these tools, we can quickly see which module has problems in enterprise operation.

Refined operation and refined marketing layer. There are four things at this level: 1. Establish user-based data extraction and operation tools. Operation and marketing personnel can extract data (users/customers) through simple condition configuration (such as choosing male, 18-24 years old, with specific hobbies) and conduct marketing or operation activities for users/customers behind the data; 2. Improve customers' response to activities (such as click rate) through data mining. Common algorithms include decision tree and logistic regression. 3. Customer life cycle management through data mining. Different from traditional customer life cycle management, big data can mark and warn customers in different life cycles in real time, and push effective activities to customers in different life cycle stages as commodities in time; 4. Personalized customer recommendation. Personalized recommendation algorithm is mainly used to recommend different goods or products according to different interests and needs of users, so as to maximize the efficiency and effect of promoting resources.

Data assists market communication. At this level, through "sexy" data analysis and mining, there are two main ways to help the product spread: one is fun data information map, I believe everyone doesn't like to read the public relations soft articles of products, but prefers to watch fun content. Especially on the Internet, netizens aged 10-29 account for more than half of all netizens in China (55%, CNNIC 20 13 data), and these users are young "diaosi", so these audiences prefer "sexy" content.

Taobao once found that Xi 'an had the largest proportion of netizens by counting the regional distribution of users who bought bras with C cups or above, and released this data, saying that girls in Xi 'an had the largest breasts, which caused the spread of many "diaosi" netizens. In March of this year, Tencent first disclosed the data map of "Escape from Beishangguang" based on more than 800 million active users, and found that 1 1% users fled from Beishangguang after the Spring Festival.

Another way of data-assisted market communication is to directly use data products for external use. For example, the Baidu index or the migration map made by Baidu during the Chinese New Year. From the data of Baidu's 8-hour migration map in Dongguan, we can see that after leaving Dongguan, most people went to Hong Kong. Then can we simply get a message that there are the most people going to Dongguan from Hong Kong? ...

Business management analysis and strategic analysis layer. I won't talk about these two levels here, because they are more similar to many traditional methodologies of strategic analysis and business analysis. The biggest difference is that the data comes from big data. But there are two aspects to pay attention to:

1. Many enterprises mistakenly put what the "business operation monitoring layer" and "user/customer experience optimization layer" do in the business analysis or strategic analysis layer. I think "business operation monitoring layer" and "user/customer experience optimization layer" are more realized by machines, algorithms and data products, while "strategic analysis" and "business analysis" are more realized by people. Many enterprises give people what machines can do, which leads to low efficiency in finding problems. My suggestion is that what machines can do, they should also do a good job of "business operation monitoring layer" and "user/customer experience optimization layer". On this basis, people can do experience analysis and strategic judgment that humans are better at;

2. In the rapidly changing Internet field, it is difficult for data to predict the general development direction of business in the choice of business strategic direction. If someone says that the general direction of WeChat is studied through data mining and analysis, it is estimated that product managers will laugh. In essence, data can play a better role in refined marketing and operation, but it plays a smaller role in creative things such as product planning and advertising creativity. But once the product idea comes out, it can pass the gray scale test and verify the effect with data.

In my opinion, it would be very good if the status quo, problems and reasons can be clarified with data through machines, algorithms or artificial means, so that the decision-making layer can make better "head-slapping" decisions based on these situations.

In short, this paper only outlines the landing plan of big data in enterprises. There are more details and methods not shown. In addition, the landing of big data in different industries may be quite different. Therefore, colleagues from all walks of life are welcome to discuss with me.

The above is what Bian Xiao shared about how big data landed in enterprises. For more information, you can pay attention to Global Ivy and share more dry goods.