The introduction
We have entered the second half of the Internet. In the first half, in the early days of the Internet, you never knew who was sitting across from you. At that time, most people were still early users of QQ. In the second half, Internet companies are not new, most companies have become Internet. They are already using the Internet to advertise their products and e-commerce to sell their products. – Mr Wang
Case studies – How to do user analytics
Let’s say you enter a catering company that sells lamb kebabs. The boss says competition is getting fiercer and fiercer now. If you want to do well, you have to understand what customers like. So on your first day at work, your boss asks, “Can you analyze user data to help our business?
User portrait
What is a user portrait? Make a sketch of the user portrait of your own business, and tell him who these users are, where they come from, and where they are going.
User portrait modeling
There are three steps: 1. Unify the user id. 2. Label the user, that is, label the user
Why a unique logo?
Let’s take an App as an example. It connects “all user behaviors from the beginning of using the App to placing orders to after-sales service”, so as to better track and analyze the characteristics of a user. To design a unique identity, you can choose from the following items: user name, registered mobile number, contact number, email, device number, CookieID, etc.
How to label users?
There are many labels, and different products, the range of choice of labels is also different, so many labels, how to divide can not only convenient memory, but also to ensure the comprehensiveness of the user portrait?
Here summed up in eight words – “user consumption behavior analysis”.
Consumer tag: it refers to buying habits, buying intentions and whether it is sensitive to promotion. Behavior analysis: time period, frequency, duration and access path, that is, the habit of using APP. Content analysis: In particular, the content that stays for a long time and is viewed frequently is analyzed to find out what content users are interested in, such as finance, entertainment, education, sports, fashion, technology and so on
It can be said that user portrait is user portrait modeling in the real world. Massive data are labeled to obtain accurate user portrait, so as to solve problems more accurately for enterprises.
So what value can user profiling bring to enterprise data analytics? Dividing business value by three phases of the user life cycle?
1. Customer acquisition: How to attract new customers through more accurate marketing
Sticking-in: personalized recommendation, search sorting, scene operation, etc. 3. Sticking-in: prediction of churn rate, analysis of key nodes to reduce churn rate
If the process of user portrait modeling is divided into data layer, algorithm layer and business layer according to the stages of data flow processing. You’ll notice that on different layers, you’ll need to label them differently.
The data layer refers to the tags in the user’s consumption behavior. We can put “fact labels” as objective records of the data.
The algorithmic layer refers to the user model calculated through these behaviors. We can label the “model label” as a classification identifier for the user’s portrait.
The business layer refers to the means of getting customers, sticking customers and keeping customers. We can put a “prediction label” on it as a result of business correlation.
Therefore, the tagging process is to calculate in the algorithm layer through the “fact label” of the data layer, type the classification results of the “model label”, and finally guide the business layer to get the “prediction label”.
According to the steps described above, the user portrait is modeled
User portrait of lamb kebab shop
User unique Identification: Mobile phone number User tag: Name, gender, region, contact information, age Consumption tag: catering taste, average consumption price, coupon usage percentage Behavior analysis: ordering time, online consumption, weekly and monthly consumption frequency Content analysis: correlation analysis (frequently matched products)
Rule:
- Attract customers: attract new, accurate marketing to obtain customers, find the advantage of the publicity channels
- Sticky guest: Scene operation, personalized recommendation, improve the frequency of users to use, for example, through red envelopes, preferential incentives and other ways to encourage preferential sensitive groups
- Customer retention: forecast of attrition rate, reduce attrition rate, reduce customer attrition rate by 5%, increase company profit by 25% ~ 85%