In the era of data operation, the operation mode has changed from extensive to refined in the past. User portraits are popular, and I am embarrassed to say that I am doing refined operations without engaging in user portraits. The construction of various user portrait labeling systems teaches you to build user portraits from 0 to 1 and other articles are widely spread. A few days ago, I heard some students think that portraits are labels, labels are portraits, and user portraits and user groups are different names of the same subject. When designing product architecture, the boundaries are unclear and the functions are staggered. So I think we should go back to the most basic questions and make these concepts clear.
First, the user portrait
The concept of user portrait is not new, even before the Internet era. It was first put forward by AlanCooper, the father of interaction design: "the PersonaReactionRepresentation of the target user." It means that real users can be represented by a target user model established by a series of attribute data.
Conceptually, user portrait is an abstract summary of user characteristics. For example, I want to know the characteristics of users who bought Beijing Universal Resort Hotel, so as to be more accurate in product design, marketing strategy and advertising. Just like two people in love, only by understanding each other's temper, personality and eating habits can they get along happily. User portraits are the result of highly abstract aggregation and should generally be used for data analysis and decision-making. It can be divided into personal portraits and group portraits.
Personal portraits are mainly used in customer service scenes to help customer service staff quickly master the characteristics of consultants, so as to flexibly adjust their speaking skills and improve the one-time service solution rate. With the implementation of the personal information protection law, the data security of personal portraits has become more and more important. Group portrait refers to the distribution of portrait characteristics of a certain group of users, such as the age distribution of WeChat users in the official WeChat account. Is it more in the z era or more in the post-80 s?
Second, portrait labels.
Used to outline the feature dimension of the user portrait. When a new classmate joins your team, what dimensions will you get to know him quickly? Hometown, age, single/married, work experience, hobbies, etc. For example, there is a life cycle label in the user portrait label. According to business characteristics or algorithm mining, the last time the task was put in was more than 180 days, which was defined as lost users. The data label is the time of the last order, or the time since the last order. Generally speaking, portrait tag is a more abstract and understandable compound tag formed by data tag after analysis and processing.
According to the characteristics of tag data and different processing methods, portrait tags can be divided into:
Statistical label
: Indicators that can be directly obtained through statistical analysis of data, such as cumulative consumption amount, 0- 100, 100-300, 300-500.
Rule class label
: Convert statistical values into business rules to form more intuitive tag values, such as high-frequency consumer users, which are defined as the number of consumer orders exceeding 5 in the past six months.
Algorithm predicts category labels
Statistical data can't be obtained directly, but need to be obtained by data mining algorithms, such as user price-sensitive labels, and need to be obtained by a series of statistical algorithms and machine learning prediction algorithms.
Third, the labeling system.
The essence of label is also an evaluation index, but it is more detailed in dimension, generally it is the label of user dimension or commodity resource dimension. For example, the number of orders in the last 30 days is an indicator, from macro to micro, that is, all, business lines, traffic portals, categories, goods and users. That is, the label system emphasizes the index value of a single user or a single commodity.
Label system is a label classification based on business scenarios, such as primary classification: basic attributes, behavior attributes, marketing attributes, risk control attributes, etc. , and then split and enrich level by level. The labeling system should be easy to expand, understand and use. The tag system is more comprehensive and rich than portrait tag, and the portrait tag obtained by using basic tags is also the input of the tag system. Compared with portrait tags, data tags are more flexible and have stronger crowd selection and stratification capabilities.
Fourth, user grouping.
User grouping refers to selecting target users according to specific conditions, conducting insight analysis to see user characteristics, or reaching these users directly. User grouping depends on tag assets. For example, recall the lost users, select the target users according to the following conditions: the last visit time is more than 180 days, and the number of historical orders is less than or equal to 1, and then use marketing means such as SMS and push to reach the recall strategy. The input of user grouping is label, and the output is user collection. The application scenario is mainly refined operation.
Verb (abbreviation for verb) user insight
User insight is to analyze the characteristics of the target user group, such as viewing the portrait characteristics of the single user in an activity. Or compare and analyze the operation activities of different groups of people, judge the difference of user conversion effect of different label screening, and adjust the next operation strategy. The input of user insight is the crowd, and the output is the portrait characteristics of the crowd.
Sixth, the relationship between user portrait, portrait label, label system and user grouping.
According to the above definition, the relationship between these names is sorted out as follows:
By collecting data sources such as user attribute data, behavior data, transaction data and commodity data, a data label system is formed. The tag system can continue to abstract business rules and process tag values to form portrait tags. It can also be directly used as a screening condition for crowd selection on CDP platform.
Portrait label is an abstract data label, which can be used for the analysis of individual portraits and group portraits of users, and can also be used as a label condition for user circles.
User screening conditions for user grouping can come from data tags and portrait tags, and application scenarios include: crowd portrait analysis, refined operation and precise marketing.
Seven. abstract
The concepts of user portrait, portrait label and user grouping are simple and easy to understand, but together, can you accurately distinguish the relationship and boundary between them? When operating applications or designing data products, it will be clearer to understand the differences between them.