Refined operations: How to establish a user tag system?

With the disappearance of traffic dividends, the cost of acquiring new users is getting higher and higher. In pursuit of higher ROI, the focus of products and operations has shifted from acquiring new customers to refined operations of existing users.

The basis of refined operations lies in labeling management of user information. When it comes to user label management, the first step is to label all users of the product. This is the most important starting point for user operations and the cornerstone of operational strategy formulation.

1. What is a “label”?

"Label" is an abstract classification and summary of a certain characteristic of a specific group or object. For example, the label "college students" is actually a summary of all college student groups. This label can be further divided into grade, major, etc., and a certain group of users can be found through labels at different levels.

2. What is the labeling system ?

Simply put, it is to label personalized users with standardized labels based on their "basic attributes", "behavioral characteristics", "social networks", "psychological characteristics" and "interests and hobbies", and then sort and aggregate the labels to form typical user labels, and then do precision marketing or personalized recommendations based on different user labels.

Its essence is a process of de-differentiation . Let’s take a simple example: suppose a financial product is planning a marketing campaign. First, list the products and services involved in the financial scenario, filter out the target customer group through user tags, and then further combine user preference tags, such as investment preference, risk preference, product preference, etc., to carry out differentiated marketing.

Having said so much, many students may ask, what is the specific role of the label system? What goals can it help us achieve?

3. What is the purpose of establishing a labeling system?

The editor believes that the role of establishing a labeling system can be roughly summarized as follows:

1. Increase new customers

A certain reading APP attracts new users by inviting friends to receive red envelopes. It labels users who have already attracted new users, and then labels users who have not shared as "users without attracting new users."

The product operation department then pops up and pushes ads to users labeled as "users with no new user acquisition behavior", increases the reward amount, etc., continuously stimulating this group of users with no new user acquisition behavior to spread the word, in order to achieve the goal of "every user is a promoter".

2. Increase retention

In order to better increase user downloads and reduce download costs, a novel APP uses different novel materials for different distribution channels (the novel materials will match the interests of users on the distribution platform).

For example, if suspense novels are used for delivery on platform A and fantasy novels are used for delivery on platform B, the novel APP recommendation algorithm will push the corresponding content to users as soon as they download it. In this way, when users from platform A open the APP, the first thing they see is suspense novels; and users from platform B will see fantasy novels.

3. Reduce churn rate

For example, in an e-commerce APP, if a user does not log in for 14 days, the user's status will be marked as a "potential lost user" and a coupon will be automatically sent to the user to wake him up; if the user is not woken up, the user's status will be marked as a "lost user" and then recalled by PUSH plus SMS push of large coupons.

The user tag system is of great help to operational work. How do we build a complete user tag system? Let me share my understanding of the process of building this system.

4. How to build a user tag system from 0-1

The process of building a label system is divided into the following parts:

1. The basis of the labeling system: data collection

Data collection mainly includes four categories: user attributes, user behavior data, user content preference data, and user transaction data.

But the data we get from the Internet is not completely true. For example, user A is male, but according to his historical consumption records on an e-commerce app, all he purchased were women's products. Such data will mislead operations. Therefore, the collection of user data is far from enough relying solely on the operations team; the technical team also needs the ability to analyze and calculate data using big data.

Through big data algorithms and models, with user behavior data as the basis, combined with business data and other data sources, we help enterprises build user intelligent tags, enable businesses to achieve self-service creation, maintenance and management of user tags, and make user portraits more accurate and closer to real user portraits.

2. Behavior Modeling: Obtaining Model Labels

After obtaining the original data, we can perform secondary processing and behavior modeling on the user data through a computing engine based on the user analysis model and other auxiliary data to obtain the model label.

There are many categories of labels themselves, so at this stage, many models are needed to label users:

(1) User status model

User status is an important attribute for evaluating the closeness of the connection between a brand and a user, and consumption frequency is usually used as the main basis for judgment. After understanding the "level of intimacy" between each user and the brand at this stage, we can carry out different marketing operations for users in different states.

(2) User Value Model

Judging the value of a user to a product is very useful for improving user retention. The RFM model is a widely used customer relationship analysis model that can easily, quickly and effectively quantify user value and profit-generating capabilities.

The RFM models are:

• R = Recency The most recent consumption
• F = Frequency
• M = Monetary

Through these three core indicators, we build a user value model.

The picture comes from the Internet

The different quadrant areas in the figure correspond to different customer groups, which can be roughly divided into eight categories: important value customers, important retained customers, important development customers, important retained customers, general value customers, general retained customers, general development customers, and general retained customers.

We can carry out targeted marketing based on different customer value attributes. For example, up-selling and cross-selling to customers on the right side of the R quadrant. As for important retained customers, since they were once very valuable and we don’t want them to leave, we can specifically recall this group of people.

(3) User grouping model

User status and user value are based on the analysis of user consumption behavior, while the user segmentation model differentiates customer groups based entirely on product business scenarios.

For example, the home appliance industry will guess the user's current purchasing needs based on the time of purchase, and divide users into three categories within the system: "facility upgrade needs", "new installation package purchase needs", and "replacement needs". It will push marketing services with different stimulation points such as "high-end recommendations", "package purchase discounts", and "old for new" to users with different needs.​

The supermarket industry will divide users into digital users, maternal and child users, fresh food users, home appliance users, etc. according to the categories they purchase, so as to identify target groups and send targeted coupons.

3. Data output: label visualization

If the operations team wants to make personalized recommendations based on user groups, they must meet the requirements of label visualization. For example, after searching for a certain tag, the tags related to it will be directly displayed; and two or three levels can be displayed, with multiple levels to clearly and intuitively see the association situation; and further, after clicking, the detailed situation in each tag will also be displayed.

Image source: Digital Business Cloud MA

For example, the historical browsing records of tags, opening channels, sharing status, target user groups, etc., so this also needs to rely on the output of the relevant tags combined with the user system.​

4. Effective management mechanism: label maintenance

For many companies, it is not difficult to generate user tags, but after these tags are generated, they are often easily overlooked and not maintained. In fact, tags also have a life cycle, from demand proposal, to generation, to approval, to execution.

Therefore, the label system must also have a clear update rule, including:

(1) Label update cycle: real-time update, monthly or three-month update, etc.;

(2) Label update dimension: under what circumstances is the update of a specific user triggered, such as under what circumstances is the risk rating of a certain type of user updated;

(3) Tag update permissions: who can update this tag library;

(4) Elimination of useless tags: For example, there are only 70 tags that will be used in the tag library, but there are 100 tags in the tag library. Then the other 30 tags are invalid tags and are purely occupying resources. We can delete these tags.

Label maintenance is a very important and systematic project, which should be taken seriously.

The above is the basic process of building a user tag system. Next, we will discuss how to apply user tags to achieve refined operations.

5. How to apply the user tag system

We imagine how labeling will reshape marketing scenarios from the perspectives of user lifecycle management, personalized push, data analysis and monitoring.

1. Automated management of user lifecycle

Based on user behavior and status characteristics, a user life cycle management model is developed to monitor changes in users at different stages and conduct corresponding marketing activities for users in different cycles. This is the primary application of the label system.

But in fact, changes in marketing strategies can never keep up with the speed of user interest migration. The sources and exit nodes of users are dynamic. Therefore, the process of labeling users should also be dynamic. Once we set a label condition, all users who meet the condition will be automatically included in this label and enter the marketing activities for this type of labeled people.

Image source: Digital Business Cloud MA

Such automated management allows operations personnel to grasp user changes in real time, freeing them from complicated data processing work and truly unleashing marketing creativity.

2. Personalized marketing campaign configuration

The establishment of a label system lays the foundation for personalized marketing activities. Through multi-dimensional user tags, we can accurately grasp the needs of different groups and adopt different types of personalized push, personalized recommendation, personalized real-time marketing and other operational strategies and service methods for different target groups, so as to realize the exploration and satisfaction of users' personalized demands and improve user experience .

(1) Personalization of push channels

Based on the interactive feedback from users of mainstream reach channels such as SMS, email, APP push, and public account template messages, information is pushed on the channels preferred by users, improving the experience while saving push costs.

Image source: Digital Business Cloud MA

(2) Personalized push content

Based on the user's historical order/browsing data, personalized recommendations based on the user's consumption/interest preferences are implemented, and some strong user personal attributes are appropriately added to the copy to increase the intimacy of interaction based on the thousands of faces of each person.

(3) Personalized push time

Based on the user's historical click/purchase time, determine the time when the user is most likely to be touched by marketing, and send marketing information to the user at the prime time.

Image source: Digital Business Cloud MA

3. Data analysis and monitoring

After completing precision marketing and refined operations, companies can monitor the operation of various labels in real time and view marketing results in all aspects and in real time. Based on the conversion, effect and impact of the tag user operation strategy, we continuously optimize the strategy to make user operation truly personalized, precise and individual.

Image source: Digital Business Cloud MA

In summary, the scientific, efficient, flexible and self-service visual management label system has greatly reduced the threshold for label creation, helping customers to achieve refined customer lifecycle management, personalized push, marketing monitoring and other applications.

VI. Conclusion

In the era of big data, the most direct way to interpret users is to build a complete user labeling system. Although the construction of user tags is very cumbersome and requires unraveling the business and responding to various changes, it has a far-reaching impact on the development of the enterprise. The refined operation of the enterprise is inseparable from the construction of a tag system.

Author: Activity Box

Source: Activity Box Operation Society