How to complete user segmentation? Teach you 4 methods

When an Internet product has a large number of users, users are generally stratified in order to better refine operations based on user characteristics. The author of this article introduces 4 methods of user segmentation for your reference and learning.

Why do almost all Internet products perform user stratification when the number of users is large?

This is mainly because, when the number of users is large, the contradiction between the personalized needs of thousands of people and the universal operating strategy will become the main contradiction of the current product. Differences in user characteristics lead to personalized demands and also give rise to the need for refined operations.

When it comes to user segmentation, these words may come to mind: core users, seed users, paying users, free users, active users, churned users, content creation users and content consumption users.

For example, TikTok users can be roughly divided into content creators and content consumers, and of course the two can be further subdivided.

The core interest of a product is profit. Whether this profit is money or traffic, it is necessary to segment users, adopt different operating strategies for users at different levels, and conduct refined operations to maximize the utilization of operating resources and maximize product benefits.

The essence of user segmentation is a means of refined operation that segments users based on user characteristics, user behaviors, etc.

The following are four common methods of user segmentation.

1. User value segmentation and stratification

User value segmentation and stratification are divided into two dimensions: first, user value segmentation based on user life cycle definition; second, user value segmentation based on user key behaviors .

The definition of user life cycle must be related to the user value growth path. Different types of products have different value growth paths. According to whether the product is paid and the frequency of product use, the value growth paths of different products can be divided into four quadrants:

Everyone should be familiar with the user life cycle model. Generally, the user's life cycle is divided into five stages. Of course, this does not mean that every user will go through the complete life cycle. This is just a general model.

(1) Introduction stage : After the user registers, he or she is just getting started and is not familiar with the product or the value it can bring. The definition in terms of data is that the user has just registered and has not yet experienced the core functional process (the core functional process needs to be defined in advance and buried for statistics)

(2) Growth stage : Users have a certain understanding of the product, recognize the user value it provides, have established initial usage habits, and will use the product regularly. The data definition is that the core functional process has been experienced, and the usage frequency and usage time are greater than or equal to the defined minimum threshold, for example, logging in three times a week and using the system for 10 minutes each time.

(3) Mature stage : Users have formed a high degree of dependence and habit on the product, with significantly high frequency and duration of use, and can contribute relatively high value. The data positioning is that the usage frequency and usage duration are greater than or equal to a certain threshold (determined by the product), or the payment frequency and value reach a certain threshold

(4) Dormant stage : They were once mature users, but now no longer visit or use the product, or their visit frequency is decreasing. The data definition is that the product has not been used for more than 10 days (specific analysis will be conducted based on specific circumstances).

(5) Churn period : users who have not logged into the product for a long time or have even uninstalled the product. The data definition is users who have not used the product for more than 30 days (customized).

Another way to segment and stratify user value is to divide them based on their key behaviors. The most typical and commonly used method is the RFM method. RFM represents three key user behaviors:

1. R (Recency), distance from the most recent transaction
2. F (Frequency), transaction frequency
3. M (Monetary), transaction amount

The RFM method can be used to divide users into 8 types

(The picture is not very high-definition, please make do with it)

We need to analyze the RFM value of each user and then determine the type of user. The stratification using RFM method is mainly divided into the following steps:

To crawl data, you just need to set the crawling principles of the three dimensions of RFM, and then ask the developer or data analyst for help.

There are three common methods for defining the median of the three dimensions of RFM:

1. The mean or median of all data
2. Based on the important value of a business node, such as the R value of investment and financial management, it is generally 1 month, because there is money to invest only when the salary is paid
3. According to the 80/20 rule, 80% of users are concentrated in the low-frequency and low-amount range, and 20% of users are concentrated in the high-frequency and high-amount range.
4. Means clustering algorithm, it would be best if the data analyst knows this

I will not go into detail here about how to conduct data analysis. There are many articles on the Internet that teach you how to use RFM.

The core logic of the RFM method is to find out the key behaviors that affect user value, and then conduct cross-analysis and user segmentation . Therefore, the RFM model does not necessarily have the meaning mentioned above, and it can have different definitions in different fields. For example:

1. In the financial field, R represents the time of the most recent investment, F represents the investment frequency, and M represents the investment amount;
2. In the live broadcast field, R represents the time of the most recent live broadcast viewing, F represents the viewing frequency, and M represents the total viewing time;
3. In the gaming field, R represents the time of the last game played, F represents the frequency of the game, and M represents the duration of the game. It can also be defined as: R represents the time of the most recent game recharge, F represents the recharge frequency, and M represents the recharge amount.

RFM simply represents a hierarchical way of thinking. For any product, we can define the key behaviors that affect users, then define the indicators of these behaviors, and then cross-analyze these indicators to complete the user stratification.

2. AARRR Model Layering

Those who have heard of growth hacking must be familiar with this model. The AARRR model can not only be used for growth, but also for user segmentation.

1. Get user: download without registration, or complete registration but no further action. During this stage, we should pay attention to the registration conversion rates of different channels and optimize the allocation of channel resources.
2. Improve activity: Registered but not completed the core process experience of the product. At this stage, it is necessary to strengthen the guidance of users to complete the core process.
3. Improve retention rate: Users have experienced the core process, but the retention time is not high. Analyze the retention problem and then provide specific operational strategies.
4. Viral spread: users whose activity frequency exceeds a certain threshold. Stimulate users to spread the word through tool optimization
5. Obtaining income: For users whose activity and retention time exceed a certain threshold, we will strengthen the guidance of payment for specific users and combine it with specific scenarios.

AARRR is a relatively rough user stratification model, which is suitable for the early stages of a product. At this stage, the number of users is neither large nor small, and the company’s data system may not yet be established .

3. User identity segmentation and stratification

When talking about user identity, the first word that comes to everyone’s mind is KOL. In content communities, users are generally divided into at least two categories: KOLs and ordinary users. For these two types of users, the operation strategies must be different.

Only when the behavioral characteristics and demands of users in the product field vary greatly, it would be more appropriate to use identity segmentation for stratification. For example, Weibo can be divided into at least: celebrity users, KOL users, active users and ordinary users.

How to sort out the user identity segmentation model of the product? Ask yourself three questions:

1. Are there relationships between users?
2. Will a certain type of user form a user class based on the scarcity of the content they contribute?
3. Can users achieve class advancement in a natural way?

If there are no relationships between users, then the identity segmentation hierarchical model does not apply.

If there is a relationship and user stratification occurs due to contribution or scarcity, then a user stratification model should be built based on contribution or scarcity.

If users of different levels can advance naturally, then build a user stratification model based on the advanced level.

4. User Demand Segmentation and Stratification

User demand segmentation and stratification are mainly divided into two dimensions:

1. User attributes mainly rely on basic user data, including gender, age, occupation, income, etc.
2. User personalized needs mainly depend on user behavior data, personal consumption preferences, and personal scenario preferences.

Therefore, user demand segmentation and stratification is mainly achieved by analyzing whether users have obvious differences in their needs in these two dimensions. There are two ways to judge: experience insight and data.

In specific operations, you can use a single dimension for differentiation, or you can use two dimensions for cross-analysis.

Choose a dimension for differentiation . For example, the typical product Mayu will push different content to users in different states: preparing for pregnancy, pregnant, or hot mom. For example, matchmaking products will segment and stratify user needs according to different age groups and genders.

Select two dimensions for cross-analysis . For example, for shopping products, different products will be pushed to users based on their gender, age, and consumption preference attributes.

The main purpose of user segmentation is to facilitate subsequent refined user operations, and the ultimate goal is to maximize product revenue at the lowest operating cost.

The two core elements of user stratification are: first, users at different levels can be defined with clear data labels and attribute labels, so that user labeling automation can be achieved; second, the user operation strategies at different levels are targeted and stable.

Four common methods for user segmentation:

1. User value segmentation and stratification, including user life cycle and RFM method
2. The AARRR model is suitable for the early stages of a product and is a simple and rough stratification method.
3. User identity segmentation and stratification is suitable for products where users are connected and have obvious class segmentation due to contribution or scarcity.
4. User demand segmentation and stratification, in simple terms, refers to whether user demand for products varies due to different user characteristics.

Author: Jarvan

Source: Jarvan