A method for evaluating and classifying the value of active APP users based on the RFM model

Labs Guide

With the rapid development of the Internet, various products emerge in an endless stream, and the competitive pressure of products is gradually increasing. Users' requirements for Internet products are becoming increasingly stringent, and their attention to user experience is also increasing. Only products that meet user habits and are recognized by users will have an advantage in the competition. Therefore, it is very important to carry out refined management and operations for different types of users. The prerequisite for refined user operations is to classify users using appropriate standards, and then divide users into different groups, and then specify differentiated operation strategies for users in different groups. The RFM (Recency Frequency Money) model introduced in this article is a user grouping method commonly used in data analysis. It uses RFM to complete user grouping, and then implements different operation strategies for different users to achieve refined user operations.

Part 01 Overview of RFM Model

The RFM (Recency-Frequency-Monetary) model is an important tool and means to measure customer value and customer profitability. It is widely used in consumer products, where R represents the time interval between the last consumption, F represents the number of consumptions within a specific time range, and M represents the total consumption within a specific time range. For non-consumer products, this definition is no longer applicable. From the perspective of user activity and behavior, the RFM model can be transformed to define R as the number of days between the last active consumption, F as the number of active days within a certain time range, and M as the value score within a certain time range. The value score can be calculated using the user's important behaviors in the product. Using the three indicator dimensions of R, F, and M, users can be divided into eight value types: important value users, important development users, important retention users, important retention users, general value users, general development users, general retention users, and general retention users. These can be used to perform refined operations on users. The overall process is shown in Figure 1.

Figure 1 Overall process of user value assessment and classification method

Part 02 User data acquisition module

The client login interface is used to report data and count the details of user active data. The client tracking interface is used to report data and count the details of user access behavior data on the client. User active data and user behavior data are aggregated to obtain user data sets. The specific acquisition process is shown in Figure 2.

Figure 2 User data acquisition process

Part 03 RFM indicator calculation module

The calculation of RFM index is shown in Figure 3. After obtaining the user data set (user active data, user behavior data), the user active data is used to calculate the number of days between the user's last active time and the number of times the user is active within a specific time range (nearly N days). Using user behavior data, calculate the user's usage of specific functions within a specific time range (nearly N days). The functions here can be customized according to product features (as shown in Table 1). For example, the functions can be generally divided into two categories: core functions and additional functions. The core functions include commenting, liking, forwarding and sharing, adding to shopping carts, placing orders, etc. Additional functions include participating in lucky draws, participating in check-in activities, participating in user surveys, etc. The type and quantity of functions are adjusted according to the actual product situation and operation goals.

The R index score is calculated using the number of days between the last active user and the rules shown in Table 2 (the score division rules here can also be adjusted according to actual conditions). When the number of days between the last active user is 21 days or more, the score is 0; when the number of days between the last active user is 11 to 20 days, the score is 20; when the number of days between the last active user is 6 to 10 days, the score is 60; when the number of days between the last active user is 4 to 5 days, the score is 80; and when the number of days between the last active user is less than 3 days, the score is 100.

The number of times a user is active within a specific time range (nearly N days) is used, and the F index score is calculated using the rules shown in Table 3 (the score division rules here can also be adjusted according to actual conditions). When a user is active 1 to 2 times, the score is 20 points; when a user is active 3 to 5 times, the score is 40 points; when a user is active 6 to 8 times, the score is 60 points; when a user is active 9 to 15 times, the score is 80 points; when a user is active more than 16 times, the score is 100 points.

Using the user's usage data on specific functions within a specific time range (nearly N days), the M indicator score is calculated using the rules shown in Table 4 (the score division rules here can also be adjusted according to actual conditions). When the user uses i core functions and j additional functions, the score for this item is (80×i/I)+(20×j/J), and the full score for this item is 100 points.

Figure 3 RFM indicator calculation process

Table 1 Function usage examples

Table 2 R index scoring rules

Table 3 F index scoring rules

Table 4 M index scoring rules

picture

Part 04 RFM model building module

As shown in Figure 4, after the RFM index is calculated, the user's RFM feature set is formed. In the RFM model construction module, the three indicators R, F, and M need to be used to divide users into different categories. In this part, the threshold of each indicator needs to be calculated first. Based on the RFM feature set of the entire user, the mean of the three indicators R, F, and M of all users is calculated, and the mean is used as the threshold of each indicator.

Figure 4 RFM indicator calculation process

Part 05 User value assessment and classification module

After the RFM model is built, the model can be used to classify users. The user classification criteria are shown in Table 5:

(1) When the user's R, F, and M indicators are all high, the user is classified as an important value user;

(2) When a user's R and M indicators are high but the F indicator is low, the user is classified as an important development user;

(3) When the user's F and M indicators are high and the R indicator is low, the user is classified as an important retention user;

(4) When the user's M index is high and the R and F indexes are low, the user is classified as an important retention user;

(5) When the user's R and F indicators are high and the M indicator is low, the user is classified as a general value user;

(6) When the user's R index is high and the F and M indexes are low, the user is classified as a general development user;

(7) When the user's F index is high and the R and M indexes are low, the user is classified as a general retention user;

(8) When the user's R, F, and M indicators are all low, the user is classified as a general retention user.

Table 5 User value classification rules

From the classification rules, we can see that: the M index determines whether a user is an important user; when the user's usage interval is short and the frequency is high, the value is high; when the user's usage interval is short and the frequency is low, the user development operation strategy can be focused on; when the user's usage interval is long and the frequency is high, the user retention operation strategy can be focused on; when the user's usage interval is long and the frequency is low, the user may be lost, and the retention operation strategy needs to be focused on. By using the user value evaluation and classification results, targeted operation strategies can be formulated for different types of users to achieve refined user operations.