User Classification Move Metrics¶
Overview¶
Based on game activity and purchase data, this feature automatically distinguishes activity strength and purchase strength through K-Means clustering, and defines user classification types accordingly.
It provides movement status by type based on the classification type at the first access date and last access date among users who accessed during the selected period, allowing you to check user usage flow.
- This metric is updated every day at 7 AM Korea Standard Time (KST).
- It can be checked through SDK integration without separate log transmission.
What Can You Do?¶
Business/Marketing Users¶
- You can quantify churn flows such as whale -> dolphin or light -> non-access during the selected period and decide when to launch reactivation campaigns.
- You can immediately identify the most common user movement patterns through the most frequent classification movements.
Data Analysts¶
- You can visually explore movement patterns at the individual user level with a dot movement map showing user distribution before and after classification moves.
- By changing the period, you can compare user type changes before and after specific events or updates.
Quick Start¶
- In the left menu, go to User > User Classification and click the user classification move tab.
- Set the Project and Period to analyze.
- In the User Classification Type Movement table, check the number and ratio of users moved from the initial classification type to the final classification type.
- Explore individual user movement paths visually in the dot movement map.
Full Features¶
Key Concepts¶
| Concept | Description |
|---|---|
| Activity Strength | A user's game engagement measured based on access and play data for the 3 days including and before the reference date. Divided into 4 levels: high / medium / low / new |
| Purchase Strength | A user's spending tendency measured based on cumulative purchase data from the first access date to the reference date. Divided into 4 levels: high / medium / low / non-paying |
| Initial Classification Type | The user classification type based on the first access date within the selected period |
| Final Classification Type | The user classification type based on the last access date within the selected period |
| Non-Access | Users whose last access was 3 days before the end date of the selected period (dormant for 3 or more days) |
Metric Terms¶
Activity Strength¶
A user's activity strength is measured based on data for the 3 days including and before the reference date, using the user's access date as the reference date.
Example
Example: For a user who accessed on January 10 -> measurement uses data from January 8 to January 10 (3 days)
Activity strength consists of 4 levels (high, medium, low, new), and the measured items are as follows. However, if activity strength is new, the user is a new user on the reference date regardless of the items below.
- Number of logins in the 3 days including and before the reference date
- Number of access days in the 3 days including and before the reference date
- Daily average number of logins in the 3 days including and before the reference date (login count / login days)
- Total game playtime in the 3 days including and before the reference date (seconds)
- Average game playtime by hour from 00:00 to 23:00 in the 3 days including and before the reference date (seconds)
- Number of rewarded ad views in the 3 days including and before the reference date
- Whether a push was opened in the 3 days including and before the reference date
Purchase Strength¶
A user's purchase strength is measured based on cumulative data for the full period from the user's first access date to the reference date, using the user's access date as the reference date.
Example
Example: If a user who accessed on January 10 first accessed on January 1 -> measurement uses data from January 1 to January 10
Purchase strength consists of 4 levels (high, medium, low, non-paying), and the measured items are as follows. However, if purchase strength is non-paying, the user has no purchase history from the first access date to the reference date regardless of the items below.
- Time from the user's first login to first purchase (days)
- Total number of payments made by the user from first login to the reference date
- Average payment amount per payment from first login to the reference date (user total payment amount / user total payment count)
User Classification Type¶
User classification types are defined based on activity strength and purchase strength, and when criteria overlap, users are classified into the higher-priority type.
| Type | Classification Rule |
|---|---|
| Whale User | Both activity strength and purchase strength are "high" |
| Dolphin User | Either activity strength or purchase strength is "high" |
| Middle User | Either activity strength or purchase strength is "medium" |
| Light User | Both activity strength and purchase strength are "low" |
| Non-Paying User | Purchase strength is "non-paying" regardless of activity strength |
| New User | Activity strength is "new" regardless of purchase strength |
Metric Details¶
User Classification Type Movement¶
Targets users whose classification type on the first access date differs from the classification type on the last access date.
Example
Example: If the period is set to January 1 to January 10, and a user who accessed on January 1, 3, 5, and 10 is classified as whale on January 1 and dolphin on January 10 so that the type changes, they are included in the display target. If they remain whale on January 10 and the type does not change, they are excluded.
You can check the number and ratio of users who moved from the initial classification type (based on the first access date) to the final classification type (based on the last access date) during the selected period.
- User Count: The number of users classified as type A at the initial classification time who moved to type B at the final classification time.
- User Ratio (%): The ratio of users who moved to type B at the final classification time relative to the number of users of type A at the initial classification time.
Example
Example: If in the period January 1 to January 10, out of 1,000 whale users at the initial classification time, 100 users moved to dolphin at the final classification time, it is displayed as 100 users (10%).
The Non-Access type in the final classification type refers to users dormant for 3 or more days, meaning users whose last access is 3 days before the end date of the selected period.
Example
Example: If the period is set to January 1 to January 10, users whose last access was before January 8 are classified as non-access.
Most Frequent Classification Movement Status¶
In the user classification type movement table, you can check the classification movement type with the maximum user count among movements excluding those whose final classification type is non-paying or non-access.
User Distribution Before Classification Move & User Distribution After Classification Move¶
The colored dot movement map lets you check how each user's initial classification result moved to the final classification result during the selected period. 
- Users whose classification movement result is non-paying or non-access (dormant for 3 or more days) are shown in the user distribution before classification move graph, but not in the user distribution after classification move graph.
- Each dot in the graph represents one user whose classification moved during the selected period, and the color of each dot matches the user classification type at the initial classification time.
- If you hover over each dot in the graph, you can check the user's initial classification type and final classification type.
Notes & Tips¶
- The classification movement metric is aggregated only for users whose first access date and last access date have different types. Users with no type change during the period are not shown.
- The non-access type is determined based on the end date of the selected period. The longer the period, the higher the proportion of non-access users may be.
- Since the metric is updated every day at 7 AM (KST), real-time data for the current day is not reflected.
- If you find a type with concentrated movement patterns, combine it with User Classification Metrics to check the characteristics of that user group.
Related Menu¶
- User Classification Metrics - Check user classification definitions and activity/purchase strength criteria
- Segment - Create segments based on user classification and use for targeting campaigns