Quick Tracking:Retention Analysis

Overview

Retained, the general meaning is that after the user starts using the product, after a period of time, still continue to use the product.

Retention analysis can customize the initial behavior and subsequent behavior for retention calculation according to different business scenarios and product stages, help enterprises analyze the stickiness of users using products, adjust strategies according to retention analysis results, guide users to discover product value, retain users, and realize real user growth. After query through retention analysis, you can save it as a report and add it to the self-made dashboard for display and statistics.

Applied scenarios

Retention analysis can solve

• Measure the change of product stickiness to users over time, and verify whether the user target group has completed the estimated behavior event within the expected time, such as observing the retention of new users after the activity is launched for 1 day or 7 days to determine the effect of the activity.

• Observe the continued attraction of features that are valuable to the product. For example, choose the initial behavior to "play a short video" and the return event to "start" to understand the retention of the user's second opening of the app.

Take 「Statistics on User Retention from Different Sources」 as an example

1. Select analysis subject: login user

2. Custom retention: select the 「Application Start」 event for both initial and subsequent behaviors

3. Select 「Source Types」 grouped as 「Channel Attributes」

4. Select time range

5. Click 「Start Analysis」

Actions guide

Page composition

The retention analysis feature consists of the following components:

1. Historical query list area: Users can click to select the list of historical saved queries.

2. Information configuration area: the user can select events, indicators, attribute settings, grouping and time selection.

3. In the Analysis Results section, you can view the visualization charts and detailed data of the analysis results.

Select analysis subject

In the retention analysis module, in addition to user behavior concatenation by device ID, user behavior can also be concatenated by logon user ID and entity ID to analyze user behavior. The drop-down list includes "device ID", "account ID" and "entity ID". By default, "device ID" is selected.

• Device ID: the unique ID generated by QuickTracking for each device

• Account ID: the unique ID of the user account that you specify by using the API.

• Entity ID: the user ID that is generated by Quick Tracking. You can use ID-Mapping to associate the device ID with the account ID on a one-to-one basis. This way, you can connect the account before and after logon.

Custom retention behavior

1. Click the drop-down to select specific events as the initial behavior and subsequent behavior.

2. Add filter conditions:

• Supports property filtering for events.

  Category name	Category description	Description
  System attributes	These attributes act on all events and is tracked by the SDK	System preset properties, such as the version of the SDK
  Global attributes	These attributes are applied to all events and is reported by users, such as the region where you are hungry.	The attributes entered by the user in the global attributes of the tracking management platform.
  Event attributes	These attributes are applied to all events, which are reported by users or tracked by the system. For example, the payment amount of a payment event	Attributes entered by the user in the event attributes of the tracking management platform

• Supports different filtering symbols based on different types of attributes

  Attribute type	Filter symbol	Symbol definition	Example
  String	Equal to	Represents equal to one or more specific values	The device brand is equal to Huawei or Apple
  	Not equal to	Indicates that one or more specific values are excluded	Device brand excludes Huawei or Apple
  	Contains	Find values that contain certain characters in an attribute value	The character Huawei is included in the device brand
  	Does not contain	Exclude values with certain characters in attribute value	The character Huawei is not included in the device brand.
  	Empty	Find data without attributes	No device brand attribute value
  	Not empty	Find data with property values	with device brand attribute value
  Numeric types	Equal to	Equal to a specific value	Order amount equals 1000
  	Not equal to	Not equal to a specific value	Order amount not equal to 1000
  	Greater than	Greater than a specific value	Order amount greater than 1000
  	Smaller than	Less than a specific value	Order amount less than 1000
  	In... with...	Between two specific values	Order amount between 1000 and 5000

Select an attribute group

For retention analysis, you can select attributes to group metrics. For example, if you want to track statistics on the retention of subsequent payment behaviors after users from different channels register, you can also use the two attributes for cross analysis.

Group computational logic description

Compute without grouping:

• The date within the selected time is tiled, and the retention calculation is performed for the triggered user on each day.

• For example, on July 4, 2023, the number of people who triggered the 「Initial Behavior」 application startup (preset) device ID (analysis subject) is 2012, and on July 5, the number of people who triggered the 「Subsequent Behavior」 application startup (preset) device ID (analysis subject) is 956, and the retention rate is 「Subsequent Behavior」 number (retention) /"initial behavior" number. The percentage of the number of 2012 who triggered the 「Initial Behavior」 after 1 day is 47.51% (retention rate);

• And so on 2, 3, 4, 5, 6 days after the retention rate.

Calculate with grouping:

• Grouping: For example, within the calculation time (20230704-20230712), the 「Initial Behavior」 is triggered and the grouping value is equal to "IOS";

• Number of users: the number of analysis subjects (device IDs) that trigger the 「Initial Behavior」 and whose group value is equal to "IOS" within the calculation time (20230704-20230712);

• For example, within the calculation period (20230704-20230712), the number of people (retained) and percentage (retention rate) of these analysis subjects (device IDs) that trigger the 「Initial Behavior」 application startup (preset) and group value equal to "IOS" and the "subsequent behavior" application startup (preset) and group value equal to "IOS" one day later.

• And so on 2, 3, 4, 5, 6 days after the retention rate.

• The difference between grouping and non-grouping is that the date after 1 day or 2 days is not a specific date when grouping is available. For example, in the 20230704, a device A triggers the 「Initial Behavior」 and meets the grouping value, and in the 20230706, the 「Subsequent Behavior」 is triggered again and meets the grouping value, then it is classified as retained after 2 days. In the 20230707, a device B triggers the 「Initial Behavior」 and satisfies the grouping value, and 20230709 triggers the 「Subsequent Behavior」 again and satisfies the grouping value, it will also be classified as retained after 2 days.

Add a global filter

If you select two step event metrics or more, global filtering supports common filtering based on common attributes between different events. The specific filtering capability and the attribute filtering capability of the same single event are set.

Add user cohorts

When you need to look at the data of a specific group of people, you can use the 「Filter User Group」 to achieve. For more information about how to create a user group, see Audience Insight.

Select time range

You can select the time range and time granularity as required. You can select a time range in the 「Relative Period」 or 「Fixed Period」 mode. In the Relative Period mode, the default time is set to the past seven days and is displayed by day.

• The 「Relative Period」 of time is based on the date range pushed forward by an anchor point and will change over time. It has three dimensions: the past X days, weeks, and months. You can also customize the time filter conditions for the past X days, weeks, and months. The day is a complete natural day, and the week is selected from Monday to Sunday, and the month is the natural month (from the 1st to the last day of each month)

The following list describes the rules:

A. Past n days: Push forward the complete n days based on the current time.

B. Past n weeks: Push forward n complete weeks based on the current time. If the current time is the last day of the week, the past n weeks include the week in which the current time is located. Example: If the current time is the 7.20 (Tuesday), then the past week is 7.12-7.18 (Monday to Sunday). If the current time is the 7.18 (Sunday), then the past week is 7.12-7.18.

C. Past n Months: Push forward n complete months based on the current time. If the current time is the last day of the month, then the past n months include the month in which the current time is located. Example: If the current time is 7.20, then the past month is 6.1-6.30; If the current time is 6.30, then the past month is 6.01-6.30.

• 「Fixed Period」 You can directly select the start date in the calendar box, and click OK to select the current time range for data analysis (the maximum selection range of a fixed period is 366 days)

View analysis charts and detailed data

After you set query conditions and click 「Start Analysis」, you can view the analysis results and retention trend charts.

• Click 「Export Data」 in the upper right corner to download the Excel file.

• Save common metrics for subsequent re-query. You can click the 「Save」 button in the upper-right corner.

• You can click the toggle in the upper right corner to view the retention rate after T days (the rate at which users have subsequent behaviors on day T after the initial behavior)

Detailed Data

In the detailed data, there are detailed data of statistical results, and the retention table view supports switching between 「Retention Number」 and 「Retention Rate」. The Number of Users column shows the number of deduplication devices/login users that have the initial behavior.

Save data to report

1. Enter a report name

2. Select the time period for saving the report.

-「Impact of Different Time Periods on Reports」

• Select a relative time period. The report date can be queried according to the time selected on the dashboard.

• Select a fixed period, the report date will not change according to the dashboard selection time

• No period is selected, the report date follows the time selected on the dashboard

3. Click 「OK」 button to save the data to the report list.

Add a report to the dashboard

For an already saved report, you can choose to add the report to the dashboard:

1. Enter 「Report Name」

2. Select the added dashboard

3. Select the type of display you want the chart to display.

4. Select the display layout of the chart in the self-made dashboard

5. Advanced settings can be selected according to the dashboard rules to be displayed.

6. Click 「OK」

Retention analysis computational logic

1 . Custom retention meaning

In Quick Tracking, N-day retention is used, that is, to analyze whether a user who triggers an event on a certain day /week /month triggers an event after N days /weeks /months, and the triggered user is recorded as a retained user.

For business goals, you can further define the retention of users and clarify the initial events and return visits of users.

• Initial event: Only devices or logon users who have triggered this event on the current day can participate in subsequent custom retention calculations. You can set specific event attributes and attribute values through the conditional filtering function to delineate more detailed user groups.

• Return visit event: A user is counted as a retained user only after a return visit event is triggered. You can use the conditional filtering feature to set specific event attributes and attribute values to set stricter return visit behavior.

Note: If a user triggers multiple initial events and return visit events, only one count is counted on the same day, week, or month, which is the number of deduplication devices /logon users.

2. Confirm the retention interval

The time range is limited to the date when the initial behavior occurs. In Quick Tracking, you can view the retention status of users 1,2,...,7, 14, and 30 days after the date when the initial behavior occurs.

For example, suppose the initial behavior is defined as the X event and the return visit behavior is the Y event. During the selected time range from 05.01 to 05.08, user Xiao A has an X event in 5.01, and the following events occur every day:

When calculating the retention of the 05.01 number, user Xiao A will be counted as the retention users after 2 days, 3 days, 6 days and 7 days.

3 . Set conditions for retention

We can use different filtering conditions to limit retention. The filtering logic is as follows:

• You can filter events by event filter or global filter.

• Global filters act on initial and return events

• If an event filter condition and a global filter condition are configured, the filter condition logic is: event filter condition&global filter condition

Example: Observe the retention of cosmetics A re-purchase by female users in the age stage of "18-35" in Zhejiang Province. Select the initial behavior as "successful payment" and the subsequent behavior as "successful payment". You can select the filter conditions "province=Zhejiang&age group=18-25 &product ID=cosmetic A".

4. Comparative analysis of retention

Add a group filter to the query condition to compare the comparison retained under different group values

• Supports preset attributes and global attributes to group retention.

• Group conditions act on initial and return events

• If a user meets multiple grouping conditions within a time range, it is classified into multiple groups.

5. Interpret the retention metrics

Initial behavioral users: the number of deduplication devices and logon users whose initial behavioral users occurred on a given day, week, or month

Retained Number: the number of deduplication devices /logon users that have undergone subsequent actions after N days, weeks, and months.

Retention rate: refers to the proportion of retained users who have subsequent behaviors after N days /weeks /months in the initial behavior users.