Flows Analysis
Flows Analysis is an exploratory analysis that visually presents the sequence of user behaviors and key nodes in a Sankey diagram. With Flows Analysis, you can view how behaviors flow in and out before and after each key node, quickly identify the primary and secondary factors that affect conversion, and improve your product accordingly. Flows Analysis helps you answer questions such as:
- What is the golden path users take through the product, from their first page view to completing a purchase?
- What is the first core event users participate in after they register?
- What is the last core event users participate in before they churn?
Define paths and configure analysis conditions
Flows Analysis is based on sessions, so before you explore user paths, you need to understand how a session is defined. First, select the events to analyze. You can select up to 30 events as the nodes that make up a session. Adding filter conditions to events isn't supported. For the selected events, you can add an event breakdown. After the breakdown, different property values of the same event are treated as different nodes. After configuration, you can select one of these events as the initial event, that is, the starting node of a session. Starting from the starting node, the Flows Analysis model keeps looking for the next node. If no next node is found within the session interval, the session ends. After a session is interrupted, a new session starts the next time the user triggers the initial event, so the same user may generate multiple sessions within the selected time range. Note that even if a user triggers the initial event again during a session, that event is only a node in the session and doesn't start a new session. You can set the session interval, with a maximum of 24 hours and a minimum of 1 second. The same behavior sequence is divided into different sessions depending on the session interval setting. For example, a user's behavior sequence is as follows (A is the selected initial event, and A1, A2, and A3 indicate that the user triggered it three times):
| Event | Event time | Time between adjacent nodes (minutes) |
|---|---|---|
| A1 | 2023-01-01 10:00:00 | |
| B | 2023-01-01 10:05:00 | 5 |
| A2 | 2023-01-01 10:15:00 | 10 |
| C | 2023-01-01 10:30:00 | 15 |
| D | 2023-01-01 10:50:00 | 20 |
| A3 | 2023-01-01 11:20:00 | 30 |
| B | 2023-01-01 11:40:00 | 20 |
Depending on the session interval you select, the sessions are divided as follows:
| Session interval (minutes) | Number of sessions | Session details |
|---|---|---|
| 5 | 3 | A1-BA2A3 |
| 10 | 2 | A1-B-A2A3 |
| 20 | 2 | A1-B-A2-C-DA3-B |
| 30 | 1 | A1-B-A2-C-D-A3-B |
You can also set an event as the ending event, which is commonly used to analyze what churned users did before they churned. The session traces back from the ending event, continuously looking for the previous node. If no node is found within the session interval, the session is interrupted.
When you use custom events to build paths, the same underlying event may belong to multiple events. To keep the attribution logic clear, check Events Priorities.
All events in Select Events appear on the right side of the pop-up shown above. When a session contains an underlying event that belongs to multiple events, the underlying event is attributed to whichever of those events is higher in the order you specify on the right side of the pop-up.
As with other analysis models, you can also filter data. Flows Analysis supports filtering by user properties, user tags, or cohorts.
View user inflow and outflow in the Sankey diagram
A session consists of multiple steps. After all sessions of all users are aggregated, you can view the overall picture. The Sankey diagram shows the data of up to 10 steps, going forward from the initial event (or backward from the ending event).
In each step, each rectangle represents a different node, and the height of a rectangle represents the number of sessions that pass through that node. Each step shows only the 7 nodes with the most sessions, and the data of the remaining nodes is merged into More.
You can click any node in a step to view detailed information. In the node details, you can see the number of sessions that pass through the node at step N and the distribution of subsequent nodes. If there is no subsequent node, the session is counted as Churn. You can click the user details to go to the User List, or save these AE user IDs as a result cohort.
The lines between step nodes show what proportion of the sessions that pass through node A go through node B in the next step. You can click a node to highlight the session paths that pass through it and see the flow more clearly. For example, if you select a set of core gameplay events for analysis and use registration as the initial event, the data shows that 60% of sessions flow through "Beginner Event" after the registration event, which indicates that it is the gameplay on the UI that attracts new users the most. Beyond that, 13% of sessions go straight to Gacha after registration, so you can further analyze whether these users are familiar with card games and used to raising their combat power through Gacha.