Behavior Sequence Analysis
Behavior Sequence (BehaviorSequence) queries the behavioral timeline of a specific individual user — a user-level drill-down investigation tool. It corresponds to the backend AnalyticsService.BehaviorSequence.
1. What Questions It Answers
- What has a specific user recently done (e.g. a user who filed a complaint, or one with abnormal payments)?
- What is the order of a user's behaviors within a given session?
- Troubleshooting: the user claims "I clicked but nothing happened" — which events did they actually trigger?
2. Difference from Path Analysis
| Model | Granularity | Perspective |
|---|---|---|
| Behavior Sequence (BehaviorSequence) | A single user | Investigating individual cases, seeing what one person did |
| Top Conversion Paths (PathSankey) | A group | Mining mainstream paths |
3. Backend API
gRPC: AnalyticsService.BehaviorSequence
| Field | Type | Description |
|---|---|---|
time_range | TimeRange | Analysis time range |
user_id | uint32 | Target user ID (required) |
app_id | uint32 (optional) | Filter by app |
event_name | string (optional) | Event name filter (empty = all events) |
limit | uint32 (optional) | Maximum number of events returned, default 100 |
Response BehaviorSequenceResponse: user_id, events[] (SequenceEvent, sorted by time ascending).
Each SequenceEvent contains: timestamp / event_name / session_id / session_seq / referer / platform / channel.
HTTP (admin proxy)
POST /admin/v1/analytics/behavior-sequence
Content-Type: application/json
{
"timeRange": { "startMs": 1718169600000, "endMs": 1718774399000 },
"userId": 1001,
"limit": 50
}
4. Interpreting Results
The returned events are sorted by timestamp ascending — a single timeline. Combined with session_id + session_seq, you can reconstruct the user's exact order of actions within each session:
- Events with the same
session_idbelong to one session. session_seqis the index within the session, letting you pinpoint "which step of the session".referershows the source page;platform/channelshow the device environment.
5. Typical Scenarios
Complaint investigation
A user reports "my order failed" — query that user's BehaviorSequence to see whether they actually triggered submit_order, and whether pay_success/pay_fail followed, pinpointing where they got stuck.
High-value user behavior insights
Query the behavior sequences of top-paying users to understand their usage habits (e.g. whether they browse frequently, whether they make good use of search).
Risk tracing
Combined with risk events, examine the full behavior chain of a risk-flagged user before and after the risk occurred.
6. Notes and Caveats
- userId is required: the behavior sequence is keyed on
user_id; anonymous (non-identify-ed) users cannot be queried. limittruncation: heavy users may generate far more events thanlimit(default 100); increase it for long-period investigations, but mind performance.- Ascending order: events are returned from earliest to latest for easy reading along the timeline.
- Across sessions: one query includes events from multiple sessions within the time range; distinguish them by
session_id.