Paths
The Paths insight visualises which events flow into and out of a chosen event — rendered as a Sankey diagram. It answers “what do users do right before and right after they signup_started?”.
Open Paths in the sidebar (/paths).
Filters
| Field | Notes |
|---|---|
| Start event | Required. The event the Sankey is anchored on. |
| End event | Optional. If set, only paths that reach this event are kept. |
| Max steps | Slider, 3–10. How many transitions to render either side of the start event. |
| Min count | Number, >= 1. Edges with fewer transitions are pruned to keep the diagram readable. |
| Date range | Filters the events considered. |
Reading the Sankey
Each node is an event. Each edge between two nodes is weighted by the number of transitions — wider edges = more users took that step. Hover an edge to see the exact count.
Sankey edges show pairwise transitions (event A → event B), not full N-step paths. If you need to test a specific ordered sequence with a conversion window, use a Funnel instead.
Common use cases
- Drop-off analysis — set start event to your “abandonment” event (
payment_cancelled,signup_abandoned) and look at the events users fired right before it. - Recovery routes — set start event to a friction event (
error_shown) and look at outbound edges to see what users do next. - Discovery patterns — set start event to a hero feature (
ai_chart_created) and look at the inbound side to see how users reach it.
Limits & gotchas
- Max steps caps total nodes (start + N outbound + N inbound). Wider diagrams become unreadable past ~10.
- Min count is a hard prune — edges below it are dropped before rendering, not greyed out.
- No property filters yet — the only scoping is via date range. Pre-filter by emitting more specific event names if needed.