Journeys & model scores

Prev Next

Journey widgets reveal movement between steps. Outcome widgets show the decisions recorded on events. Model widgets help inspect score distributions. Use them together to distinguish a change in traffic from a change in its assessment.

Journey visualisations

Add a card from Journey Visualization, select a journey in its dropdown and set the dashboard dates. Expand the card for a larger view. The built-in All Cards → Journey board provides examples.

Circular Sankey

Circular Sankey displays step transitions as bands and can represent loops. Wider bands indicate more transitions. Use it to see the dominant paths and repeated movement between steps. The range control helps reduce a crowded diagram to a subset of links.

Journey transitions in a circular Sankey
The example includes login, checkout and gift-card steps. A “(none)” origin can represent events without a recorded previous step.

Chord

Chord arranges steps around a circle and connects related steps with ribbons. Use it to compare the relative weight of connections when a left-to-right diagram is crowded. Hover over a connection to inspect it; compare the same journey and dates when switching visual styles.

Journey transitions in a chord diagram
Ribbons emphasise which steps are connected most heavily.

Network

Network lays out journey steps as nodes with directed connections. Use it when the structure of the path matters more than the compactness of the diagram. Select the journey, expand the card, and inspect the connections between steps. Dense graphs may need a narrower period to remain readable.

These views describe recorded step relationships. They do not prove that every visitor completed a whole journey, and counts of transitions are not interchangeable with counts of distinct customers. Missing previous-step information can limit the visible path.

Journey Real Time

Add Journey Real Time and select an active journey. The card streams recent activity for that journey. Use it to observe an active test or check that new events are arriving.

This is a live monitoring view rather than a replay of the dashboard's historical period. Select a historical time-series widget when you need to compare earlier traffic.

Traffic Outcomes

Add Traffic Outcomes to compare event volume by journey/step with recorded decision outcomes over time. Hover over the chart to inspect a series and time bucket. Expand it when the legend has many steps.

Traffic outcomes over time
This view helps relate changes in step volume to changes in recorded outcomes.

An event without a decision may be shown as undetermined. This can be expected for collection-only steps. A pass, review or reject is the outcome recorded when the event was processed; changing a dashboard does not re-run historical decisions.

Events Summary

Events Summary provides a predefined overview including Total, Pass and Reject, plus journey/profile and outcome summaries. It is useful for a quick starting view when its fixed definitions fit your use case. For an explicitly defined KPI, use Aggregate Stats with a named attribute and condition.

When the comparison period has no observations, a percentage change can be undefined or shown as infinite. Read the absolute counts and compare the underlying periods before reporting a change. Check suspicious percentages against an explicit count rather than interpreting formatting alone.

Model Score Distribution Box Plot

  1. Add Model Score Distribution Box Plot, or open a Model Distribution card.
  2. Select its settings control.
  3. Expand the journey and step, then select the models to display.
  4. Close the selection popover to apply the selection.
  5. Set the date range and inspect each box plot.
Choose a journey, step and model
Model selection is organised by journey and step and is saved for the selected node.
A populated model score distribution
The box represents the middle half of the scores; the median lies within it. The outer extent shows the returned low and high values.

Read the score scale according to that model's definition. A negative or positive score is not, by itself, the final decision. Select comparable model/step combinations and equivalent periods when investigating a shift.

No models selected for this node means the card needs a selection. No data for this model yet means the selected model has no usable scores for the queried period. These are different situations.