Flow visuals enable us to explore user behavior across a sequence of discrete time-stamped events, towards a defined goal. They only plot data that achieves this goal.
Flows are an implementation of a Sankey plot: blocks represent individual events, and the width of the connection is proportional to the size of the measurement connecting any two events. The plot is defined by a path, or sequence, through time-stamped events. To learn how to identify and define events of a dataset, see Events.
For contrast, see Funnel Chart.
For convenience, we divided this discussion into the following sections:
Keep in mind the following best-practices when defining the path of a flow visual:
The following steps demonstrate how to create a new flow visual on a dataset Campaign. This dataset is based on data previously imported into Arcadia from the datafile campaign.csv. For an overview of shelves that specify this visual, see Shelves for Flow Visuals.
In the visuals menu, find and click Flow.
All shelves, except Filters and Top K, are mandatory.
Populate the shelves from the available Fields:
timestamp
and place
it on the Time shelf.user_id
and place it
on the Session shelf.Start
and place it on the
Path shelf. Similarly, add All non-purchase application
events
and Purchase
, in that order.Record Count
and place
it on the Measure shelf.Click Refresh Visual.
The flow visual appears.
Click (pencil icon) next to the title of the visualization to edit it, and enter the new name.
Campaign - Flow
.At the top left corner of the Visual Designer, click Save.
Note that on the visual we just constructed, each connection has only two segments. This
does not model the real business case, so let us refine the visual. To model more of the
path that satisfies the event conditions, we need to modify the intermediate event,
All non-purchase application events
, on the Path
shelf. Note that setting a higher maximum models more of the data.
On the Path shelf, click the All non-purchase
application events
field.
In the Event Properties menu, set Min Rage parameter to 0.
Similarly, set the Max Range parameter to 5.
Click Refresh Visual.
The flow visual appears, showing multiple paths towards a Purchase
outcome. Note that in this form, the numbered steps include the
Purchase
event along the way, and each Purchase
block
terminates the flow of connections that flow to it, in a step-down manner.
Notice also that we can specify the Goal of the flow visual, to generate a plot with a single terminating goal block. To do that, perform the following steps:
Purchase
field.Click Refresh Visual.
The flow visual appears, showing multiple paths towards the Purchase
event.
Note that in this form, the numbered steps exclude the
Purchase
event, which collects all flows on the extreme right, and
then terminates them.
To remove the Goal option from an event, click the Purchase
field on the Path shelf, and de-select the Set as
Goal option.
You may wish to examine a particular business case scenario, such as how card impressions influence the final decision to purchase. To see this path, simply click on a flow leading to the first card impression, and the flow highlights the entire path.
If the path is complicated, you can change the order of the nodes in any step of the flow visual, simply by clicking on that node and dragging it up or down.
In Figure Changing the Node Order, with Goal, we demonstrate how you could re-arrange the flow visual, step-by-step, to highlight a specific path you wish to examine.
In Figure Changing the Node Order, Without Goal, we removed the goal, and arranged the nodes to place all card impressions at the top, and all purchases at the bottom of each step.
The flow visual is very good at demonstrating a sequence of events, and the proportion of events that pass through a particular node. However, because it does not use an axis, there is no relative scale that conveys the actual size of measurements on the chart. Thus, tooltips become a very important component of the flow visual.
Consider the Tooltips in the following chart.
Lastly, remember that flow visuals only show data that achieves the desired goal, in the specified number of steps. The following image shows only 5 intermediate steps (specified as maximum), and contain 48 purchases altogether.
Contrast this with a visual that covers 10 intermediate steps (specified as maximum); it contains 51 purchases.