Exploring, predicting and reporting with SAS Visual Analytics and SAS Visual Statistics

sankey diagram - Late Entry

Reply
New Contributor
Posts: 4

sankey diagram - Late Entry

Sankey diagram is very useful for tracking the trajectory of the event, even a cohort of customers .... however, currently I am facing an issue, sankey disgram can deal with drop off earlier case, but how to deal with the late entry cases?


 For example, there are 5 steps for my event ( ie. step 1 to 5, entry at step 1 and ended at step 5), a part of the cohort can finish at step 4, and a part of the cohort can enter from step 2 or step 3, how should my dataset be structured for the late entry situations?

 

 

SAS Super FREQ
Posts: 110

Re: sankey diagram - Late Entry

Hi, The path analysis algorithm does take care of this for you. Given your transaction data set - late entries are determined by their sequence appearance. Have a look at the blog at https://blogs.sas.com/content/sascom/2014/08/19/path-analysis-with-sas-visual-analytics (a little dated but it shows the structure of the expected data set). In this example the customer 'Jane' views web-page B as the first entry whereas everyone else starts with page A. So Jane is coming in one step later as the rest. Obviously this is very basic path - but the same applies to larger paths and scenarios.

Hope this helps. Falko

New Contributor
Posts: 4

Re: sankey diagram - Late Entry

Posted in reply to FalkoSchulz

Thanks Falko,

 

but I'd like the B (Jane's start point) at the same column as John's B, for example A, B, C, D are time related event, we'd like they are line up as a time series.

 

Please see the attached mock-up graph, the red circle part

 

sankey.png

SAS Super FREQ
Posts: 110

Re: sankey diagram - Late Entry

Ah I see. Sorry I misunderstood your question. No, unfortunately in path analysis related events need to have the same start sequence in order to join. If you have a path with only a portion of the events - it's considered a separate path. What you may be able to do - is to use dummy values for the events the customer is not yet part of the main path (e.g. missing value). So a table structure like this with Jane joining in at C:

 

va_82_sankey_simple_path_data.png

 

..would produce a diagram like this (screen shot is using the newer version of VA 8.x so slightly different):

va_82_sankey_simple_path.png

Not exactly what you were after but hope this helps.

 

Cheers, Falko

New Contributor
Posts: 4

Re: sankey diagram - Late Entry

Posted in reply to FalkoSchulz

Thanks !

 

that's the method I am using. I use "." instead of "Missing", it works perfectly in the Event view (only showing a dot , no flow from . to real event), but not looks nicely for the pathway views , ie including the . event (blank event) in the pathways.

 

Anyway , thanks a lot ! hopefully SAS VA new version will consider this scenario .

 

Cheers,

xiaoping

Ask a Question
Discussion stats
  • 4 replies
  • 229 views
  • 2 likes
  • 2 in conversation