Sankey diagrams for Multiple Dimensional relationships

nasa.gov

Majority of visualization objects can accommodate either one or two dimensions at a time. This would suffice for a good majority of requirements related to data visualization.

Sankey diagrams or flow diagrams can be customized to show different perspective among multiple dimensions across a single measure. In the example below, one can see the relationship between different dimensions & attributes related to Customer and Product connected via sales volume.

Although a sankey based representation is good to show relationships, it easy to fall into traps while building or interpreting data.

  • The representation is effective only when the measure is pro-rated to 100% (e.g. Volume of Sales based on Gender and Channel should be the same denominator)
    • Data gaps should be avoided between stages (E.g. Gender = Female & Channel)
  • Flow or relationship should be viewed one stage at a time instead of cumulative approach
    • Inference such as, more than 50% of Female customers shop online also own their house is wrong.
  • Order of dimensional placement is important for interpretation and can be swapped or removed

Due to slightly confusing nature of the visualization, it takes time for users to acclimatize with interpreting the results. Sankey diagrams are one of the few options available to show multiple dimension relationships. Due to this confounding nature, it is still in early stages of mainstream BI reporting adoption.

Resources

  1. Create your own Sankey diagram using Google Charts API