Item explorer is an approach to provide insights into a ubiquitous class of business questions like:
For this class of business questions, the exponential number of combinations poses a severe practical challenge. Due to the explorative nature, visualization is well-suited for such business questions. More specifically, a visualization can provide a unique representation for both revealing insights and for intuitive user interaction based on business knowledge or own hypotheses.
Let's consider a grocery store which tries to understand the shopping patterns of customers. In theory, a store selling n different products can observe 2n-1 different shopping patterns! To make it more manageable, we consider a higher level of abstraction, namely 20 different product categories, such as eggs, cereal, candy, etc. Still, there are 220-1 ≅ 1 million possible shopping patterns:
For this setting there are three obvious approaches to start with, however, each comes with its limitations:
The basic idea of item explorer is to visualize the product categories in an intuitive way and to allow the user to interactively explore the combinatorial space. The choice of the underlying visualization is the good old bar chart! Huh, why that? The decision in favour of the most widely applied, old-fashioned visualization comes from the fact, that the cognitive ease and intuition to read and quantitatively interpret bar charts is still unmatched. Yes, let's go back to the very basics ... but at the same time let's push the envelope to accompany the visualization with rich and meaningful interactions without sacrificing the ease of use and interpretation.
Item explorer represents each item/ product category by a bar in the bar chart. Following the example above, each bar denotes the number of customers buying that specific product - independent of all other products. E.g. a bar referring to candy denotes the number of customers who bought candy no matter what they bought in addition. This means each bar has a clear meaning but customers who bought more than one product are contributing to the frequency of all bars of the corresponding products.
The interactive exploration is supported in two ways:
Feel free to jump right to the links to a demo, a video, and the source code on github.