If tables of complex data lack sufficient visual analytics illustrations on which recommender system to use, users will go to another website for better illustrations. The more complex the analysis of similar record attributes become, the more difficult it is for users to make informed decisions. Du et al. include seven illustrations of visual analytics that can help readers to better understand the paper.

The first illustration is a graphical display of the EventAction user interface that is split into two parts: review recommendation (of action plans) and review similar records. The more the user switches between the tabs, the more the display expands. The expansion stops when the user achieves the desired outcome.

In the next illustration, the user sets up eight similarity criteria and includes “paper” and “advisors” as temporal criterion. The user specifies that the first two years are not selected and the advisors are not late in submitting the papers. When choosing the latest similar record distribution, the user is shown the related temporal data.

To take a deeper look into the outcome, a hierarchical tree of similar records is shown as a donut, followed by a compact overview of similar records and a heatmap of temporal events ranked by frequency of similar records obtained. They provide more flexibility in getting more data details than the standard graphs.

To help in fine-tuning the recommendations, an activity history, outcome estimations, and action plans are visually presented. The next illustration is a simple workflow of tasks to be taken to prepare for an action plan. It is followed by a chart of average time for locating the records.

The last illustration is the “channel attribution analysis” case study that finds records of marketing value, such as customer attributes like age and previous product purchases. The study shows the desired action plan of getting the customers to purchase more products.

The authors’ paper is detailed--the approach to the problem, the descriptions of user interface and backend systems, case studies, guidelines, and limitations. They provide illustrations to help readers better understand their visual analytics approach. Those interested in the challenges of setting up an adequate approach for recommendations that can be visually explained should read this paper.