Business Forecasting Using Deep Learning
The relative change as a percentage of prior quarter and the absolute change from the prior quarter have been plotted for all dealers. The horizontal axis represents the number of dealers and the vertical axis represents either the relative or absolute change in performance.
The plot has been set up so that individual classes can be observed. Select the class from the dropdown menu.This can just as easy be any categorical variable, i.e., region, salesperson, credit rating, etc. Hovering the cross hairs over a point reveals the dealer identity. The range can be zoomed using the box tool and reset with the reset tool.
The visualizations purpose is to demonstrate that the hightest growth or decline comes from a small subset of the dealer population. The majority of the dealers continue with a baseline growth path. The visualization displays those dealers expected to have high performance contributing to client growth and those dealers who will decline and drain client performance.
The unfortunate fact is that the high and low performers are rarely the same quarter to quarter. Timely identification of these changelings can result in the client refocusing the resources now applied evenly to all dealers regardless of performance into a more effective concentration.
Comparing the Registered Class with the Platinum Class reveals that a large portion of the expected growth and decline will occur in the Platinum Class. The Platinum Class demonstrates extreme ranges in low and high performance with few occupying the midrange of the order input curve. Applying resources to correct under performance and instead supporting high performance is a good management choice for the next quarter. It would also be a good management choice to further examine the causes of low performance in each class. It is also likely that pruning the Registered class could result in cost savings and better resource allocation for the client.
The class plots demonstrate that class identities make a big difference in performance forecasts. Additionally, it is apparent that a small portion of the dealers in each class contributes the vast majority of the change - both upward and downward. The client will do well to specifically focus its attention to the unexpected performance by dealers identified within each class. Note: the selections available could include any categorical variable such as region, territory, etc. and further dealer focus is possible.