Graphing Uncertainty in Marketing ROI Statistical Models

Presented multiple novel concepts for imbuing levels of statistical uncertainty into data visualizations. The visualizations graphed the marketing channel efficacy through relative ROI. Relayed unique data dimensions visually, leveraged strong conceptual understanding of perception and interpretation of data visualization language. Refined mock-ups were created in Adobe Illustrator.

A bar chart using the Y-axis to depict Total Spend and the X-axis serving as a categorical delineation. The width of each bar is generated by the impact mean value. The softness of the gradient is a depiction of the confidence or uncertainty of the model.

A scatter plot in which the Y-axis is again the Total Spend. In this chart the X-Axis is now a linear interval scale of the impact Mean. The channels are color coded and plotted along both Axis. The size of the points is not linked to a value, however the width of the whiskers extending out from the points are linked to the confidence or uncertainty of the model.

This example uses a hybrid lollipop-bubble chart where in the Y-axis is the Total Spend, the X-axis is nominal categorizing the different channels, the size of the points is linked to the Mean value and the width of the point’s gradient is representative of the model’s confidence. This direction also lends itself well to a time series. The third example of this direction shows that by replacing the X-Axis with time and switching to color coded channels the viewer can see changes over time.