Skewed Data

Has anyone found a good way to represent two pieces of data that may be skewed? For instance, when building a bar chart to show sales of a production against the budget of a production. The issue lies in the budget and sale amounts of productions. One budget may be 6 million, and another 200,000. To put them on the same chart ion a realistic scale doesn't work out so well. 

  • In the Tessitura Community, we have a bunch of Skewed data scenarios.  Here are some general techniques for doing this kind of thing.

    • Switch a scale to Logarithmic.  This is good for showing very small values and very large values on the same scale so that you can see detail in both.  However, for most folks, the non-linear nature of the scale is very confusing.  I tend to do this when looking at the early dates in an on-sale period when I want to see the end of the on-sale period as well.
    • Eliminate or segregate outliers.  If what you really care about is the more run of the mill data then remove those from the data you are plotting.  And put a note in the chart that you have done so.  Or even put a small pivot table or second chart next to the chart showing the outlier's actual values.
    • Finally one of my favorite types of plots for showing the distribution of Skewed data is a Box and whisker charts or better yet is a good Violin Chart.  We can do box and wisher charts in Tessitura Analytics.  Violing Charts are not available.

    However, in your case, it sounds like you want to show the actual versus budget.  You could consider showing something like % to the budget.  and put that valus on a second right axis.  So one set of data (the actuals) is on the left axis. And on the right axis % of Budget completed.  I've done this kind of thing in the past.  However, I've not tried this exact scenario with Tessitura Analytics.

    One of the things I do when dealing with a problem like this I try to draw on just a piece of paper the chart I imagine I would like to achieve.  Then work out how to actually make it work.  If you want to share a drawing here or over in the Analytic Coffee Group there are others who might jump in with ideas.

    --Tom

  • Drawing on Tom's second to last paragraph, I would create a stacked bar using % based reporting = budget/revenue.  This would normalize the comparisons and provide an easy graph explanation and a simplified foot note citing the total budget for each production.

    Cheers, Mark