In fact, the worst manager I ever had (who repeatedly asked me to define standard deviation, and he had a PhD) appreciated my displays the most.

I thought that might be the case, something a manager asked for. I always find it difficult to work on these types of projects.

My favorite example request was trending data over time using pie charts. There was no business value but a boss wanted exactly what he requested and his peers all thought it was good. I do not do that anymore.

Anyway, I was curious about how all this came about. Thanks for the detailed answer.

- Dave

I think the purpose that was originally put forth was to do something a manager asked for; making a manager happy has some value, not necessarily business value. I did come up with a methodology, but then dismissed it as confusing and unenlightening.

I proposed another method to track sales of all products, so at least one could find trends across the time periods. Sometimes giving your manager something different but better, especially if the manager imagines they helped guide its development, also has value.

I have to admit, since someone emailed me out of left field asking for free advice on how to make the chart their manager wanted, I was more interested in steering them toward something more useful. I was not expecting to turn it into a larger project with more flexible exploration capabilities. This was simply a quick and dirty example of an improved approach.

If there had been more information (and greater interest), including a breakdown of products by type or other factor or details on the sales, etc., I may have been inclined to set up a worksheet to feed a pivot table, to allow grouping, sorting, and other analyses.

call me dense but….

- What is the purpose of this ranking/charting exercise?
- What insite will one get from charts of the top ten sales of 14 products by week?
- Are any products related?
– For example, Does a sale of Product A lead to sales of Product B?

I know you were asked to do this but I can not see the value in it.

If I had all this sales data, I know I would like to know what is the ratio of sales to new customers vs existing customers and a whole bunch of other metrics but without knowing the business need of this data I find it difficult to think of visual means to display the data.

@dmgerbino

Sure, you would make a stacked area chart (or stacked 100% area chart). You can’t use real times since the X axis of an area chart only recognizes integer numbers. Instead you would use convenient X axis units (say 1 = 1 minute) and apply a dummy axis, or just use 1440 rows of data (1440 minutes/day) and in the first column leave blanks everywhere except where you want labels to appear.

You could even handle some sort of interactive features usig chart events as I’ve described in a number of posts.

Is it possible to do a chart like this one here from NYTimes, using any type of data metrics?

New York Times: How Different Groups Spend Their Day

http://www.nytimes.com/interactive/2009/07/31/business/20080801-metrics-graphic.html

OJ

“There are many ways to make the output more flexible than I’ve shown in this simple demo:”

one more thing that comes to mind: using sparklines to show weekly sales. This takes less space and provides quick glance. That leaves the space for important analytical questions like: “which products are top 10 on week n?”, “what are the top n products?”; “which products performed badly?”

I was provided with six weeks of dummy data, and it is intended to be extended to 52 weeks. I think top ten was something they thought of because whatever amount of products per week times 52 weeks is just too much for a column chart. The fact that Item L dropped to zero the week after it placed first is more important (to me) than seeing that it fell out of some arbitrary Top N list. As if I’d even notice this in the column chart.

There are many ways to make the output more flexible than I’ve shown in this simple demo:

- Chandoo’s suggestion to sort by a given week
- The ability to highlight (“brush”) a group of lines
- The ability to adjust the time scale (start, stop, duration)
- Plotting of cumulative totals, not just weekly values
- Aggregating items by category (e.g., A and B are peanut butter, C through F are jams, G to I are bread, etc.)

As an onging concern (up to 52 weeks in a year), how should the problem be tackled? One possibility is to break down the analysis into four-week intervals, and display the top ten items per week in four bar graphs. A user control would allow selection of the week range, updating the charts accordingly. No doubt, there are other ways the tackle the problem.

Thanks again for another thought-provoking article.