All true. And speaking as a former plant engineer, the plant manager is not likely to provide for such a fancy package, and unfortunately in many cases, not likely to care. This data was generated and analyzed by a more forward-looking facility, and still Excel is the tool of choice.

Much of your dataset meets plant requirements (Response<0.5), with at least 20% of Responses wonderfully between 0 and 0.06, but you have several outliers over 2. This is a rough dataset to interpret by subgroup means (marginal or binary interaction), because any unfortunate combination with other factors is likely to perturb such a mean, and it may not be fair to hold the factor or factor pair of interest responsible.

This dataset would be a great opportunity to use your box plots and other methods from Tukey’s Exploratory Data Analysis (EDA) approach to visualize the scatter of your data. With a good statistics package, one could create a scatterplot matrix from your raw Excel data in minute or so, and we could see not just the averages but all 71 points scattered for every individual or pairwise combination of factors. If you can figure out a way to turn a raw dataset with arbitrary N factors and 1 response into such a matrix plot using just a general tool like native Excel, I would be impressed. (But I’d still recommend turning to the specialist — a real stat package — when the plant manager gets around to asking an engineer to use the dataset to quickly fix any problems.)

When I got to the section of the article titled “Panel Chart (almost)” then it struck me. Although the PivotChart was designed to chart PivotTable data, in fact it has no knowledge of the underlying PivotTable structure (unlike Tableau). Therefore, if the PivotTable has more than one dimension (which is of course very common), you end up with nonsensical charts like the line chart shown in your article. The PivotChart should be able to understand the multiple item levels in the PivotTable and create individual charts automatically. If it were capable of doing so, it would be one heck of a visual analytical tool.

Anyway, with two row dimensions, the technique you describe elsewhere (i.e. making a copy of the outer dimension(s) as another field in the source data and using that field(s) in the column area of the PivotTable), works well to create the separation shown in the first couple of charts in your post here. However, start adding additional dimensions (row or column area) and you end up with the kind of mess as shown in you crowded chart.

Having had to do something similar to what you’ve done here (but with sales data broken down by product, region & date), I ended up creating multiple copies of the PivotTable data so that I could filter each table (and thus chart) by the data I wanted to isolate on individual charts. The resulting charts were then stacked on each other. Although the result looked good, the process was ugly. At least I didn’t have to create formulas.

Sorry ’bout this long-winded response :)

Oh, by the way, I have to say that the using the duplicate field technique in a PivotTable allows one to create cycle plots quite easily.

I thought of the data arrangement, and was starting to rearrange it, when I decided it was late and I needed sleep. It would be a great follow up topic, wouldn’t it?