If you're only going to include two time points (i.e., two of your pies),
I'd make a clustered column chart. Two clusters, one for each date; five
series, one for each division. You can readily compare heights of all ten
bars. Looking at the pies I made with some dummy data, I formed a mental
assessment of the percentages of each wedge. When I made the column chart, I
was surprised by how small the small divisions were. Apparently smaller
values are overestimated when reading a pie chart, but not when comparing
heights of bars in a column chart.
But I'm not sure from your last post whether you need two charts, or 24 (12
months x 2 years). While one or two pie charts are bad, an array is awful.
There would be no quantitative information transferred at all.
This requirement to explicitly show three categories that comprise a total
of 5% of sales takes away attention from the two other categories which
account for the vast majority (95%) of sales. I would make a panel chart
like this one:
http://peltiertech.com/Excel/ChartsHowTo/PanelUnevenScales.html
with different height panels. I'd put the panels in order so the largest
division is at the top and the smallest at the bottom. I'd make the scales
of the bottom three go from 0 to 10% (any smaller range makes the bottom of
the chart look too cluttered), and the larger two from zero to, whatever,
60% or so, to account for the greatest value in either division's sales, The
total height of the chart would then be 60+60+10+10+10, or 150%, and 10% in
one would be the same height as 10% in another.
When I get a chance I'll write this into a blog post. For now I think this
may give some ideas how to proceed.
- Jon