Warning: This is a bit off-topic.
Using R, I'm trying to develop a way of visualizing a marketing campaign
that is divided into lists, each with a mailing and control group. Within a
mailing group, we have respondents and sales conversions (think
"successes")
and the overlap between the two. We also have conversions among those
subjects in the list's control group, who were withheld the mailing. There
are several lists to a campaign, with varied numbers of subjects in the
mailing and control groups.
I want some way of illustrating this information visually, a display that
communicates:
1) The relative sizes of each list's mailing and control groups
2) The lists' relative importance to the overall campaign
3) The effect each list's mailing has on proportion of conversions to sales
4) The volume of the overlap relative to the number of respondents and
number of conversions
5) Confidence intervals about all of these figures
I've taken a stab at developing my own type of display using R and grid,
which you can see, along with comments and the raw data set, at
http://r.loyaltymatrix.com. Unfortunately, my display feels clunky and
nonintuitive. Any suggestions? Please feel free to comment or to share any
ideas, either through e-mail or the commenting feature in the blog or on
this list.
As a subtext, I've looked into rmeta, which has a funnelplot function and a
plot for the Mantel-Haenszel test, meta.MH. Neither is quite what I want.
The funnel plot tells me nothing about the relative sizes of the mailing and
control groups; I would like to be able to see it when a control group is
tiny. By the same token, it shows no confidence intervals, which are
important if the control groups are too small to make any real statement.
plot.meta.MH shows the intervals, but doesn't offer any information about
the size of the lists and their relative importance in the campaign.
Additionally, neither utilizes the information I have available about
respondents (as opposed to conversions) and their overlap with converted
customers.
Sorry to bother everyone with a somewhat off-topic note, but it is my hope
that someone may have dealt with this question previously and may have some
wisdom to share. Thanks for any help you can provide,
Kevin
