I know they are often frowned on, and I agree that they can be very misleading - but there are times when a stacked area chart can be a really effective representation, IMHO. Often when I want one of these I also want it to be "stepped" (i.e. representing measures that have constant values for a period and then a discrete delta corresponding to an event). A stepped line representation is also often useful.
It is just possible to cajole Tableau into generating stepped lines and stepped/stacked areas, but it's really quite difficult to do and the result isn't very satisfactory - especially for the area views (I do it using polygons, but highlighting, tool-tips, etc. don't really work very well). "First-class citizen" support would be nice. One day.
I have a classic example that I'm working on at the moment. I'll just briefly outline what I'm doing, both as an example of why I think there's a case for this chart type and also to see what other ways anyone can suggest of representing this data.
I'm analysing the performance of transactions in a system. Each transaction consists of a series of activities. I have the start and end time for each activity for each transaction.
Large numbers of transactions may be in flight concurrently. I'm trying to understand the queuing behaviour.
One of the many visualisations I'm using is a stacked area chart showing the number of instances of each activity (colour) in flight over time. One of the the main reasons for wanting the measures stacked is that the overall height of the stack represents the total number of transactions in flight at the time. I need area rather than just stacked bars because the spacing of the events (activity start and completion) is very non uniform, so I need to use a continuous time axis. Using bars, the gaps between bars give the impression of no activity, whereas really they represent no visible change of state, but the system may be very busy.
I also find a small-multiples view with "stepped" lines quite effective - but with quite a few values for the "multiples" dimension, it can be hard to see the time correlation clearly. Also, the overall number of concurrent transactions is very hard to see, whereas it's much clearer on the stacked view.
I've attached a couple of images showing the stepped/stacked area and the stepped line views I've managed to come up with. I've also got lots of other views (scatters and gantt charts are both very effective for this sort of analysis), but I'd be interested to hear anyone's thoughts on other ways of visualising the concurrency - especially if they don't require quite so much effort as these did. (The source data is in 6 tables with a few thousand rows but the stacked area view uses a 350 line custom SQL statement with 8 UNIONed clauses and a total of 72 table instances returning over half a million rows to get the steps and stacking to work - and I still can't quite get what I want due to the Jet issue I just posted about on the Datasources forum.)
Can anyone suggest any other effective (but easier) ways of conveying concurrent activities like this?