I'm just going to present the data without a lot of discussion for now. The lesson I'd like everyone to take away from this is: it is possible to improve your score with a lot of hard work (I put in 10-15 hours per week for about ten weeks). Don't get discouraged if you haven't broken 700, or 650, or 630, or whatever after four weeks of studying. It takes a while to really get it down.
My first GMAT score, with about three hours of studying, was 620. I should mention that I took the GRE a few years ago, so I didn't come in completely cold - I had some relevant experience (the SAT doesn't count in my case, that was way too long ago). The trend looks smooth and steady, but keep in mind there are only two data points between weeks 0 and 6, so there's no way for it to appear otherwise. Notice the dip in my score in practice exam 4. I was pretty bummed after that one; I thought my previous score had been a fluke. The rebound in practice exam 5 was carried by my Verbal score; my Quant score didn't budge between practice exams 4-5 (see below).
All data come from Computer Adaptive Tests, not paper-and-pencil tests.
The percentile graph is telling - it shows similar improvement in both my Quant and Verbal scores. I was surprised at the improvement in my Verbal score since I had a good handle on the English language going into it. But the GMAT is a fickle beast, and there are a lot of test-specific quirks to master (for example, "correct" American English idioms). A big part of the challenge is learning how to think like the test makers. Another key challenge is getting the timing of the exam down.
I'll follow this post up with some more analysis, but I wanted to get these graphs posted since I think they're pretty interesting in their own right.