The "oh, shit, I am pregnant" use case is laughable. Again, the pregnancy testing kit is far more accurate than the statistical analysis and it can be used few days before the first day of the period. No service can do that.
The only case where stat analysis of the cycle is applicable is for people who do not or cannot use contraceptives and do not want to end up being pregnant. This is a negligibly small group as is, and I suspect that its subset of Web 2.0 aficionados is going to be close to zero.
Incidentally, while a pregnancy test will tell a woman if she is pregnant before the period, she has no reason to worry until a day after it should have arrived (ignoring condom failure).
Why would any woman use a simple and easy cycle tracking service when better methods are available at the drug store? For the same reason you might google someone rather than paying $50 for a background check.
Consider a woman with a noisy cycle (28 days +/- 5) or with a mean different from 28 days (29 days +/- 2). Such a woman may become nervous about pregnancy on a regular basis: "I'm 2 days late, and my flawed human memory doesn't tell me that this occurs 30% of the time for me. Oh shit, am I pregnant?!?" The service can remind her that this is normal.
It's also useful for women with regularity problems (a detailed timeseries would be useful to show to the doctor), or women trying to get pregnant (not always as easy as condom ads suggest).
Also, consider that you may know women who need this service, but they may not discuss the need with you.