In this post, I'm defining success as "The number of people who find it useful is roughly equal or greater than the number of people who I thought would use it." I'm explicitly not defining it as financial success - most of the projects I've worked on have been hobby or volunteer efforts, and the commercial ones tend to flame out for implementation difficulties.
By this definition, I count 5 and a half successes (auto-uploader and new DB system for FictionAlley, Scrutiny and roomdraw ticker at Amherst, "Write yourself a Scheme in 48 hours", and a Netbeans plugin for my current employer is the half), 3 and a half failures (inAsphere, RejectedByYC/Bootstrapacitor and the current incarnation of Diffle), and 7 projects that failed because I or my employer failed to finish implementing them (a MUD, a Gnutella API, SchoolBrain, SecureTunnel, a Scheme-like programming language, collaborative editing, a Ruby plugin for Netbeans, and a Haskell-like programming language). The interesting part is the failures that were implemented fully, so I'm going to skip the successes and the bad implementations.
inAsphere.com was a teen-content dot-com run exclusively by teenagers, founded in 2000. Had the normal forums, link directories, articles, and polls. We launched it after a month and a half, folded in cross-x.com and its userbase (an earlier site founded by one of the founders), then found that rather than gaining new members, we were losing the existing cross-x.com userbase. The investors pulled out after 3 months and the project was canceled.
IMHO, it failed because: A.) we were pompous, elitist, and corporate - the market statement was "A teen content site for intelligent teens", thus implying that anyone who didn't like use was stupid. B.) we really weren't all that smart. C.) we had very little to differentiate us from every other dot-com in 2000 D.) we talked down to users - while they had felt like they were part of a community in cross-x, it just felt like they were part of a corporate machine with inAsphere.
RejectedByYC/Bootstrapacitor was a side-project my startup did after YC rejections came out. You can see the initial announcement threads at http://news.ycombinator.com/item?id=16234 and http://news.ycombinator.com/item?id=18233, though the site itself is down now. Aside from some of the obvious comments in the threads, I'd put its failure down to: A.) unclear positioning - because of the name and manner of announcement, many people thought it was intended to compete with YC, while it was really intended to serve as an additional resource for those of us who couldn't get into YC/TechCrunch/other programs. B.) Chicken-and-egg problem - the site gains much of its value by having a community of dedicated & driven entrepreneurs, but no dedicated & driven entrepreneur will sign up unless others already have. C.) Poor understanding of requirements - honestly, I don't think it actually "fit" its intended purpose all that well. As an entrepreneur now, I wouldn't use it myself, which is a pretty damning indictment. Ironically, my sister just told me today that she thought that idea has more potential than what I'm currently working on, so perhaps there really is something there, and we just weren't willing to spend enough time on it.
I'm reluctant to include Diffle ( http://www.diffle.com/ ) as one of the failures, because it was never meant to succeed! The site as it exists now is basically a bunch of functionality that we need for our real idea to avoid losing out for lack of silly stuff (much like how Google Video lost out to YouTube for a bunch of stupid and easily-correctable reasons). I could knock off that functionality fairly easily while still employed at my day job, so I figured I might as well put it up on the off chance that people like it. After all, having something up is better than having nothing up, and we have a completely different name and domain name ready in case we tarnish our brand. It also gives us a bit of a psychological boost, knowing we have a site that we can just plug our real engine into.
In retrospect, I don't think I would do it this way if I had to do it all over again. If your real idea is any good, you'll have plenty of users begging for you to add the features that you omitted. And you can find out much quicker if your real idea is any good, and switch to a different one if necessary. Plus, feedback on something that's not your real product is basically useless, because you draw a completely different audience. Friends and family that you show it to assume you're doing something different than you actually are, and they make all sorts of erroneous assumptions based on that.
Finally, I'm including the Netbeans plugin I wrote for work. This actually does get used - my boss loves it, and gave me a nice bonus for it. However, it's not being sold, which is what I thought would happen with it. The reason is kinda instructive:
When I started at my employer, I thought the company did something different than what it actually does. Like many bootstrapped startups, there's a split between the founder's vision of what the company should be, and the products/services that actually bring in the money to execute on the founder's vision. I had built a product based on the founder's vision, but I thought that the founder's vision was bringing the profits in, and so was a little disappointed when nobody except my boss uses my software.
Also, I think I should mention an observation about the 5 successes: in every single case, I was a member of the community that would eventually be using the software, and I basically just built something I would use myself. I was an author at FictionAlley before I was a programmer for FictionAlley. I was a student at Amherst before I wrote its course-evaluation system. I was interested in learning Haskell before writing a tutorial for it, and in fact basically wrote the tutorial as a way of teaching myself Haskell.