The single biggest problem (IMHO) for a startup is traction. Getting people to stop whatever it is they are doing and use your product is impossible at least 90% of the time. I know this because it has happened to me several times.
So here is one advise I would like to share with startups: build a product that does not require users to come to your website. In other words, do not change your end users' current behaviors. The best way to do this is to integrate with the applications they are already using. One application that is common to all internet users is Email. We all go it. If your audience can make use of your product by simply sending an email then you are ahead of the curve. Doubt it? Ask Posterous, CCBetty, or Snoozemail. So if you are readying to build a CRM app, or Search engine etc.. figure how to make it work from Email or IM or similar products others have already accepted.
Edit: And in regards to CRM, what small businesses really need is not a light version of SalesForce, its a product that can make sense of forwarded emails.
My startup is in the b2b sphere (specifically, industrial research, design, and manufacturing). Our product is slowly growing and getting traction, but that's only because we don't expect to have tens of thousands of users. Each of our users is worth more to us than the average twitter user, so we put more time into recruiting them to the site (generally it's a personal contact 50%+ of the time that gets them there).
So in our case, we can afford to put the time in with individual users. We thus don't really need to integrate with existing products, because our market doesn't demand it. In fact, in the IRDM area, I've found that almost everything is proprietary and very secretive; every company protects its trade secrets tightly (not because they are advanced or cutting edge... just because they hold on to every minor edge they can, regardless of the wisdom of doing so). In short, it's very old school business-types who think their secret sauce is in their product rather than in their relationships and experience.
But I would definitely like to second the "figure how to make it work from Email or IM or similar products others have already accepted." Our app has put a lot of focus on email interaction, which I think is crucial for us to succeed.
I guess all I'm saying is that you really should make sure you know what your ideal user is like. If they're like mine, then you have a little more leeway because they're a little more used to the non-integration of different components (i.e. they are willing to try anything).