Print on demand books sell as well as their marketing makes them sell. The reason the majority of books published with print on demand don't sell well is because they are either low demand academic frontlist or trade backlist books, or, they are subsidy published and self published books from authors who don't do any effective marketing and give up after a short time. Publishers who use print on demand for their frontlist books and invest serious time and effort into marketing and producing books for which there is a market do quite well, and have very low costs. For new publishers, I believe it’s far easier to learn how to promote books published with print on demand and only move to offset later if the economics make sense.
One of the problems with traditional pre-market publication when using LSI, Replica or Booksurge for print on demand is they just tend to blow up once a year as their systems are changed. Publishers who get too clever with their timing (and I've been guilty of this myself) can end up with a bunch of orders cancelled by Amazon and no Ingram availability right when the publicity hits. I know some publishers would like to believe that potential customers will keep a good ad stuck to the refrigerator door for a couple months if the book isn't available when they go first go looking, but I have my doubts.
Learning how to promote books online requires just as much study and work as any other type of marketing. I've seen the sales increase for each one of my titles on a year-over-year basis, doubling or more in the first two years and than smoothing out some. That's the dead opposite of the traditional model, where most new books, if they get that far, don't survive the trial period on the shelves, and see the majority of their lifetime sales in the first six months. One of the reasons my sales tend to go up over time is that it takes a while for the increased visibility at places like Amazon to kick in, and word-of-mouth requires mouths. You can jumpstart your Amazon sales with paid promotions, I've never used one myself, and it also requires careful study to be cost effective. I’ve found the key to promoting books with a publisher website is very simple. You have to give away a lot, by posting major excerpts from your books and associated material. This is the only way that people who find your site will link to you as a resource. Most publishers remained terrified of doing this and just never experiment with it in depth. Websites that are essentially online advertisements for books don't attract anybody. You don't need super high numbers of visitors if excerpts are compelling enough to encourage buyers. It does take a lot of effort, but the cost is minimal if you do the work yourself, unlike say, paid advertising, where you can't.
When I started publishing and selling books on the web ten years ago, I thought there would be a limited window of opportunity before the web would be simply flooded with quality sites from kids who grew up online. Nothing could be farther from the truth, the quality seems to be slipping and most kids who look like they know what they are doing online are just in it for the movie and music downloads. I've helped a number of people launch sites recently, and now as ever, content is king. If you put out the best content and work at keeping it current, you'll get more readers and sell more books. Blogs offer a pretty nice content management system for those lacking design savvy, and they also help undisciplined writers add content on a regular basis.
Some people avoid trying to promote books online because a failure to attract visitors stares you in the face every day you check your metrics. Traditional advertising is far gentler on the ego. If you run an ad in a magazine or send out a big mailing and you don’t sell books, you can always find something else to blame it on. I don’t know what kind of marketing isn't hard work and doesn't take many months to come to fruition. With online marketing, the cost is measured strictly in time.