This article on InfoWorld raises the idea that SaaS may be "throw away" software and as such, its long term viability is questionable.  Simon Jacobson, an AMR analyst, is paraphrased in the article as saying..

According to Jacobson, some companies are using SaaS as an interim solution. Try it for a year, two, maybe three, and when the time is right, dump it for something better.  That something better could be another SaaS application, yes, which in turn could be thrown away when it gets replaced by the new, more innovative solution that comes along.

Ephraim Schwartz, the article's author, continues on with that notion and concludes there is a high likelihood that SaaS could very well end up being another flash in the pan.  His conclusion stems mainly from the idea that a company can bypass initial up front investments in traditional software, ERP is the example used in the article, for two or three years of SaaS "goodness" and then dump their SaaS commitment and move on to something bigger and better.  To be correct, this idea relies heavily on the notion that a SaaS "commitment" easy to break off, with time and dollars being the two primary dimensions used to illustrate the lower commitment threshold.

But that really is not accurate.  From a pure IT perspective, yes, SaaS typically has lower startup costs and a much faster time to deployment model than traditional solutions.  But Schwartz misses out on two key things when he looks at short SaaS "flings".  First, after two years, there is going to be a significant amount of time invested in training employees and shaping business process around that short term SaaS solution.  When I look at the massive amount of time and energy that was spent on getting Microsoft employees and processes up to speed with our Siebel deployment (shudder, groan... run to bathroom) I laugh at the idea that a company would want to go through that type of pain every two years.  Yes, smaller companies would experience less angst, but the long term ramifications of switching major business applications would have a big impact on the bottom line regardless of the size of the company.

Secondly, and I would argue more importantly, think of the important data that would have to be moved from the choice du yesterday to the choice du jour.  After two years, a company is going to have a lot of customer data, business data, you name the data, invested in a given platform.  There is absolutely no incentive for a SaaS provider to make it easy to extract or migrate that data in any meaningful way.  It enables the very scenario that Schwartz describes so any SaaS vendor that has the mantra of "we make it easy for you to leave" is just asking to go out of business.  No one is going to do that.  The whole idea with the "try before you buy" model that many SaaS vendors have is based on the notion that a) you will find there offering acceptable and b) once you actually invest time/resources (not necessarily dollars) it won't be easy for you to leave.  Even if you are able to pull the data out of a particular SaaS vendor, you still have to get it into some other solution.  That is not going to be an easy proposition either.  Vendors may start to make such tools, but rest assured their competition will continue to try and make it hard for you.  The SaaS model actually makes it very easy to do just that.  If Vendor A comes up with a solution to migrate Vendor Bs data, all Vendor B has to do is change their data formats ever so slightly.  The end user will never see the changes (the beauty of an application in the cloud), but Vendor A's migration tool will suddenly be SOL.

SaaS and it's cousin S+S are long term directions IMHO. There are still issues to be resolved on a variety of fronts before they are truly "universal" and even then, there will remain many valid reasons to run applications on premise.  Maybe some of those reasons will limit SaaS success, but to say SaaS is "throw away" and doomed to failure ignores a lot of the "real world" realities that business face today.