Until the model has matured, and standards of expectation and performance are settled, SaaS can be a slippery ride. Just because the press is shouting "Surf's up!" doesn't mean that you can just go do it, dude.
Since the balance of negotiating power changes on contract signing, it's critical to cover the details up front. For example, it's not uncommon for an agreement to contain early-out options for both parties if the relationship can't be made to work (e.g., performance for buyer, profitability for service provider). What are the guardrails, including equitable financial resolution, for these decisions?
An agreement might initiate payment for services upon "acceptance" (e.g., production over time at committed service levels without any severity level 1 incidents). What are the service levels, and what incidents qualify for severity level 1 treatment? How does the agreement resolve the challenge of enabling ongoing modification?
This discussion has an added complexity when the service provider uses a common instance of the software, an underpinning of most SaaS models. How does the purchaser keep his business processes  unaffected if the provider goes out of business? At the end of the agreement, what transition help and data is the service provider required to give, for what compensation?
All these questions (and many more) are easily resolvable, but they require a detailed understanding by both parties of the services to be provided and the conditions of, and compensation for, provisioning those services. It's not uncommon for the pre-contract effort to take as long as the implementation effort.
Source: Intelligent Enterprise, http://intelligententerprise.com