A typical mistake that I have seen a lot of startups make is, “start development” first without a cohesive “Blueprint”.
A “Blueprint” for a web bases service should comprise of detailed wireframes (a term used for grayscale diagrams) of each and every page that a user will go through in order to complete the task that you plan to launch. But is this enough? No, not at all. Unless you test these wireframes, you will not know that the workflow that you have designed for the user will work exactly the way you want it to. Is the user following the prescribed path or are they getting distracted with some other element on the page? is the workflow too long? By compromising on the number of steps would you be compromising on the usability? Etc. etc. are the questions that one needs to address with usability testing.

So, why don’t we launch the service and then test it with live audience?

How many times have you been to a portal or subscribed to a service which had a tedious workflow? even though you tried it once, did you go back?
Testing on live audience might make sense to some but think about the cost of acquisition of a user…now that’s something that will help you take a call.

Hence we believe that its important to know what you want to deliver to your end user and start development based on that.