It's difficult to link social media accounts to purchases. But that's going to change in the future. There's been a lot of efforts from American Express on Twitter and Facebook that kind of pioneer this. We're a ways off from the whole purchase behavior stuff on a detailed level. Unless you have tracking and cooperation from vendors and such (like Anand is suggesting).
Simply put -- you don't go to Twitter or Facebook even to purchase things. You might go there to be influenced about a purchasing decision though. So yes, while people say one thing and do another -- what do the people listening do? How does what was said affect others?
"Intent" and "sentiment" is perhaps the best we have at the moment, but I wouldn't count on it for DIRECT purchase behavior so much as I would reputation management which could affect purchase behavior. Brands (small and large) can get a sense for how they are perceived across social media and that still really means a lot. I wouldn't be quick to dismiss that..Because it can tell you a lot about retention and satisfaction.
Ok, great someone purchased once... What happens after? Are they happy? Are they likely to purchase again? Are they encouraging their friends to purchase? Or discouraging others?
These are all things that happen outside of a direct sale and they are just as important. Probably more important for forecasting to be frank.
In simple terms, you'd want to basically look at your reputation (sentiment, volume of mentions, etc.) at a specific point in time (maybe a sales quarter) and compare that to actual sales data. Of course this isn't a quick process. You gotta keep doing it over and over. Over time, you might (and it's not a guarantee) be able to see a correlation between the social data and your sales.
You may wish to do more on social then and see how it affects your sales...And here starts the long experimentation process.
A little more quickly and granular - Google Analytics can help tell you your ROI for social when people come from social networks and make purchases. You can track that with Google Analytics. Though what Google won't tell you is sentiment. It does a fair job with gender and demographics, but even still you're going to want to compare all of your data. What happens is one thing kinda proves another. It's important to look at all the various signals and not focus strictly on one.