One of traditional publishing's many, many failures is their basic lack of data capture - the fact that they have no way to measure how an individual book is a success or failure compared to others, and thus no way to derive why a book succeeds or fails. The rest of us don't get to lean on multi-billion dollar media conglomerates and whine about Amazon; we have to figure it out as we're already in motion, in order to keep improving our successes and minimizing our losses.
A quick note on success: if you truly believe that publishing your book was only done for the love of it, then you're likely to be offended when success is measured in monetary terms and volume of sales. That's fine for you, but not for anyone who wants to increase sales, or cover the mortgage. Math is a language used to describe the universe, and in order to describe it, we need to use terms in common. Unit volume and dollar volume are the terms I'm choosing. Fuzzy feel-good and raising awareness are not quantifiable terms, and therefore are not useful terms.
So, how do you track the time you spend building an audience, the time you spend writing, the time you spend editing, the time you spend looking for cover art / working on covers? How do you track the size of your fanbase, and whether or not they're actually interested in buying the things you write? How about the reach and effect of promotions? Welcome to metrics.
While I should hope that facebook and twitter, google+ and myspace and every other format I'm rolling into MyTwitFace have the ability to keep metrics, I'm going to tackle blogs because I know blogs. Wash, rinse, repeat, apply as you can to other social media.
First, how much traffic do you get on your blog? Sitemeter.com will let you track the number of visitors and pageviews, the sites referring people to you, and your most visited posts. It doesn't, however, capture the audience that uses an RSS feed to view your posts - for that, see feedburner. Neither do a great job at capturing outclicks - people going from your post to an amazon book page, for example.
If you are in a state or country that has an agreement with amazon.com (and if I understand correctly, .co.uk also has this option), get thee to the Amazon Associates program and sign your happy site up right now! This program lets you give site-tagged links for products that people can click on, and you get a kickback for driving traffic to them. However nice the money is, though, it's almost irrelevant compared to the data. With this linking account, you can actually see how many people who clicked out from your site went and bought your story!
This is how I can tell you exactly how many people clicked through a book announcement each day and bought the newly-released book. This is how I can tell you the rough effectiveness of a newsletter mailing. Mailchimp can tell me how many people clicked on the link in the newsletter, but the Amazon Associate program can tell me how many of those people actually bought the book. Seriously, it's like Amazon whipped aside its veil of corporate secrecy, and is giving you straight, raw, glorious data on how effective any and every promotion you do is in ACTUAL SALES through your links, separated out from the casual browser on amazon itself.
When you know how many people are visiting your site on average each weekday, and how many people bought the latest release through said site (use the first 7 days unless you click-to-sale ratios are very different from ours), then that tells you the size of your story-purchasing fan base versus the whole. When you know how many people over the course of a non-release month clicked on your book cover icon on your sidebar (and you DO have your book covers with a link to retailers on your sidebar, right where they can see it, don't you?) compared to how many bought your stories straight off amazon.com, you know the background sales level of just keeping your blog traffic steady. When you can tell how many people opened a newsletter, how many clicked to the page, and how many of those bought the story - that tells you whether your ripest area for improvement is on the newsletter, or on the cover/blurb. If you do a guest post and they let you use your associate tag instead of one of their own (be very polite and ask, if you're not sure), then you can look at the spike and see how effective guest-posting was. If you're on a forum and embed the link in your sig with a cover or two (assuming it's not against forum etiquette), then you can measure the traffic from your frequent posting days to see if the visibility drives sales.
What you can measure, you can improve and measure again to make sure it was an improvement.