Some of the advice I’ve seen around social media measurement boils down to “Don’t pay attention to x. You should be measuring y.”

Don’t pay attention to retweets; you should be measuring follower growth. Don’t pay attention to follower growth; you should be measuring post likes. Don’t pay attention to post likes; you should be measuring FlegmaRank, our proprietary new index based on a secret algorithm that boils eighty thousand different variables into a single integer between 0 and 1.

It’s enough to drive an online campaigner to drink… or, more productively, to the bookshelf. This stuff is why I was so pleased to draw the cartoons for Beth Kanter and K.D. Paine’s Measuring the Networked Nonprofit. And why I loved reading Katie’s Measure What Matters. And (this’ll take you back) Avinash Kaushik’s Web Analytics: An Hour a Day. (I’d link to it, but he has a more recent follow-up, Web Analytics 2.0.)

What all that advice really should boil down to is this: what do you want your online efforts to achieve? How do you believe they’ll do it? (That is, what’s your theory of change?) How can you measure each stage of the mechanism underlying your theory? How can you benchmark against peers, competitors and past performance? And how can your measurements help you assess your model against real-world results, and adjust accordingly?

Answer those questions, and you’ll know which metrics matter. (They may be the usual suspects. They may not. Chances are they’ll be some of each.) Everything else is noise.