If, like me, you drifted away from Hanson's newsletter because it became too much of a cheerleader for internet radio issues and became less relevant to 'our' concerns and problems, it's time to wake up and pay attention again, at least to his blog:
Suppose we used to survey 10 Chicagoans every week to find out what percentage are Chicago Blackhawks fans. That means over the course of a calendar quarter, we’ll talk to 130 people. That’s probably a decent sample size to answer a question like this. (I could do the math on standard deviation and margin of error, but let’s save that for another day.) The analogy for Arbitron’s new approach is to survey maybe 30 Chicagoans — but survey them every week. “Hey, that’s 390 data points every quarter! Better than the old system’s 130 data points!” No, it isn’t. In week #1, 30 people is too small a sample size for the question — we could get a result that’s way off of reality. And surveying the same people every week for 12 more weeks is going to give us, basically, the same wrong answer every week. Yes, the results will be more stable than under the old, 130-person system, but that doesn’t mean it’s a better approach... ...Because Arbitron was too cheap to throw around a few $2,000 bills toward consulting contracts, their stock price took a $200 million hit yesterday — and they’ve put a not-insignificant portion of radio’s annual $20 billion in ad revenues is now at risk, too.