|December 15th, 2012|
|experiment, contra, bida [html]|
For a while I've written off our advertising as too hard to test, but now I'm not sure. If we were to randomly run ads before some dances and not before others, we could see how this matched to our attendance. The signal is pretty noisy, though. I wonder how long we would need to run the test to tell?
Since May 2011 (1.5 years) we've had 353 clicks, or about 10 clicks per dance. Our average attendance is about 100 people. I don't know enough stats to say how long we would have to run a test (at two dances (trials) a month) to find whether there's an effect and if so whether it's large enough to be worth it.