"Decide what you’re testing upfront (e.g., “If A, then B”) so you don’t succumb to wishful thinking and forced causation. This is also true for online analytics."

Measuring Social Media ROI: A Case Study (Plus: Tweet to Beat Winners)