How can Tweets be tested?
Unfortunately, it’s not possible to A/B test tweets – yet (maybe someday there’ll be an app for that). If you’re handy with numbers, Rand Fishkin from SEOmoz has a process for hypothesis testing tweets you can try. (Interestingly, Rand found shorter tweets produced higher click through in his analysis).
Again, click through isn’t the perfect measure of engagement or interest. You can’t count how many people read your tweet and got value out of it without retweeting or replying to you, but you can measure which tweets spurned the most unfollows with Tweeteffect. It’s a good idea to make note of content that causes an unusual exodus, regardless of the length, link placement, use of adverbs or time of day.