AB Tasty offers the automatic calculation of a statistical reliability indicator for each objective. This index (Khi-2 index) allows you to extrapolate the results of a test over a longer period of time.
Thus, an objective with a very good reliability rate has very good chances of behaving, “on average,” like the historical test data.
Statisticians use the threshold of 95% per use, which seems to be a good threshold to reach before making any definitive decisions based on the tool.
More details are available here on statistical considerations: http://en.wikipedia.org/wiki/Chi-squared_test