One thing organizations that do A/B testing have to agree on is how to treat inconclusive experiments. 

Inconclusive result means statistically you can neither say it is positive nor negative, and therefore instead of deriving conclusions on whether a feature is good or bad, one should analyze the goals and come up with the next hypothesis until it becomes conclusive or be regarded as failure and move to the next idea.

Sometimes I think A/B testing is like the stock market, you should be disciplined in your approach for your investments to yield the best results. There is an element of loss aversion, and fear of appearing as a failure. Discipline solves most of this.