If your A/B test shows a statistically significant increase in user engagement but a drop in system performance, how would you make a launch decision?
