You are reviewing results from an A/B test on a growth surface in a digital product, and the treatment changes a conversion metric with a fairly low baseline rate. Different stakeholders are quoting the result as either an absolute percentage-point lift or a relative percent lift, and the framing changes how large the impact appears.
How would you decide whether to report absolute lift or relative lift when presenting experiment results? Explain when each is more appropriate and how you would make sure the interpretation stays statistically and practically sound.
Relative lift can overstate business importance when baseline is smallAbsolute lift can under-communicate proportional improvement across experimentsPeeking can bias which framing gets emphasizedSRM can invalidate either presentation