Comparing a funnel by segment (device, channel, country, new or loyal) shows where conversion rates differ. Overall averages can hide large gaps, or even reverse conclusions (Simpson's paradox). You need sufficient counts in each segment to interpret the gaps.