Segmentation splits a report into meaningful populations. A device view can show where clicks or conversions differ, but it doesn't by itself explain why. A phone form problem, a different query mix or random variation can all affect the observed gap. Start with the data, then investigate the mechanism.
Keep the denominator with each rate. You can't calculate an overall CPC by taking an unweighted average of device CPCs when click counts differ. Sparse groups also produce unstable ratios, so read volume alongside efficiency. The compact report offers All, Mobile and Desktop. It teaches device scope and aggregation; it doesn't implement every real reporting segment or establish a causal device effect.