Compare Trends and Useful Segments

A rate is a level. Whether it is rising or falling is a different fact, and a reader given only one of them will draw the wrong conclusion roughly half the time.

A segment at a low rate that has doubled over the period is a success story. The same segment at the same rate having halved is an emergency. The level is identical. Every action you would take is opposite.

So report both, together, always. This costs one column.

The trend can invert the headline

In a real analysis, one segment is presented with a low conversion rate and, in the same breath, the fact that the rate has been falling sharply across the period. Those two facts together support a conclusion neither supports alone: this is not a chronically weak segment, it is a deteriorating one, and whatever changed is recent and probably findable.

Without the trend, the reasonable reading is that the segment is simply poor and should be deprioritized, which is precisely the wrong response to something that used to work.

A blank segment is a finding

Carried from the previous guide and repeated here because it shows up first in a report: a large blank bucket in a grouping field is data quality surfacing, not a category. Show it, label it, do not analyze it as though it were a segment.

Which segments are worth cutting

Not all of them. A segment is worth reporting if somebody could act differently on it. Segments nobody can act on differently are decoration, and each one costs the reader attention that the actionable ones needed.