The planning team has three different questions: how many buyers a channel reaches, which delivery points are near a depot, and how a small fitted model behaves on observations it didn't see. Each question needs its own output contract. A distinct estimate isn't a row count, geographic distance isn't driving distance, and a prediction isn't an observed result.
Use the fresh filters and boundaries in each task rather than reusing a number from a previous exercise. The hidden data changes, so the query must express the rule. Some tasks ask for an approximate function even when an exact result would be easy on this small dataset. That requirement practices the syntax; it doesn't imply that approximation is the right choice for every small report. The final choices revisit the limits that belong beside an analytical result.