The short answer
They design territories from data and then defend the design politically. Accounts get scored on multi dimensional signals, size, industry, technographics, growth triggers, not just postcode, and the carve is optimised so expected opportunity per rep is roughly equal rather than account count being equal. Field patches are also optimised for travel time, because hours in a car are hours not selling. The operating discipline that separates the good from the rest is a fixed annual cadence, a published set of rules for what happens to in flight deals when a patch moves, and rep input collected before the design is locked. Then they measure the design itself: not just attainment, but how evenly attainment is spread.