One side odd, one side even

American streets almost universally alternate parity: odd numbers on one side of the street, even on the other. This lets a courier know which side of the street to drive down before reading anything else.

Generated test data should respect this convention — an address like "442 Main St" is fine, but validators rarely check parity, so both odd and even numbers are useful in a sample set.

Blocks, not counting

Rather than counting doors one by one, most cities number by block: each block adds a fixed step, often 100. House 1521 sits in the block between 1500 and 1599, which immediately tells you the fifteenth block of the street.

That structure is why thousands-style numbers (like 1600 Pennsylvania Avenue) are completely normal in US cities — the grid absorbs them.

Why test data should vary numbers

Reusing a single house number across hundreds of records is a red flag for deduplication logic and rate limiters. Good sample data varies the number while keeping street and ZIP stable, so you can watch your system group and deduplicate correctly.