How do you know when a segment is too small?

When you can't learn from it or act on it meaningfully.

For testing: You need at least a few thousand recipients to get statistically significant results. A segment of 500 might show a 15% open rate, but that number could easily be 12% or 18% with normal variation. You can't trust the data.

For campaigns: If creating custom content for a tiny segment takes more time than the revenue it generates, the segment is too small to justify separate treatment.

For deliverability: Very small sends can actually look suspicious to mailbox providers. Reputation builds through consistent volume patterns, not sporadic micro-campaigns.

A practical minimum depends on your total list size. For lists under 10,000, segments below 500 are probably too small. For lists over 100,000, segments below 2,000 start losing value.

Small segments can still be useful when merged with similar groups or when used for exclusion rather than targeting. But if a segment is so small that it exists alone with custom treatment, question whether it belongs.