What’s the right granularity for location segmentation?

The right answer isn't "as granular as possible." It's "as granular as your business actually needs, and no more than your data can support."

Start with why location matters for your specific program. A global SaaS company usually needs country-level segmentation for language and timezone differences. A retail chain needs city or zip code level for store-specific promotions. A national service with regional pricing needs state or province. There's no universal answer because there's no universal business model.

The practical test: does location genuinely change what you'd send? If you're sending different product catalogs to subscribers in Texas versus California, state-level segmentation is justified. If you're sending slightly different shipping estimates, country-level might be enough. If you'd send the same email regardless of which US state someone is in, state-level segments add maintenance overhead with no return.

Watch out for overly thin segments. A segment of 50 subscribers in a specific zip code sounds precise. It's also too small to generate statistically meaningful engagement data, too fragile for A/B testing, and may harm your sender reputation if you're sending too-small batches. A rough rule: you want at least several hundred subscribers in a segment before drawing conclusions from its performance data.

For most senders, the right progression is: start at country level, then add regional sub-segmentation (state/province/city) only where your content or offers genuinely differ. Test whether more granularity actually improves engagement before building the infrastructure to support it permanently.

Your location data quality also limits how granular you should go. Country data inferred from IP is usually reliable enough. City-level data inferred from IP is less so. If your city-level segmentation is based on shaky inference, you'll create personalization experiences that feel wrong to subscribers who got the wrong city assigned. That's worse than no location personalization at all.