How to handle sensitive attributes (gender, location, income)?

Before collecting any sensitive attribute, the first question to answer is whether you actually need it for segmentation that can't be achieved another way. That's not rhetorical. Sensitive data carries real privacy and legal risk, and the threshold for "worth it" should be higher than for general behavioral data.

What "sensitive" means in two senses:

Legally, GDPR defines a specific list of "special category" data that requires explicit consent and has processing restrictions: racial or ethnic origin, political opinions, religious beliefs, health data, sexual orientation, and similar. These need a documented lawful basis beyond standard consent and should only be collected when you have a genuinely specific use case (a healthcare provider personalizing health information, for example).

Ethically, plenty of attributes aren't on the legal list but still carry discrimination risk or feel intrusive to subscribers: income, household size, relationship status, precise geolocation. These don't require explicit consent under most laws, but mishandling them can produce discriminatory targeting effects or damage trust if subscribers feel surveilled.

Gender: If you're collecting it (say, for product sizing or salutation preferences), make the field optional, offer more than two options, and don't use it for messaging assumptions. Don't write different email copy based on gender unless the relevance is direct and obvious. Avoid reinforcing stereotypes.

Location: Country or region is usually enough for timezone, language, and offer relevance. Precise geolocation is a different category. If you're asking for it, explain clearly why. "We use your city to show you local store events" is a reasonable disclosure. Using location inferred from IP without disclosure is the gray area to avoid.

Income: Rarely worth collecting explicitly. If you need to segment by economic behavior, purchase history is a better proxy. It reflects actual behavior rather than self-reported or inferred numbers that are often wrong and can produce discriminatory effects in targeting.

Regardless of the attribute, give subscribers easy access to view, correct, and delete their sensitive data. For the broader discrimination risk that sensitive attributes can create in segmentation, the segmentation discrimination guide covers proxy discrimination. And for the compliance and consent framework that applies here, the privacy implications overview is the right foundation.