Customer taste profiles: personalise recommendations without being creepy

Personalisation works when it feels like good service and fails when it feels like surveillance. The difference is how you gather information and what you do with it.

Three kinds of data

• Zero-party data: what customers tell you on purpose, such as answering "what's the occasion?" or choosing favourite styles. This is the best kind.
• First-party data: what you observe on your own store, such as purchases and pages viewed.
• Third-party data: bought or inferred from other sources. It is the riskiest and least reliable, and we do not recommend it.

A taste profile built mostly from zero-party answers is accurate, respectful and easy to explain.

What a taste profile contains

For a fashion or home brand it might hold: preferred occasions, styles, colours, materials, budget band, sizes and a few favourite products. Keep it small. Each extra field must earn its place by improving recommendations.

Collect it the right way

1. Ask simple questions with tap-to-answer choices, in a chat or on your site.
2. Explain why: "So I can suggest pieces you'll like."
3. Make it optional. Shoppers can skip it and still buy.
4. Record consent and the purpose.

Use it to help

• Recommend products that match (how recommenders work).
• Send restock alerts for items they wanted, with their permission (opt-ins).
• Open conversations with relevant new arrivals.
• Improve your own range: if many people want a style you do not stock, that is a signal.

What to avoid

• Guessing sensitive characteristics, such as health, religion or finances.
• Making the customer feel watched: "we saw you looked at this three times".
• Using the data for a purpose you did not tell them about.
• Keeping profiles for ever without a reason.

Principles from India's DPDP Act

Without turning this into legal advice, the Act's core ideas are good practice anyway: get clear consent for a specific purpose, collect only what you need, keep it secure, and let people withdraw consent and ask for correction or erasure. See DPDP for online stores. If you sell to Europeans, the GDPR adds more (GDPR guide).

Store it safely

Keep profiles in your store's own protected database, limit who can access them and log access. A platform that gives each store its own database makes this separation simpler (dedicated databases).

Give customers control

Offer a simple way to view, edit and delete their preferences, for example a "reset my profile" command in chat. People trust brands that make this easy.

What to do next

Write down the five questions that would most improve your recommendations, and drop any that do not. Then ask them in your next customer chat, and measure whether recommended products sell better.