Skip to content
Glossary

Personalization

Personalization is the practice of tailoring offers, copy, and product surfaces to each shopper's intent, history, and segment. Done well, it feels like attention. Done poorly, it feels like surveillance.

In context

The personalization spectrum runs from segment-level ("first-time visitors see free shipping banner") through cohort-level ("returning customers from paid social see cross-sell carousel") to 1:1 real-time ("this specific shopper, who has clicked running gear three times in the last 30 days, gets a hydration-pack recommendation").

Real-time 1:1 used to require heavy infrastructure. AI agents and embedding-based retrieval make individual-level personalization practical for smaller stores too.

How Zubby uses this

Zubby personalizes three surfaces: the agent's reply (the shopper's language, the page they are on, country-targeted promotions, and, for verified shoppers, their order history), product recommendations (re-ranked against the shopper's implicit and explicit signals), and lifecycle email/SMS (cart contents and recently viewed products).

Stored order history is only used when your storefront has verified who the shopper is through a supported verified identity flow, such as a signed login token. An email address typed into chat never unlocks a shopper's profile history. A guest can separately request a specific order using the order-number and checkout-email checks described in the order guide; that is not broad profile personalization or permission to change an order.

Illustrative example: use the current product page and a shopper's stated budget to suggest an alternative, without collecting their email. Use prior-order context only when identity is verified and the history is available; missing data should lead to a question, not an invented preference.

Product recommendations in Zubby