Prompt
I’m tired of our homepage recommendations not changing with user behavior
Latest observation
Oct 13, 2026
OpenAI APIWeb search: off
That makes sense — static homepage recommendations can feel stale fast.
A good fix is to make recommendations respond to recent user behavior in near real time, for example:
- pages viewed
- categories browsed
- items clicked, saved, or purchased
- session recency and frequency
A practical approach:
- Track key events on the site.
- Update a user profile or feature vector continuously.
- Use a ranking model that re-scores homepage modules based on those signals.
- Add fallback logic for new or sparse users.
- Measure impact with A/B tests and monitor CTR, conversion, and bounce rate.
If you want, I can help you design:
- a simple rule-based version,
- a production ML-based recommender,
- or a real-time architecture for this.