The shop had traffic. Average order value was flat. We shipped “frequently bought together”, a personalised home and smarter search. AOV +67% in a quarter — not magic, better relevance.
What worked
1. Collaborative filtering on order history 2. Merchandising rules (margin, stock) on top of pure ML 3. Search with synonyms and typo tolerance 4. A/B — not “turn on for everyone on day one”
What didn’t
Recommendations for out-of-stock items (frustration) and aggressive pop-ups that killed UX.
Conclusion
AI in e-commerce should show up on the receipt, not only in a slide deck.
_Want to lift AOV?_ We’ll review catalogue and data — then propose the first experiment.
Keywords
AI ecommerce
shop personalisation
recommendation engine
AOV
AI retail
