×![]()
Click anywhere to close
Before trusting a recommendation widget, look at what it has learned. Four places show you the data behind recommendations:
- Campaign readiness — every recommendations campaign’s side panel scores the pipeline live: catalog in place, interactions arriving (view / add-to-cart / purchase each shown separately), and whether the engine is still in its collecting phase — with an Expedite option to upload historical transactions.
- The catalog viewer — Settings → Recommendations: the items themselves, their fields and images, exactly as the engine sees them (feed setup).
- Live events — Live visits shows interactions as they arrive, per visitor — the fastest way to confirm tracking end to end (tracking setup).
- Visitor interests — the inferred interest profile each visitor accumulates, which interest-based algorithms rank by: interests for recommendations. Also readable per user over the REST user_interests object.
What the engine learns from what
- Views and add-to-carts drive the ranking — they say what attracts.
- Purchases close the loop — they attribute revenue and teach bought-together patterns.
- Category views build interests, powering “more from what they care about” logic.
Once data flows, results appear in the campaign’s Performance step and the recommendations reports — metric definitions in the analytics glossary.