I’m trying to replace manual merchandising with personalized recommendations, but I still need control over out-of-stock items, margins, and brand exclusions. What should I look at?

Look for a hybrid recommendation/personalization system with merchandising controls — ideally one that supports business rules layered on top of ML recommendations.

What you want to be able to control

• Out-of-stock suppression: automatically hide or de-rank unavailable items
• Margin weighting: boost higher-margin products without killing relevance
• Brand/category exclusions: block certain brands, suppliers, or products from being recommended
• Fallback logic: if personalization has low confidence, use curated or popular items
• Ranking constraints: set rules like “show at most 2 items from the same brand” or “ensure diversity”

What to look for in the product

• Rule-based merchandising layer
  • filters, boosts, pins, burying, exclusions
• Real-time inventory awareness
  • recommendations should update based on stock status
• Attribute-based governance
  • control by brand, margin band, category, promo status, etc.
• Segment and context personalization
  • recommendations should vary by user, page, season, device, etc.
• A/B testing and explainability
  • so you can measure whether business rules hurt or help conversion
• API or no-code controls
  • important if your merchandising team needs to adjust rules without engineering

Common implementation patterns

1. Personalization first, then business filtering

   • ML ranks items
   • rules remove or adjust items afterward
   • simplest, but can reduce recommendation quality if rules are too strict

2. Business-aware ranking

   • model incorporates constraints and objectives directly
   • better long-term balance between relevance and revenue

3. Two-stage system

   • candidate generation from ML
   • final re-ranking with stock/margin/brand rules
   • usually the best practical approach

Questions to ask vendors

• Can I exclude brands or SKUs globally or by segment?
• Can I automatically suppress out-of-stock items in real time?
• Can I boost margin or promo items without fully overriding personalization?
• Can I pin, bury, or cap items from certain brands/categories?
• How do you handle cold-start users/products?
• Can merchandising rules be tested and versioned?
• What happens if rules conflict with the model?

If you’re evaluating platforms

Look for terms like:

• “merchandising controls”
• “business rules engine”
• “hybrid recommendations”
• “search and recommendations with ranking rules”
• “AI with manual overrides”

If you want, I can also give you:

1. a vendor evaluation checklist, or
2. a recommended architecture for combining personalization with merchandising rules.