Skip to main content
Learn how to effectively organize your products using collections and groups for better AI content generation and catalog management.

​
Why Taxonomy Matters

Good product organization:
  • ✅ Improves AI content quality (better context)
  • ✅ Enables efficient bulk operations
  • ✅ Allows collection-specific prompt customization
  • ✅ Makes products easier to find and manage
Poor taxonomy = poor AI context = generic content

​
Collections vs. Groups: Quick Recap

FeatureCollectionsGroups
SourceFrom your e-commerce platformDefined in Cernel
AssignmentManualAutomatic (rule-based)
PurposePrimary product taxonomyDynamic segmentation
Use CaseWhat customers seeInternal organization
Full comparison →

​
Best Practices for Collections

​
1. Keep Structure Meaningful

Good Structure: 
Clothing
├── Men's
│   ├── Shirts
│   │   ├── Formal Shirts
│   │   └── Casual Shirts
│   └── Pants
└── Women's
    ├── Dresses
    │   ├── Formal Dresses
    │   └── Casual Dresses
    └── Tops
Poor Structure: 
Products
├── Blue Items
├── Items on Sale
└── Everything Else
Organize by product type and use case, not by temporary attributes (color, sale status).

​
2. Assign Primary Collections Correctly

Every product needs a primary collection for AI context. How to choose:
  1. Most Specific: “Formal Dress Shirts” > “Shirts” > “Clothing”
  2. Use-Based: “Evening Gowns” > “Dresses” (defines purpose)
  3. Permanent: “Casual Wear” > “Sale Items” (doesn’t change)
1

Open Product

Navigate to product detail view
2

Go to Collections Tab

See all assigned collections
3

Set Primary

Click “Set as Primary” on the most appropriate collection

​
3. Avoid Over-Categorization

Bad: Product belongs to 15 collections
  • “Clothing”
  • “Men’s”
  • “Shirts”
  • “Blue Items”
  • “Cotton Products”
  • “Summer Collection”
  • “New Arrivals”
  • “Under $50”
  • …
Good: Product belongs to 3-5 meaningful collections
  • “Men’s Casual Shirts” (primary)
  • “Summer Collection”
  • “Cotton Apparel”

​
4. Ensure Consistent Depth

Keep taxonomy depth consistent across categories: Inconsistent (problematic): 
Electronics
└── Smartphones (depth 1)

Clothing
└── Men's
    └── Tops
        └── Shirts
            └── Formal
                └── Long Sleeve (depth 5)
Consistent (better): 
Electronics
└── Smartphones
    └── Android (depth 2)

Clothing
└── Men's
    └── Formal Shirts (depth 2)

​
Best Practices for Groups

​
1. Use Groups for Dynamic Segmentation

Good Use Cases:
  • “Needs Enrichment”: Products where description is empty
  • “High-Value Products”: Price > $200
  • “New Arrivals”: Created within last 30 days
  • “Incomplete Data”: Missing key attributes
Poor Use Cases:
  • Duplicating your collection structure
  • Creating groups that never change

​
2. Design Clear Conditions

Good Conditions: 
Luxury Items:
- price > $500
- collection contains "Premium" OR brand in ["Gucci", "Prada"]

Needs Description:
- description is empty OR description length < 50 characters
Poor Conditions (too complex): 
Random Group:
- (price > $100 AND color = "Red") OR (price < $50 AND material = "Cotton") OR (brand = "Nike" AND NOT size = "Large")

​
3. Monitor Auto-Selected Primary Groups

Cernel automatically selects primary groups based on specificity.
1

Spot-Check Products

Open random products and view their primary group
2

Verify Logic

Does the auto-selected primary make sense?
3

Adjust if Needed

If consistently wrong, revise group structure or conditions

​
Common Taxonomy Issues

​
Issue 1: Products Without Primary Collection

Symptom: Enrichment fails with “No primary collection” Solution:
1

Identify Affected Products

Filter products without primary collections
2

Assign to Collections

Add each product to at least one collection
3

Set Primary

Mark one collection as primary
4

Retry Enrichment

Products can now be enriched

​
Issue 2: Wrong Primary Collection

Symptom: AI generates off-brand or irrelevant content Example:
  • Product: Luxury silk scarf
  • Primary Collection: “Blue Products” (wrong—focuses on color, not product type)
  • Better Primary: “Women’s Luxury Accessories”
Solution: Change primary collection to the most product-defining category

​
Issue 3: Messy Collection Structure

Symptom: Hard to find products, inconsistent AI quality Solution: Reorganize collections
  1. Plan new structure on paper
  2. Create new collections in your e-commerce platform
  3. Reassign products
  4. Sync to Cernel
  5. Update primary collections
  6. Re-enrich products with new context

​
Issue 4: Primary Group Keeps Changing

Symptom: Primary group changes unexpectedly Cause: Groups are dynamic—when product attributes or group conditions change, membership recalculates Solution:
  • Accept it: This is by design for dynamic groups
  • Manual override: Set primary group manually if stability is critical
  • Adjust conditions: Make group conditions more stable/robust

​
Taxonomy Strategies by Catalog Size

​
Small Catalogs (<1,000 Products)

Approach: Simple, manual organization
  • Collections: 2-3 levels deep
  • Groups: 5-10 dynamic groups for utility
  • Primary Assignment: Manual review and assignment
  • Maintenance: Easy to keep clean

​
Medium Catalogs (1,000-10,000 Products)

Approach: Structured taxonomy with some automation
  • Collections: 3-4 levels deep
  • Groups: 15-30 groups for segmentation
  • Primary Assignment: Mostly automatic with spot-checking
  • Maintenance: Quarterly reviews

​
Large Catalogs (10,000+ Products)

Approach: Highly automated, rule-based
  • Collections: Imported from platform, kept synchronized
  • Groups: 30+ groups with complex conditions
  • Primary Assignment: Fully automated
  • Maintenance: Automated monitoring + exception handling

​
Workflow: Cleaning Up Taxonomy

1

Audit Current State

  • Count products per collection
  • Identify orphaned products (no collection)
  • Find over-categorized products (too many collections)
2

Define Ideal Structure

  • Sketch target taxonomy on paper
  • Aim for 3-4 levels max
  • Group by product type, not attributes
3

Implement in Platform

  • Create new collections in your e-commerce platform
  • Move products to correct categories
  • Delete obsolete collections
4

Sync to Cernel

  • Trigger manual sync (Shopify)
  • Or re-upload feed
5

Update Primary Collections

  • Bulk update primary collections in Cernel
  • Validate with spot-checks
6

Re-Enrich

  • Re-enrich products to regenerate content with improved context

​
Using Taxonomy for Prompt Customization

Leverage your taxonomy to customize AI behavior:

​
Collection-Specific Prompts

Example: 
Base Prompt (Organization Level):
"Write a product description highlighting features and benefits."

Luxury Collection Override:
"Write an elegant description emphasizing craftsmanship and exclusivity."

Budget Collection Override:
"Write a practical description emphasizing value and affordability."
When enriching:
  • Products in “Luxury” collection → Get luxury-specific tone
  • Products in “Budget” collection → Get value-focused tone
  • Other products → Get base prompt
Learn more about hierarchical prompts →

​
Maintenance Schedule

Weekly:
  • Check for new products without primary collections
  • Verify recent enrichment quality
Monthly:
  • Review auto-selected primary groups
  • Clean up obsolete collections
  • Update group conditions as needed
Quarterly:
  • Full taxonomy audit
  • Reorganize if structure has drifted
  • Re-enrich affected products

​
What’s Next?

Collections & Groups Reference

Detailed guide to taxonomy concepts

Primary Taxonomy

How primary collections and groups are selected

Bulk Operations

Use taxonomy for efficient bulk enrichment

Customizing Prompts

Create collection-specific prompts
Need help restructuring your taxonomy? Email [email protected]