Skip to main content
Every product in Cernel has a primary collection and a primary group—the single most appropriate category classification for that product. Understanding how these are selected and why they matter is crucial for generating high-quality AI content.

Why Primary Taxonomy Matters

Context for AI Generation

The primary collection/group provides essential context for the AI: Example: A “Dress”
  • Primary Collection: “Formal Evening Wear” → AI generates elegant, sophisticated content about special occasions
  • Primary Collection: “Casual Summer Dresses” → AI generates relaxed, everyday content about comfort and versatility
Same product type, completely different content based on primary taxonomy.
Products without a primary collection cannot be enriched. The AI requires this context to generate appropriate content.

Prompt Customization

Prompts can be customized per collection/group:
  • Set a base prompt for all products
  • Override with collection-specific prompts (e.g., luxury vs. budget tone)
  • Cernel automatically uses the most specific prompt available
The primary collection/group determines which prompt is used. Learn more about hierarchical prompts →

Primary Collection

How Primary Collections Are Set

Primary collections are typically set by your e-commerce platform:
1

Platform Assignment

When you categorize a product in Shopify (or your platform), you assign it to one or more collections
2

Primary Collection Import

Cernel imports the main collection as the primary collection during sync
3

Manual Override (if needed)

If the imported primary collection is incorrect, you can manually change it in Cernel

Multiple Collections vs. Primary

A product can belong to multiple collections, but only one is primary: Example:
Product: Blue Silk Evening Gown

Collections:
- Dresses (general category)
- Evening Wear (occasion)
- Blue Clothing (color-based)
- Sale Items (promotional)

Primary Collection: Evening Wear ← This drives AI content context
Why “Evening Wear” is Primary:
  • Most specific to the product’s purpose
  • Provides clear context for tone and styling
  • Not temporary (unlike “Sale Items”)
Choose the primary collection that best describes what the product is and how it’s used, not just general attributes or temporary states.

Changing Primary Collection

1

Open Product

Navigate to the product detail view
2

Go to Collections Tab

View all collections the product belongs to
3

Set Primary

Click “Set as Primary” next to the most appropriate collection
4

Save

Changes take effect immediately for future enrichments
Changing primary collection doesn’t retroactively change already-enriched content. It only affects future enrichments for this product.

Primary Group

How Primary Groups Are Selected

Unlike primary collections (manually set), primary groups are automatically selected by Cernel using intelligent heuristics.

Selection Algorithm

Cernel evaluates all groups a product belongs to and selects the primary based on:
1

1. Specificity

More specific groups are preferred over general onesExample:
  • “Men’s Formal Dress Shirts” (specific) ✅
  • “Shirts” (general) ❌
2

2. Hierarchy Depth

Deeper nodes in the group tree are more specificExample Tree:
Clothing (depth 0)
└── Men's (depth 1)
    └── Formal Wear (depth 2)
        └── Dress Shirts (depth 3) ← Deepest = most specific ✅
3

3. Condition Complexity

Groups with more conditions are more specificGroup A (simple): category = "Shirts" Group B (complex): category = "Shirts" AND material = "Cotton" AND fit = "Slim"Group B is more specific.
4

4. Alphabetical (tie-breaker)

If multiple groups are equally specific, alphabetical order breaks the tie

Example Primary Group Selection

Product: Navy Cotton Dress Shirt Group Memberships:
  1. “All Products” (root)
  2. “Men’s Clothing” (category = “Men’s”)
  3. “Shirts” (type = “Shirt”)
  4. “Dress Shirts” (category = “Men’s” AND type = “Shirt” AND style = “Dress”)
  5. “Blue Apparel” (color contains “Blue” or “Navy”)
Primary Group Selection:
  • Winner: “Dress Shirts” (depth 3, multiple conditions)
  • Runner-up: “Blue Apparel” (depth 2, but less relevant to product purpose)
Cernel selects “Dress Shirts” as primary because it’s the most specific group relevant to what the product is, not just an attribute it has (color).
Primary group selection prioritizes product identity over product attributes. A red shirt’s primary group should be about it being a shirt, not about it being red.

Viewing Primary Group

1

Open Product

Navigate to product detail view
2

Check Header or Sidebar

The primary group is displayed prominently (often with a badge or special indicator)
3

View All Groups

See all groups the product belongs to and which one is marked as primary

Manual Override (if needed)

While primary groups are auto-selected, you can override if needed:
1

Open Product

Go to product detail view
2

View Groups

See all groups the product belongs to
3

Set Primary

Click “Set as Primary” next to a different group
4

Save

Manual override is saved and used for future enrichments
Manual overrides are rare. The automatic selection algorithm works well for most cases. Only override if the auto-selected primary clearly doesn’t make sense.

Primary Taxonomy Best Practices

Before enriching, verify all products have a primary collection assigned. Products without one will fail enrichment.Check: Run a filter for products without primary collections and fix before bulk operations.
“Evening Gowns” is better than “Dresses” for a formal dress. “Running Shoes” is better than “Footwear” for athletic shoes.The more specific, the better the AI context.
Don’t use “Sale Items” or “New Arrivals” as primary collections—these change over time.Use permanent classifications like product type, style, or use case.
Spot-check auto-selected primary groups for a sample of products. If the algorithm consistently gets it wrong for a category, you may need to adjust your group structure or conditions.
If one product category has 5 levels of hierarchy and another has only 2, the deeper one will always be considered more specific—even if it’s not.Aim for consistent depth across your taxonomy.
While collections and groups serve different purposes, aligning their structure reduces confusion.Example:
  • Collection hierarchy: Clothing > Men’s > Shirts
  • Group hierarchy: All Products > Men’s > Shirts
Parallel structures make it easy to understand product organization.

Troubleshooting Primary Taxonomy

Error: “Product must have primary collection before enrichment”Solution:
  1. Open the product in Cernel
  2. Go to Collections tab
  3. Assign the product to at least one collection
  4. Set one as primary
  5. Retry enrichment
Issue: Product has correct collections but wrong one is marked primary.Solution:
  1. Open product > Collections tab
  2. Click “Set as Primary” on the correct collection
  3. Future enrichments will use the corrected primary collection
  4. Optionally re-enrich the product to regenerate content with proper context
Issue: Auto-selected primary group seems illogical.Solution:
  1. Review the group conditions—does the product actually match them?
  2. Check if there’s a more specific group the product should belong to
  3. Consider revising group structure or conditions
  4. If needed, manually override the primary group for this product
Issue: Primary group changes unexpectedly when product attributes update.Solution: This is expected behavior—primary groups are dynamic and recalculate when:
  • Product attributes change
  • Group conditions are modified
  • New groups are created
If you need stable primary groups, use manual overrides or make group conditions more robust.

Impact on Content Quality

Good Primary Taxonomy → High-Quality Content

Example: Yoga Mat
Primary Collection: "Yoga & Pilates Equipment"

Generated Description:
"This premium yoga mat provides excellent grip and cushioning for your
practice. The eco-friendly material is durable yet comfortable, perfect
for vinyasa, hatha, or restorative yoga. Non-slip surface ensures
stability during challenging poses..."

✅ Context-appropriate content focused on yoga use cases

Poor Primary Taxonomy → Low-Quality Content

Example: Same Yoga Mat
Primary Collection: "Blue Products"

Generated Description:
"This blue mat is a great choice if you're looking for something blue.
The blue color matches other blue items. Made from blue-friendly
materials..."

❌ Content focuses on color (because that's all the context provided)
❌ Misses the actual purpose and use case
Bad taxonomy = bad content. The AI is only as good as the context you provide through primary collections and groups.

Advanced: Multi-Level Primary Taxonomy

For complex catalogs, consider primary taxonomy at multiple levels: Level 1: Primary Collection (What it is)
  • Example: “Dress Shirts”
Level 2: Secondary Collection (Style/Occasion)
  • Example: “Formal Wear”
Level 3: Tertiary Collection (Seasonal/Promotional)
  • Example: “Summer Collection”
Cernel uses the primary collection for core context, but you can configure prompts to also reference secondary/tertiary classifications for nuanced content.

What’s Next?


Next: Explore Integrations to connect your e-commerce platforms to Cernel.