Skip to main content
Bulk operations allow you to enrich hundreds or thousands of products at once, transforming your entire catalog efficiently. Instead of processing products one by one, Cernel’s bulk enrichment runs in parallel, saving you hours or even days of work.

​
Why Bulk Enrichment?

The Challenge: With catalogs of 500, 5,000, or 50,000+ products, enriching one product at a time is impractical. The Solution: Bulk operations let you:
  • Enrich entire collections or groups simultaneously
  • Generate consistent content across similar products
  • Complete catalog-wide initiatives (like SEO improvement) in hours instead of weeks
  • Scale content production as your business grows
Cernel processes products in parallel—enriching 1,000 products takes only slightly longer than enriching 100.

​
Bulk Enrichment Methods

​
1. Select All Products

The simplest approach: enrich your entire catalog.
1

Go to Products → All Products

Navigate to the full product list
2

Select All

Click the header checkbox, then “Select All [X] Products” to select across all pages
3

Click Enrich

Open the enrichment modal
4

Choose Attributes

Select which attributes to generate
5

Start Job

Launch the bulk enrichment
Enriching thousands of products consumes significant tokens and time. Consider testing with a small batch first to validate quality before processing your entire catalog.

​
2. Enrich by Collection

Enrich all products in specific categories for consistency.
1

Navigate to Collections

Click “Collections” in the sidebar
2

Select Collections

Use checkboxes to select one or more collections (e.g., “Men’s Shirts,” “Winter Accessories”)
3

Click Enrich

All products in selected collections will be enriched
4

Configure & Start

Choose attributes and launch
Benefits:
  • Consistency: All products in “Formal Wear” get the same tone and style
  • Context: The AI understands collection-specific nuances
  • Manageability: Easier to review 200 shirts than 5,000 mixed products
Enrich by collection when you want consistent brand voice and style across product categories.

​
3. Enrich by Group

Use dynamic groups for targeted bulk operations. Example Use Cases: Group: “Products Needing Descriptions” 
Conditions: description is empty OR description length < 50 characters
Action: Enrich description attribute for all products in group
Group: “High-Value Products Without SEO” 
Conditions: price > $100 AND meta_description is empty
Action: Prioritize SEO enrichment for valuable products
Group: “New Arrivals - Last 30 Days” 
Conditions: created_at > -30 days
Action: Enrich all new products as they arrive
1

Create Group

Define conditions that identify target products
2

Navigate to Groups

Go to the Groups page
3

Select Group

Check the box next to your target group
4

Enrich

All products matching the conditions are enriched
Learn more about groups →

​
4. Filtered Bulk Enrichment

Combine search and filters for precise selection.
1

Go to Products

Navigate to All Products
2

Apply Filters

Use filters to narrow down:
  • Specific collections
  • Price range
  • Attribute values (e.g., color = Red)
  • Enrichment status
3

Select Filtered Results

Click “Select All” to select only products matching your filters
4

Enrich

Process the filtered subset
Example: “Enrich descriptions for all products in ‘Electronics’ priced over $500 without existing descriptions.”

​
Bulk Enrichment Best Practices

Always enrich 10-20 sample products first. Review quality, refine prompts, then scale to thousands.Why: If your prompts produce poor quality, you don’t want to apply that to your entire catalog. Testing catches issues early.
Instead of enriching 10,000 mixed products, do:
  • Batch 1: Electronics (500 products)
  • Batch 2: Clothing (1,200 products)
  • Batch 3: Home Goods (800 products)
Why: Category-specific review is easier, and you can adjust prompts between batches.
Enrich your best-sellers, hero products, and high-margin items first.Why: These products drive the most revenue, so quality matters most. Low-traffic products can be batch-processed with less scrutiny.
If enriching 50,000 products, split into multiple jobs of 5,000-10,000 each.Why: Smaller jobs are easier to monitor, review, and troubleshoot. If something goes wrong, you haven’t committed your entire catalog.
Large jobs can take 30 minutes to several hours. Check the Dashboard periodically for failures or issues.Why: Catching problems early (e.g., API rate limits, validation errors) lets you address them before the entire job completes.

​
Managing Large-Scale Enrichment

​
Job Performance at Scale

Processing Speed:
  • ~2-5 seconds per product for simple attributes (color, size)
  • ~10-15 seconds per product for text content (descriptions)
  • ~20-30 seconds per product for multiple complex attributes
Typical Job Durations:
  • 100 products x 3 attributes = 5-10 minutes
  • 1,000 products x 3 attributes = 30-60 minutes
  • 10,000 products x 3 attributes = 4-6 hours
Jobs run in the background—you can close your browser and they continue processing. Check back later to review results.

​
Handling Partial Failures

With large jobs, some products will likely fail (missing data, API issues, etc.). Cernel handles this gracefully:
  1. Job completes with “Partial Success” status
  2. Successful products are ready for review and application
  3. Failed products show error messages
  4. Retry failures without re-processing successes
A 95% success rate on a 5,000-product job is normal. Review successes, fix failures (e.g., assign missing primary collections), and retry just the 250 that failed.

​
Bulk Review Strategies

Reviewing thousands of enriched products can be overwhelming. Use these strategies:

​
1. Sampling Strategy

Don’t review every single product—sample strategically:
1

Review Sample Products

Review a random 10% sample of enriched products to assess overall quality
2

Assess Quality

If the sample shows consistently good quality, proceed with batch-apply for the rest
3

Individual Review for Issues

Products with errors or questionable content deserve individual attention
4

Spot-Check Applied Content

After application, manually check a few random products in your store to ensure they look good

​
2. Delegate Review

For very large jobs, divide review among team members:
  • Team Member A: Reviews “Men’s Clothing”
  • Team Member B: Reviews “Women’s Clothing”
  • Team Member C: Reviews “Accessories”
Each person reviews their section independently, then all apply together. Learn more about user management →

​
3. Progressive Application

Apply in waves:
  1. Wave 1: Review and apply top 100 products
  2. Wave 2: If quality is good, apply next 500 without review
  3. Wave 3: Batch-apply remaining thousands
  4. Spot-Check: Randomly inspect final results

​
4. Filter-Based Review

Use job table filters to focus on specific subsets:
  • Review products by collection (one category at a time)
  • Review products with specific attributes (e.g., only check meta descriptions)
  • Review failures first, successes later

​
Common Bulk Scenarios

​
Scenario 1: Complete Catalog Enrichment (New Store)

Goal: Enrich 2,000 products for a new e-commerce launch. Workflow:
  1. Test on 10 sample products, refine prompts
  2. Enrich by collection (5 batches of ~400 products each)
  3. Review each batch before moving to next
  4. Apply all within 1 week
  5. Launch store with complete, professional content
Time Saved: ~300 hours of manual writing.

​
Scenario 2: SEO Improvement (Existing Store)

Goal: Add meta tags to 10,000 existing products. Workflow:
  1. Create group: “Products without meta tags”
  2. Enrich meta title + meta description for entire group
  3. Sample review (10% of products)
  4. Batch-apply all
  5. Monitor SEO performance over 3 months
Time Saved: ~200 hours of copywriting.

​
Scenario 3: Attribute Backfill

Goal: Add structured attributes (color, material, size) to 5,000 products for better filtering. Workflow:
  1. Enrich color + material + size for all products
  2. Review sample of 10% to assess quality
  3. Batch-apply products that pass quality checks
  4. Review individually any products with issues
  5. Push to platform, enable faceted search
Time Saved: Weeks of manual categorization.

​
Scenario 4: Multi-Language Expansion

Goal: Expand from English to French, German, Spanish (3,000 products). Workflow:
  1. Perfect English content first
  2. Enrich all products for French site
  3. Native French speaker reviews sample
  4. Apply French content, launch French store
  5. Repeat for German and Spanish
Time Saved: Months of translation work + $$$$ in translator fees.

​
Monitoring Bulk Job Performance

​
Dashboard View

The Dashboard shows real-time progress for bulk jobs:
  • Progress bar: “1,247 / 5,000 products enriched (25%)”
  • Status updates: Which products are currently processing
  • Error alerts: Failures as they occur
  • Estimated completion: Time remaining (approximate)

​
Job Detail View

Click into the job to see granular details:
  • Product-by-product status: ✅ Success, ❌ Failed, ⏳ In Progress
  • Attribute-level results: Which attributes succeeded for each product
  • Error messages: Why specific products failed
  • Batch actions: Retry failures, export results
Learn more about job monitoring →

​
Optimizing Bulk Performance

AI providers (Gemini, GPT) can have rate limits during peak usage. Run large jobs during off-peak hours (evenings, weekends) for faster processing.
Complex prompts with many data sources and instructions take longer to process. For bulk operations, use streamlined prompts.
If enriching 10,000 products, consider running 2 jobs of 5,000 each simultaneously (if your plan supports multiple concurrent jobs).
Only enrich attributes you actually need. Don’t generate meta descriptions if your platform doesn’t use them—it wastes time and tokens.

​
What’s Next?

Jobs & Monitoring

Track and troubleshoot bulk jobs

Review Workflow

Learn efficient review strategies

Collections & Groups

Organize products for bulk operations

Attribute Prompts

Optimize prompts for quality at scale
Next: Explore Attribute Prompts to customize how the AI generates content for each attribute type.