Bulk Enrichment for Large Catalogs

If you're managing a catalog with 2,000 to 500,000+ SKUs, enriching everything at once is impractical. This article covers strategies to work smarter: batch processing, phased rollouts, and monitoring progress without overwhelming the system.

The Core Rule: Work in Batches, Not All at Once

Trying to enrich your entire 50,000-SKU catalog in one go is a recipe for slow processing, timeouts, and frustration. Instead:

Work in logical batches — by category, vendor, product type, or priority — and enrich 500–5,000 products at a time.

This approach gives you:

• Faster feedback and ability to test prompts early.
• Easier troubleshooting if something goes wrong.
• Incremental wins you can review and refine.
• Manageable processing times.

Strategy 1: Batch by Category

Your most straightforward approach: enrich one product category at a time.

Steps:

1. Go to Workspace → Views and create a new View (e.g., "Sofas - Unenriched").
2. Filter by Category = "Sofas" AND Enrichment Status = "Empty" (or however you track unenriched products).
3. Select all products in this View.
4. Run Generate → Generate All Attributes or Generate Empty Attributes Only (depending on your needs).
5. Review the output for a representative sample (10–20 products).
6. Make any prompt refinements based on what you learned.
7. Move to the next category and repeat.

This keeps your work organized and lets you test and refine prompts before moving to the next batch.

[SCREENSHOT: Workspace → Views showing a filtered view like "Sofas - Unenriched" with product count and category filters visible]

Strategy 2: Batch by Vendor

If you work with multiple vendors, batch by vendor instead:

1. Create a View for each vendor: "Vendor A - Electronics - Unenriched."
2. Enrich by vendor, 1,000–3,000 products at a time.
3. Because vendor data often has consistent formatting, you can test prompts on one vendor, refine, then scale to others.

This is especially useful if different vendors provide different quality sources (one has detailed specs, another only has images).

Strategy 3: Batch by Priority

If categories are too broad, batch by highest-value products first:

1. Identify which products matter most: bestsellers, high-margin items, products that appear on your website.
2. Create a View: "Top 500 - Unenriched."
3. Enrich and refine on these high-impact products first.
4. Once you're happy with quality, move to mid-tier and lower-value products.

This ensures your best-selling items get the most attention and polish.

Recommended Batch Sizes and Timeline

Here's a rough guide based on catalog size:

These are estimates; actual time depends on complexity, number of attributes, and your system. Always test a batch first before committing to a full schedule.

The Phased Rollout Approach

For very large catalogs (50,000+), don't plan to finish enrichment in a day. Instead, spread it across weeks:

Week 1: Enrich your top 20% (highest-value, best-selling products)

• Refine prompts aggressively on this batch.
• Get stakeholder feedback.
• Identify what "good" looks like.

Week 2: Enrich the next 30% (mid-tier products)

• Use refined prompts from Week 1.
• Test any new sources or attributes added.
• Gradually improve automation.

Week 3+: Enrich remaining 50%

• Use proven, tested prompts.
• Run larger batches (less frequent review needed).
• Focus on consistency and coverage.

This phased approach lets you refine before you scale, reducing the risk of enriching 100,000 products with a flawed prompt.

Using "Generate Empty Attributes Only" for Incremental Enrichment

As you add new sources (images, specs, reviews), you'll want to fill in gaps without re-enriching what's already done.

Scenario: You've enriched 50,000 products for description, category, and brand. Now you're adding manufacturer spec sheets and want to enrich dimensions and materials on the same 50,000 products.

Solution:

1. Add spec sheets as a source.
2. Configure "Dimensions" and "Materials" attributes with new prompts.
3. Select all 50,000 products.
4. Run Generate → Generate Empty Attributes Only.

Only the new attributes (dimensions, materials) will generate. Your existing descriptions, categories, and brands stay exactly as they are. This saves processing time and respects your previous work.

[SCREENSHOT: Generate dropdown with "Generate Empty Attributes Only" highlighted, showing a batch of products]

Monitoring Progress

For large catalogs, track your progress systematically:

Use the Table to Track Completion

In your View, add columns for each enriched attribute. Scan visually to see:

• Which attributes are filled vs. empty.
• Which categories or batches are done.
• Where gaps remain.

Create a Tracking View

Build a View that shows you "Unenriched Products" by filtering:

• Enrichment Status = Empty OR
• Critical attributes (like description, category) = blank

Run this View periodically to see your remaining work.

Log Your Batches

Keep a simple log (spreadsheet or notes) of what you've enriched:

• Batch 1: Sofas (2,000 products) — completed, quality good
• Batch 2: Chairs (3,000 products) — completed, category tags needed refinement
• Batch 3: Tables (2,500 products) — in progress

This helps you remember what you've tested, what worked, and what still needs attention.

Pre-Enrichment Setup for Large Catalogs

Before you start enriching thousands of products, do this once:

Import All Your Sources First (Article 2.6)

Don't enrich products, discover sources, then enrich again. Get all your sources loaded:

• Manufacturer specs and datasheets
• Images (organized and properly sized)
• Vendor descriptions and reviews
• Any custom data you have

Then enrich once with everything available. This is far more efficient than multiple rounds.

Configure All Attributes Before Enriching

Go to Workspace → Attributes and set up every attribute you'll need:

• Decide on prompts, acceptable values, and data context.
• Enable "Use AI" on all attributes you want enriched.
• Test prompts on a small batch first.

Once you're confident in your setup, run full batches. Don't start enriching, then pause halfway through to reconfigure attributes — it wastes time and creates inconsistency.

Use Categories to Organize Your Work

Before you start:

1. Make sure all products are correctly categorized (at least at a high level).
2. Create one View per category or vendor.
3. Label each View clearly so you know which ones are done.

This makes batching and tracking much easier.

Common Issues for Large Catalog Enrichment

Processing Timeouts

Problem: You selected 10,000 products and the enrichment stalled or timed out.

Solution: Reduce batch size. Try 5,000 products instead. If 5,000 still times out, go down to 3,000. Stability matters more than speed.

Slow Processing on Large Batches

Problem: A batch of 8,000 products is taking 4+ hours.

Solution: Close other browser tabs and applications to free up resources. If that doesn't help, you might have too many attributes or overly complex prompts. Consider:

• Temporarily disabling "Use AI" on less critical attributes.
• Simplifying prompts (shorten them by 20–30%).
• Running smaller batches in parallel sessions (if your system supports it).

Inconsistent Output Across Batches

Problem: Sofas enriched in Week 1 have one tone/style, and sofas enriched in Week 2 look different.

Solution: Your prompt changed, or your sources were inconsistent. Before starting Batch 2, review Batch 1 output and document the style/tone you achieved. Update your prompt to say "match the style and depth of previously enriched products." If sources changed, note that in your prompt too.

Some Products Didn't Enrich

Problem: You ran Generate on 5,000 products, but 200 have empty description fields.

Solution: Those 200 likely missing required source data (e.g., no product name, no vendor description). Before re-running:

1. Identify what data is missing on those 200 products.
2. Fill in the missing required fields.
3. Run Generate → Generate Empty Attributes Only on just those 200 products.

Quality Drops on Later Batches

Problem: Your first 10,000 products are rich and detailed. Your next 10,000 look generic.

Solution: Your source data for later batches is thinner, or your prompt assumptions are breaking. Check:

• Are later batches from different vendors with fewer sources?
• Did you change which sources the prompt should use?
• Is there inconsistency in how products are structured?

Refine your prompt to be more adaptive. Test on a few products from the later batch, adjust, then re-run.

When to Use Each Generate Mode at Scale

• Generate All Attributes: You've made a significant prompt change and want to refresh everything. Best used after testing on a small batch.
• Generate Empty Attributes Only: You're adding new attributes or sources incrementally. Use this most of the time to avoid unnecessarily re-processing.

For large catalogs, "Generate Empty Only" is your friend — it respects your previous work and only fills the gaps.

Next Steps: From Enrichment to Export

Once you've enriched your catalog, you have two paths:

Exporting Enriched Data Back to Your Platform

If you imported from Shopify, WooCommerce, or another platform, you can sync enriched data back to keep your source system in sync. Head to How Integrations Work in Merchkit in Section 4 to learn how.

Optimizing for Sales Channels

Your enriched catalog is now ready to be optimized for specific channels (Amazon, Wayfair, your own site). Each channel has different requirements for descriptions, images, and metadata. Go to Section 5 to start generating channel-optimized content.

Both paths are available. Choose based on your immediate priorities.