Before You Start
You'll need your product data in CSV format. The required columns are:
id— a unique identifier for each producttitle— the product namedescription— a text description of the product
Use semicolons (;) as the delimiter. Here's a minimal example:
id;title;description
101;Organic Cotton T-Shirt;Soft, breathable crew-neck t-shirt made from 100% organic cotton. Available in black, white, and navy.
102;Wireless Noise-Cancelling Headphones;Over-ear headphones with active noise cancellation, Bluetooth 5.0, and 30-hour battery life.
103;Cast Iron Skillet 12 inch;Pre-seasoned cast iron skillet suitable for stovetop, oven, and campfire cooking.
Better descriptions lead to better matches. A title like "Widget X-500" with no description gives the AI very little to work with. Adding details like "stainless steel kitchen timer with magnetic back and loud alarm" makes a significant difference.
For more detail on formatting, see the CSV formatting guide.
Step 1: Upload Your Data
Go to the Matcher page and choose how to input your data:
- Paste CSV data directly into the text area — useful for quick tests or small batches
- Upload a CSV file — better for larger datasets
The system will parse your data and show you a preview of the products it found.
Step 2: Choose a Taxonomy
By default, Taxonomy Matcher uses the Google Product Taxonomy — a hierarchy of 6,000+ categories maintained by Google. This is the right choice if you're selling through Google Shopping, or if you need a general-purpose category system.
If you sell on a specific marketplace (Amazon, Otto, etc.) or use your own internal categories, you can upload a custom taxonomy as a .txt file. Each line should be one category path, with levels separated by >:
Electronics > Computers > Laptops
Electronics > Computers > Desktops
Electronics > Audio > Headphones
See the custom taxonomy guide for formatting details.
Step 3: Configure Options
Before starting, you can optionally enable:
- Hierarchical categorization — narrows down categories level by level for better accuracy (recommended for large taxonomies)
- AI image generation — generates product images from descriptions using Gemini AI
- Title and description optimization — suggests improved product text for SEO
Each option consumes credits.
Step 4: Run the Matcher
Click Start Matching. The AI processes each product, and you'll see real-time progress as categories are assigned.
Processing time depends on the number of products and options enabled. A batch of 100 products with standard matching typically takes 1-2 minutes.
Step 5: Review and Export
Once processing completes, you'll see results in a table showing each product alongside its matched category and confidence score.
From here you can:
- Download as CSV — for importing into your feed management tool, Shopify, WooCommerce, etc.
- Download as JSON or XML — for programmatic integration
- View on your Dashboard — access previous results anytime under your match history
Results are saved to your account, so you can come back to them later without re-running the match.
Tips for Better Results
- Write descriptive titles. "Blue Widget" is hard to categorize. "Bluetooth Wireless Speaker, Portable, Waterproof, 10W" is easy.
- Include key attributes in descriptions. Material, intended use, target audience, and product type all help the AI make accurate decisions.
- Start with a small test. Try 10-20 products first to verify accuracy before processing your full catalog.
- Use hierarchical matching for taxonomies with many categories — it consistently produces more specific results.