Advanced Features That Make Product Categorization Easier
Basic matching — uploading a CSV, getting categories back — works well for straightforward cases. But when your catalog grows or your products get more specialized, you run into edge cases that need more powerful tools.
Here's what Taxonomy Matcher offers beyond basic matching, and when each feature actually helps.
Hierarchical Categorization
Standard matching asks the AI to choose from thousands of categories at once. This works, but it's like asking someone to find a specific book by scanning every shelf in a library simultaneously.
Hierarchical categorization works differently. It narrows down level by level:
Product: "Wireless Noise-Cancelling Headphones"
Level 1: Electronics (from 10 main categories)
Level 2: Audio (from 15 Electronics subcategories)
Level 3: Headphones (from 8 Audio subcategories)
Level 4: Over-Ear Headphones (from 4 Headphones types)
Result: Electronics > Audio > Headphones > Over-Ear Headphones
Each decision has fewer options, which means each decision is more accurate. The result is significantly better matches, especially for products that could plausibly belong to multiple top-level categories.
When to use it: Any time accuracy matters more than speed — particularly for products with ambiguous names or short descriptions.
AI Image Generation
Taxonomy Matcher integrates Google's Gemini AI to generate product images from text descriptions. You get four style options:
- Cutout — clean white background, suitable for marketplace listings (Amazon, eBay, Shopify)
- Artistic — creative composition with vibrant colors, designed for social media and ads
- Mood — lifestyle context showing the product in a realistic usage scenario
- Professional — studio-quality aesthetic for catalogs and B2B materials
This isn't a replacement for professional product photography in every case, but it's useful when you need visuals quickly — for new products, catalog placeholders, or A/B testing different presentation styles before investing in a photo shoot.
Each image generation costs credits, same as matching.
Batch Operations
When you need to categorize products regularly (new inventory, seasonal updates, supplier feeds), doing it manually each time creates unnecessary work.
Batch operations let you:
- Process multiple product lists in sequence or parallel
- Apply consistent settings across all batches (same taxonomy, same options)
- Monitor progress for each list independently
- Set up recurring workflows for regular catalog updates
The main value is eliminating repetitive setup. Instead of uploading, configuring, and running the matcher each time, you define the workflow once and reuse it.
Advanced Filtering and Validation
Raw product data is rarely clean. Missing descriptions, inconsistent formatting, duplicate entries — these all reduce matching accuracy and waste credits.
The filtering system lets you:
- Validate data before processing — check for required fields, minimum description length, proper formatting
- Preview filtered results — see exactly how many products pass your criteria and estimate credit costs
- Apply business rules — filter by category, price range, brand, or custom conditions
- Skip problematic entries — automatically exclude products that don't meet quality thresholds
This matters most when processing supplier feeds, which often contain incomplete or inconsistent data. Running validation first means you only spend credits on products that have enough information for accurate matching.
Getting Started
If you're already using basic matching, the easiest way to try these features is:
- Enable hierarchical categorization on your next batch — it's a single toggle, and you can compare results against standard matching.
- Try image generation on 10-20 products to evaluate the quality for your specific product types.
- Set up a simple filter to catch products with missing descriptions before processing them.
Each feature works independently, so you can adopt them gradually based on what your workflow actually needs.