01
Product data rarely arrives perfectly organized
A distributor might receive a description such as: “3/4 inch stainless steel ball valve, 1000 PSI” Someone still has to decide whether it belongs under: Industrial Supplies → Valves → Ball Valves Another supplier may describe essentially the same product completely differently. For businesses managing thousands of SKUs, these small decisions become a large data-cleanup problem.
02
Teach AI from your existing catalog
Your existing product catalog already contains valuable examples. For example:
| Product Description | Category |
|---|---|
| 18V cordless impact driver | Power Tools |
| Nitrile disposable gloves | Safety Equipment |
| Stainless steel ball valve | Industrial Valves |
Upload historical product data to StayCharted AMT and train a model on the taxonomy your business already uses. When new products arrive, the model predicts the appropriate category.
03
Use product photos when appearance defines the category
With Vision, StayCharted AMT can learn categories from supplier or product photos as a separate image workflow. Start with photos your team has already labeled, uploaded in category folders or supplied as picture links.
For example, a photo of a backpack can be categorized as Backpacks even when a supplier’s text description is inconsistent. Use an AI Image Classifier Model, or evaluate a Dedicated AI Image Model on Business with Vision.
Text and image models are separate approaches; this does not combine description and photo inputs into one model. Photo categorization does not establish an exact SKU match, read part numbers, or replace your field-mapping rules. Review predictions before applying your catalog labels.
04
Use your taxonomy—not someone else's
This matters because product categories are rarely universal. One organization may categorize a laptop as: Computers → Business Laptops Another may use: IT Hardware → End User Devices Another might organize products by internal purchasing group rather than by product type. StayCharted AMT learns from your examples instead of forcing you into a predefined taxonomy.
05
Useful for more than one category
The same approach can be used for:
- Product family
- Product type
- Procurement group
- Brand classification
- Merchandising category
- Internal department
- Tax category
- Shipping classification
You decide what the output should represent.
06
Start manually, automate later
Smaller catalogs can be processed directly through uploaded CSV or XLSX files. Larger workflows can call the StayCharted AMT Prediction API whenever a new product is created or imported.
- Supplier Feed
- Product Management System
- StayCharted AMT API
- Predicted Category
- Product Record Updated
The product management system remains the system of record. StayCharted AMT simply provides the business-specific prediction.