Amazon shops use product categories and a taxonomy structure to make it easier for customers to find what they're looking for in their vast catalog in as little clicks as possible. Having your products categorized to the appropriate categories is best way to ensure people can find your products and makes sure your products aren't suppressed or removed from the marketplace. Having your products incorrectly categorized can even violate Amazon's selling policies. If amazon deems your product to be in the wrong category they can even change it.
Getting your categorization correct is key to selling on Amazon and our ai saas product Pumice.ai can categorize products for you with a few clicks.
Amazon product taxonomy is a hierarchical structure of categories and subcategories that Amazon uses to map products on their marketplace. It contains over 10,000 categories and goes multiple levels deep. It's similar in depth to that of Google Product Taxonomy but slightly larger and has a few top level categories that are different.
These categories are also tied to attributes that Amazon uses to further understand the products in a given category. To correctly apply these attributes to your products you need correct categorization. Search results on Amazon use both these fields to ensure the products fit the target audience of a search.
You should be as granular as possible with the categories that apply for your product. Choosing categories that are not as granular when there are better options further down the product taxonomy will make it harder for customers to find what they're looking for. As potential customers look in sub categories that are deeper and deeper its easier to understand their shopping intent and your products should match that. For instance if you put all of your shoes in the "Clothing, Shoes & Jewelry > Men's Shoes" category instead of going further down and putting shoes in " Clothing, Shoes & Jewelry > Men's Shoes > Men's Golf Shoes" buyers that are only looking for golf shoes won't be able to find your product. There's too many options to go through with the first category if they already know what they're looking for.
This isn't as big of an issue for an ecommerce store that sells their own specific items as there taxonomy isn't as large and they have control over the number of products in each category. For a marketplace with millions of sellers selling millions of products you need to make sure your products are as easy to find as possible.
Our product data automation platform Pumice.ai has a pre-built Amazon product taxonomy categorization model that uses ai to categorize products. The model knows each amazon category and understands the relationship between product data and the relevant categories. This taxonomy is one of the many pre-built taxonomies we support. We can even fine-tune it to your specific product data to better understand the relationship between your products and the right category. No more manual product categorization or weird keyword matching systems.
Let's look at how we can quickly categorize millions of amazon products to the amazon category taxonomy with ai.
Pumice offers two ways to interact with the categorization model, the dashboard or the API. The Dashboard takes in CSV files of products and the API uses a JSON format for the product data. You'll either want to export your product data from the catalog system or connect the API. We have extensive API documentation and CSV file examples in our dashboard.
Pumice.ai categorization models require just a title and descriptions but you can provide optional fields such as:
You can even use other Pumice.ai models to extract this data from your existing product data! Once you've got your product records ready to go we can start categorizing.
.png)
As mentioned before we have a pre-built Amazon product taxonomy model available. If we've added the model to your account you can find it in the dropdown menu for "Select product taxonomy" under "Dynamic categorization", or you can provide your "model_id" to the API we provide you when you're onboarded.
Once you hit run or kick off your API call you can sit back and let the ai models do their job. If you're using the batch API it will return a task_id that you can use to check when your run will be complete. Your data will be returned with the best fit category and a confidence score. You can use this confidence score to set up thresholds for deciding when manual review is needed. You should save any data that is manually reviewed as incorrect as we can fine-tune the model again to improve those categories.
Our model is trained on millions of amazon products that are correctly categorized and ranking in Amazon. We focus on the main category for each product, so our model knows the best fit for each product.
Once you've ran your products through Pumice you can upload your data to Amazon. If you still need product attributes we offer an attribute generation model that you can use to complete your product data. We usually recommend using some sort of database or PIM to manage your products and use the amazon store as an external selling channel.
Our saas product Pumice.ai uses SOTA ai models to help you automate product data and PIM related tasks. Never spend another minute manually categorizing products, and focus on real business moving tasks. Contact us today to get started with our categorization models. We'd love to show you what we've built.
Our Amazon product taxonomy services include:
Some of the other popular endpoints we offer in Pumice that are used by Amazon sellers and marketplaces.
Our product similarity endpoint allows you to compare the similarity of product records based on image and text. You can use this to understand when new products being onboarded from vendors already exist in your catalog and would be duplicates, but have different naming conventions, semantics, and attributes. Think (Impact drill vs Impact driver) (st. vs street).
This allows you to remove the manual effort needed for product onboarding and deduping the catalog, meaning you can onboard products faster and at a higher volume. A real life example for one of our customers:
newly added wine title from vendor: Walt Pinot Noir Bob's Ranch
existing product in the master SKU catalog: Walt Bob's Ranch Russian River Valley Pinot Noir
Same product from the same vendor, but different naming conventions. We don’t need to add this wine to the master SKU catalog.
We have a number of attribute extraction models that take your existing product data and extract relevant attributes based on the category and the product data. Some categories like apparel and personal care use these attributes to further filter results that better fit what the customers are looking for. We have models for:
These also help search engines better match search results to user input queries.
