How to Identify Product Attributes with AI in conbase.ai

Published on July 08, 2025

By Daniel Manco

Introduction

  • The Goal:
    • By the end of this guide, you'll be able to automatically pull accurate, structured product attributes for thousands of product records in minutes, streamlining your catalog management workflow.
  • The Problem:
    • Manually identifying attributes for each product is slow, inconsistent, and hard to scale. Teams spend hours copying specs and miss key details, which causes listing errors and lost sales.
  • The Solution:
    • Conbase.ai's AI-powered pipeline lets you identify attributes at scale using custom prompts that understand your e-commerce context. Instead of one-size-fits-all tools, you build a workflow that matches your exact catalog rules.
  • Prerequisites:
    • A Conbase.ai account
    • A CSV containing your products with columns like title, description, category, and spec_sheet
    • Basic knowledge of the attributes your store tracks (size, color, material, power, etc.)

Are you here for the first time ? Then watch a full conbase.ai workflow in action

This video walks you through a complete conbase.ai workflow, from input to final output, so you can see exactly how the platform works and how to get the most out of its features.

Step 1: Set Up Your Project and Import Data

Create a New Project

Log in to your Conbase.ai dashboard and click “New Project.” Name it "Identify Product Attributes - 2025-04" to keep things tidy.

Upload Your Products CSV

Upload the CSV with your product data. Conbase.ai supports up to 10,000 rows, so you can test and scale in the same place.

Data Prep Tips:

  • Add clear column headers
  • Remove odd characters that might confuse parsing
  • Include every field the AI needs (for example, spec_sheet if you have one)
  • Clean up obvious typos before upload

Step 2: Build Your AI Identification Pipeline

Add an Action Step

You can start from a template or make one from scratch.

Option A: Use a Template
Search the library for “Product Attribute Extractor.” It has a base prompt ready to tweak.

Option B: Create from Scratch
Click “Create New,” name the step Identify Attributes, and add it to the pipeline. We’ll use this option below.

Set Up Your Custom Prompt: The 3-Part Recipe

Part 1: Instruction

You are an e-commerce data specialist. Identify key attributes for each product. The output will contain columns with keys: size, color, material, key_features. If a value is unknown, return an empty "". Do not invent data.

Part 2: Context

Product Title: {title}
Description: {description}
Category: {category}
Spec Sheet: {spec_sheet}

Part 3: Output Columns

  • size → Output Example: 13-inch,
  • color → Output Example: silver,
  • material → Output Example: aluminum,
  • key_features → Output Example: retina display; M3 chip; backlit keyboard

Save the step when done.

Step 3: Configure and Run the Pipeline

Fine-Tune AI Settings

Model: "GPT-4o" works well for structured extraction.

Processing Mode: "Instant" for testing and for fast execution, "Scheduled" for large batches and 50% less API costs

Integration Mode:

  • Fill if empty (safe test)
  • Overwrite (if you know existing data is wrong)
  • Append (keeps old values and writes to new columns)

Select Rows and Execute

  1. Select 10 rows first and click “Run Pipeline.”
  2. Review results for accuracy.
  3. Tweak the prompt if issues appear.
  4. When happy, select all rows and run again.

Step 4: Review and Export Your Results

Preview the Output

  • Open random cells to confirm JSON structure
  • Check that unknown values are marked as "unknown," not empty
  • Look for wild guesses; refine the prompt if any appear

Export to CSV

Click the export icon. The download includes original columns plus size, color, material, and key_features.

Next Steps:

  • Import into your PIM or e-commerce platform
  • Track listing quality scores to measure impact
  • Document the prompt for future pipeline runs

Conclusion and Next Steps

You’ve automated attribute identification for your entire catalog. What once took days now happens in minutes, and your listings stay consistent across channels.

  • Time saved: compare manual vs. AI processing
  • Quality boost: fewer listing errors
  • Scalability: ready for new product drops without extra headcount

Book Your Free Conbase.ai Demo and see how else you can streamline content operations.

Ⓒ Copyright. All rights reserved by conbase.ai