AI-Powered Product Attribution for 1,000 Items Across 7 Categories

Discover how we used advanced AI models to automate product attribution for over 1,000 items across seven categories. By combining text-based extraction and image recognition, we transformed a weeks-long manual task into a scalable, accurate, and CMS-ready data pipeline.

They transformed SEO into our most effective acquisition channel, driving substantial year-over-year organic growth and helping us support 1.5 million monthly visitors across our vast 170,000-item catalog.

Client

ePlaneta

Expertise
  • Artificial Intelligence
  • Image Recognition
  • Data Processing
  • Custom Automation
  • eCommerce Optimization
Year

2025

ePlaneta is one of Serbia’s leading online retailers, offering over 170,000 products across 2,000+ categories. With around 1,500,000 monthly visitors, the platform serves a wide range of customer needs. They needed structured, category-specific attributes to power faceted navigation and improve product discoverability, especially where key details (like color, shape, or type) weren’t reliably present in descriptions.

Client words

For the past three years, Granular Group has been ePlaneta’s go-to partner for scaling our digital presence. They transformed SEO into our most effective acquisition channel, driving substantial year-over-year organic growth and helping us support 1.5 million monthly visitors across our vast 170,000-item catalog. Their strategic guidance fueled rapid expansion in key verticals—household appliances, IT hardware, footwear, and apparel—solidifying our leadership in each.

Their greatest impact has been in empowering us with clear, actionable insights. They built a bespoke controlling system that tracks performance at the product, category, and keyword level, surfacing opportunities and flagging issues the moment they arise. This real-time visibility has enabled our teams to optimize campaigns swiftly, fine-tune content, and maintain steady conversion gains across the board.

With their proactive, hands-on collaboration, Granular feels like an extension of our team. They navigate cross-departmental hurdles with ease, continuously unlocking new revenue streams and ensuring ePlaneta is poised for sustainable growth in Serbia’s competitive eCommerce market.

Branimir Kulašević, Chief Ecommerce Officer

Challenges

  • Manual Product Attribution at Scale

    Assigning attributes to 1,000+ products manually would require opening each product, reviewing details, and saving updates one by one, a process too slow and unsustainable.

  • Inconsistent or Missing Product Information

    Many key attributes (shape, color, type) were not present in product descriptions, making text-only attribution insufficient.

  • Category-Specific Attribute Requirements

    Seven different product categories required unique sets of attributes and filters, adding complexity to standardization.

  • No Unified Data Structure

    Existing product data (name, description, specs, images) was dispersed, requiring consolidation before AI processing.

Solution

  • Attribute Framework Development

    Created a complete list of required attributes and filters for each of the seven categories.

  • Consolidated Product Data Tables

    Merged product names, descriptions, specifications, and images into unified data tables for clean processing.

  • Text-Based AI Extraction

    Used AI models to analyze product content and assign attributes based on available textual information.

  • AI Image Recognition

    Applied image recognition models to detect visual attributes: shape, color, style, and other features missing from descriptions.

  • Unified CMS-Ready Output

    Generated a single structured database for all products, ready for direct import into the CMS with no manual editing needed.

Results

  • 1,000+ products attributed Fully enriched with structured attributes across seven categories.
  • 29 new filters created Category-specific filters built to improve product discoverability.
  • 100% automated processing Replaced weeks of manual work with an AI-driven workflow.
  • Dual-source accuracy boost AI text + image analysis ensured coverage of attributes missing from product descriptions.
00 Project Background 01 Defining the Attribute Structure 02 Data Consolidation 03 AI-Driven Text Extraction 04 Image Recognition Layer 05 Attribute Synthesis & Verification 06 Outcome
 
Project Background

The goal was to attribute more than 1,000 products across seven categories, each requiring unique filters and attribute sets. Manual attribution was not an option due to the scale, inconsistency of product information, and the need for structured data suitable for import into the CMS.

Defining the Attribute Structure

We began by creating a unified attribute structure for all seven categories. This included identifying the complete list of filters and attributes needed for each product type and ensuring consistency.

Data Consolidation

Before applying any AI model, we compiled all existing product data into clean, structured tables. This included product names, descriptions, specifications, and images. Consolidating these elements provided a reliable foundation for automated processing.

AI-Driven Text Extraction

Using advanced AI models, we analyzed product descriptions and other textual content to extract attributes such as dimensions, material and other category-specific details. This step addressed products with well-documented descriptions while highlighting missing information.

Image Recognition Layer

To fill the gaps left by incomplete or inconsistent textual data, we introduced an additional layer of AI: image-based recognition. Models were used to detect visual attributes such as shape, color, style elements, and other characteristics not evident from text alone. This dual-source approach significantly improved attribute accuracy.

A few examples:

AI image-based recognition. Attributes: material, color, shape.

 

AI image-based recognition. Attributes: color, style.

 

AI image-based recognition. Attributes: type, color

Attribute Synthesis & Verification

Outputs from text and image models were merged, cleaned, and standardized through rule-based checks. This ensured that each product received a complete and accurate set of attributes aligned with the predefined category structure.

Outcome

By combining category-specific schema design, AI-driven text extraction, image recognition, and structured data engineering, we transformed a highly manual workflow into a scalable attribution system capable of enriching thousands of products with consistent, accurate, and CMS-ready attributes.

Related case studies

More work with this client More SEO case studies

Explore all case studies