Improving product relationships with Machine Learning for Intergamma.
Sector
(DIY) Retailer
Products
50,000+
Employees
10,000+
Intergamma, the organization behind DIY chains Gamma and Karwei, has been a market leader in the Netherlands and Belgium for years. With nearly 400 stores, Intergamma continuously seeks ways to enhance the online shopping experience for its customers. One key challenge was improving product relationships on its websites to help customers find relevant variants, alternatives, and accessories efficiently. However, manually establishing these relationships was time-consuming. To tackle this, Intergamma partnered with Squadra Machine Learning Company to explore automation possibilities.
Background
Intergamma, established in 1968, is a prominent DIY retail organization operating throughout the Benelux region. Through its well-known brands, Gamma and Karwei, Intergamma provides a vast assortment of home improvement, gardening, and construction products. Catering to both professional contractors and enthusiastic DIYers, Intergamma focuses on offering a wide product range, competitive pricing, and a strong emphasis on customer service. With a significant presence in the Netherlands and Belgium, Intergamma has solidified its position as a leading player in the DIY retail market.
The challenge.
-
Inefficiencies in manual product matching
Intergamma aimed to create product relationships for variants, alternatives, and accessories. However, manually mapping these relationships was a time-intensive process, making it difficult to scale and maintain accuracy across the entire assortment.
-
Inconsistent product data
Efforts to automate relationships using business rules were unsuccessful due to inconsistent product data. Missing or incorrect attributes led to inaccurate product links, requiring constant adjustments and manual corrections.
-
Need for high-precision recommendations
DIY product recommendations require a higher level of accuracy compared to other industries. Unlike fashion, where a mismatched suggestion is minor, incorrect recommendations in DIY (e.g., an incompatible underfloor for a garden shed) can cause significant issues. Previous attempts with Google’s recommendation algorithm failed to meet the necessary precision.
Our solution.
-
Proof-of-concept with PowerRelate.ai
Intergamma provided Squadra with an initial dataset to test PowerRelate.ai. The AI-driven approach successfully identified product relationships, which were manually validated by Intergamma before being imported into the website.
-
AI-powered attribute matching
Using Natural Language Processing (NLP) and fuzzy matching, PowerRelate.ai automatically detected distinguishing attributes, such as color or dimensions, improving the accuracy of product relationships. Product images were also integrated to assist with validation.
-
Scalable implementation across the assortment
Following a successful A/B test that resulted in a 4.6% transaction uplift, Intergamma expanded PowerRelate.ai across its full product range. Currently, tens of thousands of product relationships are live across Gamma and Karwei’s platforms.
The result
PowerSuite has empowered Intergamma to...
Improve product relationships
Automated variant matching reduced manual effort while increasing the accuracy of product recommendations.
Enhance customer experience
Better product links made it easier for customers to discover relevant products, improving the shopping journey.
