Intergamma

Intergamma is a hardware store chain with almost 400 stores in the Netherlands and Belgium. They have been the market leader in hardware products for years. Intergamma wanted to provide their products with related products, comparable products, product families and accessory products. Establishing these product relationships manually would be an extremely time-consuming process, however Squadra Machine Learning Company has made it possible to partially automate this process.

The request

Intergamma approached Squadra MLC because they wanted to facilitate product relationships to enrich the customer experience. In many cases, all product attributes of related products are similar except for one attribute, which is where the variance is based upon. For example, for a can of paint it could be that the color, gloss level or size of the can varies. To utilize these product variants in their sales channel, Intergamma decided to collaborate with Squadra.

Intergamma had already investigated whether significant product relationships could be withdrawn from the available product data. It appeared that every time a feature was added to the product page, this feature needed to be turned off again due to multiple issues. Intergamma learned from these issues in two ways: firstly, it appeared that the product data was of insufficient consistent quality. Some fields were not correctly filled in or not filled in at all, which led to weak relations between products. Secondly, it appeared that there was a need for sufficient business knowledge to find out which attributes are needed to receive clear and effective product variations. This is a challenge due to the fact that effective product variations strongly depend on the product range.

This is where Squadra was asked to use their machine learning competencies in order to find relevant product variants, which could be validated manually. This way, high quality product relationships could be generated without putting much work in.

The solution

The project that Squadra subsequently worked on, consisted of two parts. For the first part, Intergamma had delivered a huge amount of product data (Excel files) to Squadra. By using this data and the Powerrelate.ai software, Squadra managed to capture product relationships. These relationships were validated by Intergamma employees, who obviously have sufficient business knowledge to do so. This proof-of-concept implementation resulted in improved product relationships for product detail-pages in 4 different product ranges. The effectivity was measured using an A/B test and the proof-of-concept implementation led to an uplift in sales with 4.6% which was “just a magnificent score” according to Anouk Renaud, Product Owner at Intergamma.

After finishing the A/B test, Intergamma decided to implement PowerRelate for the entire product range. Meanwhile, there are now 19,000 product relations live for Gamma in the Netherlands, 16,000 for Gamma in Belgium and 15,000 for Karwei. It is estimated that these product relations account for approximately 75% of all possible product relations.

The challenge

PowerRelate has a user interface in which an Intergamma user can view and approve the product relation types and related products found per product. The algorithm models the relations on the basis of smart Natural Language Processing (NLP), where the distinguishing properties of the product (such as color, width and length) are automatically derived from the product description. Fuzzy matching is also used to detect the similarities between values ​​such as grey/brown and brown/grey. If available, the products are provided with product images so that assessment can easily take place.

The challenge

Before Intergamma collaborated with Squadra, they have worked with Google’s recommendation algorithm. This algorithm offers product recommendations which are completely automized. However, this algorithm did not offer the desired quality in product recommendations for Intergamma, due to the fact that products in the hardware segment are often too specific to offer general recommendations to. For example, when you buy a t-shirt online in a clothing store, it would not be a huge issue if the recommended pants are not to your taste. But when you want to buy a garden shed and a recommended floor does not fit in the garden shed that you want, you may face a more difficult situation.

At Intergamma and other hardware stores the product relationships are stricter. This results in a desire for people with specific business knowledge to validate the product relations output. Compared to the test that Intergamma conducted with Google, PowerRelate is a positive surprise.

The collaboration

The collaboration was outstandingly pleasant according to Intergamma. Anouk states: “Squadra has actively been thinking along with us. They frequently suggested new solutions and asked new questions to gain knowledge about our targets and desires. They were proactive and took the lead in setting up the platform for us. All and all, I found it a successful and great alignment”.

“Squadra has actively been thinking along with us. They frequently suggested new solutions and asked new questions to gain knowledge about our targets and desires”

Nomination most innovative project

Intergamma has been nominated for the CIO Magazine Innovation Award, in the category ‘Most Innovative Project / B2B program, for the above-mentioned project.

De PowerRelate software is offered as a Software as a Service (SaaS) that can be easily integrated with existing PIM solutions. For more information or a demo, please contact us at info@powerrelate.ai or +31 (0) 85 065 3780.