ETIM / FEST

FEST represents the national associations of wholesalers involved in the distribution of plumbing and heating products throughout Europe.

Challenge  

To reduce costs and facilitate the transfer of product information from manufacturers to wholesalers, and eventually to retailers and consumers, there was a significant need for standardizing product data within the plumbing and heating sector, more broadly recognized as the Heating, Ventilation, and Air Conditioning (HVAC) industry. The absence of a standard results in data duplication and unnecessary modifications. Additionally, the need to satisfy various wholesale and retail requirements leads to costly inaccuracies in product data, complicating expansion into new markets and regions.

ETIM has been established as the definitive classification model for technical products and has been adopted as the standard for the HVAC sector. Consequently, manufacturers in this field must adapt their proprietary metadata definitions (such as specific product classes, feature labels, feature values, and measurement units) to align with the ETIM standard.

FEST operates in eighteen different countries, with just over half participating in ETIM for the HVAC industry. Therefore, there is a pressing need to encourage the remaining countries and their manufacturers to convert their proprietary product data to the ETIM standard. This conversion can be quite labor-intensive, requiring classification of products according to ETIM definitions and aligning specific feature names/values/units of measure with the ETIM standard.

Solution  

Utilizing advanced Artificial Intelligence (AI) techniques can significantly simplify these conversion processes, reducing the time necessary to adapt a manufacturer’s metadata definitions to the ETIM standard. After a brief evaluation of the available data, Squadra Machine Learning Company recommended the use of the Powerconvert.ai  tool for (1) product classification, (2) matching feature names, and (3) converting feature values.

ETIM Class Prediction  

The ETIM class prediction algorithm assigns an ETIM class based on the manufacturer’s name and product description. This service is available in two formats: a bulk classification service where users upload an Excel sheet with multiple products and receive an output with the predicted ETIM classes included, and a ‘Google-like’ web interface allowing users to input one manufacturer and product description to obtain the predicted ETIM class.

This algorithm requires training using already classified ETIM product data. As such, it functions with product descriptions in the language it was trained in. This can pose a challenge, especially since the focus is on manufacturers from Southern and Eastern Europe where data pools in these languages are limited.

Recent studies indicate a viable workaround: translating Finnish, Dutch, or German training sets using Google Translate and developing a new model based on the translated dataset. Research has demonstrated that this approach achieves nearly the same level of accuracy as models trained on native language datasets (with only a 1-2% accuracy difference).

Thus, leveraging the initial dataset alongside Google Translate, Squadra Machine Learning Company successfully created ETIM Classification algorithms for every European language.

Matching Manufacturer Feature Names to ETIM Feature Names  

The ETIM feature matching algorithm provides an efficient method to align manufacturers’ non-ETIM feature labels with the ETIM format.

This service is integrated into the Powerconvert.ai  web application, where manufacturers can upload a standardized dataset in Excel format, and the algorithm predicts the correspondences. Additionally, users can review and adjust these mappings individually, allowing for ongoing improvement as the algorithm learns from the results.

Conversion of Feature Values / Units of Measure  

After identifying and validating feature name correspondences, feature values are adjusted to the appropriate formats. For instance, if a manufacturer uses “inox steel” instead of “stainless steel,” the algorithm suggests corresponding feature values when necessary. It can also detect when units need conversion, such as changing from dm³ to m³.

Result  

With the assistance of Powerconvert.ai  , FEST is now capable of converting manufacturers’ metadata definitions to the ETIM standard. Utilizing the initial dataset and Google Translate has enabled Squadra MLC to support FEST in navigating language differences, enhancing their international online presence. By implementing this software, the organization has notably saved both time and resources previously spent on manual product data conversion.