How artificial intelligence helps speed up collaborations between retailers and brands

Découvrez les actus Alkemics

How artificial intelligence helps speed up collaborations between retailers and brands

par Pierre Arbelet8 March 20190 commentaires

When supported by high-quality data, artificial intelligence is able to offer pertinent suggestions, leading to valuable productivity gains for both retailers and brands.

Many different types of information contribute to the overall quality of product data, including product category and packaging information, labels, warning symbols, allergy information, storage information, etc. It can seem to retailers that it is becoming “humanly impossible” to verify the accuracy and completeness of product data. This is because there is such a large volume of data to check, and the complexity of the rules that need to be followed presents even further challenges. In addition, in an omnichannel context, this verification process (and any necessary corrections) need to be completed as quickly as possible – almost in real time.

If the task is becoming “humanly impossible”, how can companies accomplish it anyways? Initial approaches to the challenge are based on a heuristic approach. Concretely, this means configuring a series of rules for these verifications. Using tools such as regular expressions, it is possible to accomplish tasks such as pairing the word “wheat” in a product’s composition with a note about presence of gluten, in adherence with regulations. The problem is that these rules can quickly become complex, and adding another language means they all need to be reconfigured.

Detecting and improving the formatting of allergens in the ingredients text field


Machine learning is becoming increasingly powerful 

This is why Alkemics, while still taking a heuristic approach for certain categories of data, is investing a great deal in in Artificial Intelligence (AI), and more specifically, in machine learning. Machine learning algorithms can be considered to be self-learning: they are able to identify “patterns” based on a large quantity of data. This means that in contexts where a heuristic method becomes difficult to maintain over time, machine learning becomes more efficient as more and more verified data becomes available.This means that the pertinence of the recommendations given by machine learning algorithms directly depend on the quality and the representativity of the data provided. Unsurprisingly, the Alkemics platform, with its hundreds of thousands of products, is an ideal environment for this.

Nonetheless, it is still important to note that artificial intelligence should not be thought of as a tool that “knows the answer”, but instead as a solution to increase productivity — to a significant degree. It is not the AI that decides to put a label on a product, the user still manages and remains responsible for the data. This being said, the suggestions provided by these machine learning algorithms make it possible to accomplish a task that is seemingly becoming “humanly impossible.”

With this tool, retailers are able to browse their product catalogue by manufacturer, brand, category, lifecycle, etc. to analyze data quality. An overall data quality score is calculated based on the descriptions, regulatory data or even media required by the retailer. For product lists, as well as on each product page, AI indicates whether the data is satisfactory, incomplete, or if a correction is required. This is where productivity gains begin, thanks to a clear view of data that needs to be entered or corrected, and these gains only increase throughout the correction process.

Recommended symbols in the "Regulatory Information" section


A valuable tool for bulk data management

When they encounter incorrect or incomplete data, retailers can send a notification to the relevant manufacturer with just a few clicks, with the help of AI. This ready-to-send message includes suggestions to facilitate the process of bringing the product data into compliance. AI also assists the manufacturer who receives the message: algorithms make suggestions for the fields that need correcting, based on the product or the information that has already been filled in.

The productivity gains offered by this AI-assisted process are even more evident when it comes to large-scale data management. One example could be if a retailer received a formal notice from a consumer protection agency, requiring them to bring information about a specific allergen into compliance as quickly as possible — with the help of AI, the retailer can send a standardized note, including suggestions for corrections, for all of the affected products. All of the brands concerned by this will receive this message and be able to act quickly, and retailers can track the progress of these corrections.

AI and retail innovations

AI is already assisting users, both retailers and brands, throughout every step of the process, beginning with the creation of the product page. Once a product’s commercial name has been entered, algorithms suggest a product category, net contents values, suggested labels, regulatory formats for allergens, information about the types of diets that the product is suitable for, etc. This tool, which is able to take advantage of the thousands of available product data points, assists the user in the process of validating all of this information.

These suggestions offer significant time savings, greatly reducing the time required to create product pages that are compliant with both regulations and retailer requirements, while also improving data quality. And this is just the beginning. Multiple AI suggestions can already be accepted at the same time. Very soon, simply entering a product name will enable the tool to predict content suggestions for the various different fields that need to be filled in, such as product category, composition, allergens, labels, etc. This will also optimize analyses regarding consumer habits, for increasingly customized offerings.

NEW ON ALKEMICS – All of these suggestions will now appear at the top of each product page, allowing you to quickly view and validate them. Ensure high-quality data with just a few clicks!