image (2)

Putting data science to practice at Nutreco

About Nutreco

At Vantage AI, we serve a wide variety of clients. From startups to multinational corporations, ranging from digital natives to traditional companies. Last year, we were given the opportunity to assist Nutreco with setting up their data science capability. Nutreco is a global leader in animal nutrition and aquafeed. It was founded in 1994 and has over 10.000 employees in more than 30 different countries. Nutreco used to be a company focused on traditional data analysis, but is now in the process of transforming into a company maximizing the added value of digital. Part of this transformation consists of the creation of a data science team. We assisted in laying the foundation for this team by putting strategy to practice.

The data science team

One of the main elements of Nutreco's digitalisation process is integrating data science in their tools. Marc Jacobs, the manager of data innovation, R&D, modelling and data science centre is leading this change. “My primary responsibilities are to develop models and algorithms that add to the value proposition of digital products and services” Marc is also giving the support for the statistical expertise and skills for more than 50 researchers. 

 

"My primary responsibilities are to develop models and algorithms that add to the value proposition of digital products and services"
Marc Jacobs - Manager Data Innovation

Getting started with machine learning

At the outset there was a strong focus on strategy within Nutreco’s data science approach. A lot of emphasis was put on questions like ‘how do you start a data science project?’ And ‘how can data deliver real benefit?’ However, less attention was paid to putting data science to action. Nutreco realised that the first project had to resonate well within the broader organisation. It had to result in a tangible product that delivers real business value, which in turn could serve as a catalyst for future data science projects. Integrating this within an existing and popular project within Nutreco, called NutriOpt, would be the next logical step.

NutriOpt is an innovative, digital platform that delivers integrated solutions supporting sustainable precision farming for optimal animal performance and business success. One of the services in NutriOpt is advice on nutrition mix. This tool enables users to find the optimal nutrition mix, which in turn leads to optimal animal performance. The input of this tool are various nutritional values which are stored in a matrix. 

This is where everyone envisioned the newly set-up data science team could be of great value. If they would be able to predict future nutritional values, they could be providing advice to nutritionists on the applicability of their matrix. In effect, this would make sure that nutritionists optimize their feed based on the most accurate data available. The idea here was to predict nutritional values over time based on historical data. The resulting predictive model could then be used by all of Nutreco’s clients, whilst also being improved over time to further enhance the user experience.

Why Vantage AI

We played a key role in co-developing this first data science tool for Nutreco. Not only by assisting in the development of the algorithm, but also by doing the required engineering to put the model to production. “There were a lot of stakeholders involved from a range of teams across the business, which made it not an easy task to create this solution. Vantage AI’s experience and approach definitely made the difference when it comes to getting to a production-ready model for predicting nutritional values”, according to Marc Jacobs.

Results

In a short time we delivered an API that could be used by a system that was already in place for clients of Nutreco. This way, we could empower and accelerate current products within Nutreco without having to create a whole new product. Marc was pleased with the result: “The project was fun, but the real value in this project was to deliver a data science solution and get the organization thinking about future digital projects. Now we know what it takes to put a model in production. The model is just a small part. This was just the start!”