Boosting supply chain software with vmi weath

Can you present Weathernews France and its business activity?

The Japanese company Weathernews was founded in 1986 following a maritime catastrophe—a cargo shipwreck at the port of Onahama in Fukushima. The merchant vessel had sunk from a lack of weather information, and the Japanese merchant marine desperately needed a way to avoid future incidents. They also wanted to economize on weather-induced fuel consumption increases and to calculate arrival times in their ports.

The company’s business today is for the most part dedicated to land, air, and maritime routing. Two years ago, Weathernews purchased Climpact-Metnext so they could further develop in Europe. The company then became Weathernews France.

Why integrate weather data in a VMI?

In big name distribution, seasonal or weather-sensitive product sale is extremely dependent on climate. This is particularly true for ice cream, a summertime product, and soups, a wintertime one. To help industrial companies and distributors anticipate demand as early as possible, Weathernews took a closer look at dependency between weather data and retailer sales in order to come up with a mathematical model.

The model provided past and future indicators useful to Supply Chain players for the management of production and supply. The goal was to prevent selling out in-store, to diminish overflowing stocks, and to reduce logistics costs. Now, marketing departments can also use the data to better adapt during sales periods.

From an operational standpoint, the main challenge is to successfully constitute a mass of specialized information in a simple format that is directly usable by brands and suppliers.

What are the handling specificities for this weather data?

Predictions are usually available at a national level. But weather doesn’t have the same influence in Brittany that it does in Alsace! To make predictions more reliable and account for regional specificities, weather data must be ponderated by integrating sales history collected during prior years.

Other parameters also have an influence on consumption: Vacation periods, calendar specificities, and promotional information are all included in our prediction tool.

What constraints does this model impose?

The most important question in our model regards the level of data precision with respect to the product. Is it better to use a general indicator or an approach per product? Or perhaps a brand or format-oriented one? It is the specific challenge faced that will determine whether one or several indicators is used.

Whatever the case, the model must be both precise and robust: the more it is fragmented, the more it will lack reference data. The right level of analysis must be found. We start by finding the most aggregate data level, then send our datasets to Generix Group, who integrates them in the appropriate VMI with the right level of precision.

Anything else to add about your Generix Group collaboration?

Weathernews France partnered with Generix Group to tackle its VMI weather issue alongside a major industrial, distributor, and drink production specialist. After a kick-off meeting, our teams met once every 15 days to speak about the project’s implementation.

Using past and present weather indicators provided to Generix Group for integration in the VMI, we were able to analyze the company’s data, improve its prediction models, and come up with a desired data mesh according to business activity (drink or store sale). We performed tests on two of the distributor’s warehouses to evaluate the gains achieved, particularly with historical data integration.

Following initial feedback, we observed that the model was functioning correctly with drinks, but was not satisfactory for in-store sales. Having observed this, we refined our modeling techniques. Lastly, our experimentation showed a performance improvement for each individual warehouse, with particularly impressive gains when integrating past data.

For further reading: The 3 key performance indicators of VMI software