Context
Commodity traders are large groups like Trafigura, Glencore, Mercuria, Vitol who buy, produce, store and sell commodities like copper, aluminium, steel, oil, gas (for the challenge, food and agriculture goods are not considered). They make their decisions based on market prices, supply and demand, … but unlike pure financial trading desks, they manage physical flows, transport fleet and inventories.
They already buy and use several data sources, worth millions of euros, and build their own indices against their decision criteria.

The challenge

You have access to a proprietary, brand-new kind of data: for all dry-bulk commodities, you know the industrial sites that produce these commodities (copper, aluminium, steel, other non-ferrous metals, cement, …) or consume them (copper concentrate, bauxite, alumine, …), and you know the level of activity of each of these industrial sites. For each site, this level of activity is measured as a value between 0 and 100, and is given on a monthly basis with a history of 6 years, a weekly alert on detected events (shut-downs, restarts) and a daily nowcasting.
Making use of these new data and of any other available open data, you have to define the data product which will provide the largest value to commodity traders, and become their #1 data source.
For the challenge, you will explain your understanding of commodity traders’ needs, describe the target data product, explain and quantify the value of this data product for commodity traders, and detail how, as a data scientist, you get to this data product from available data sources including your proprietary one.

The deliverable

3-page document maximum, in word format.
AI can be used, provided that it be made explicit, with the description of the model and prompts in annex to the document.