Everyday we generate enormous amounts of data through transactions in commodity markets. There have been many estimates of the quantity of data generated, the great hype of Big Data in 2014, and how firms should be investing in this area.
I want to take you through the example below and elaborate my perspective on Big Data in relation to commodity markets.
Example: Let us consider the case of ethanol production, merchandising, and consumption. Ethanol can be produced petro-chemically or from agricultural feedstock, such as sugar or corn. The decision to produce ethanol is determined by two factors – revenues and costs of the ethanol business.
Revenues are decided by the forward price curve, and costs are decided by the source of raw materials used to produce ethanol.
Revenue Issue We need to look into the price of ethanol and volume of ethanol produced. Prices of ethanol are a function of its demand as well as its supply in the market.
Substitution effects of other oxygenate compounds in gasoline and cross elasticity between ethanol and similar products also determine the price structure of ethanol. Similarly, the volume of ethanol produced is a function of capacity utilisation, expected gross margins, inventory financing costs and so forth. It is evident that none of the metrics we use in the revenue side is as exogenous as we would like to it be and depends instead on a number of factors, making it co-dependent on many variables.
Cost issue The cost side of the business is similarly impacted by the decision to use sugarcane versus corn (as ethanol can be derived from either). If one uses sugarcane, then the issue of production optimisation of ethanol versus, say, sugar dominates the consideration of the cost of sugarcane as an input.
Similarly, the use of corn as a raw material for the production of ethanol is decided by the price of corn, substitution effects of feed demand for corn, blending economics of ethanol into gasoline and a variety of such issues. We compound this optimisation problem by taking into account the subsidy and effect of petrochemical prices on ethanol demand.
The above illustration elucidates the issue of complexity in decision making with so many variables. Almost all businesses I know of in the world now overcome this complexity by making arguably grand assumptions that help them solve for the optimal solution – creating an enormous fluctuation in their earnings, great uncertainty in planning and a suboptimal decision making, to say the least.
Three big areas There are broadly three big areas within the big data in this above-mentioned problem. The ability to gather all the data relevant to the ethanol business; the capability to conduct a robust analysis with the data available and the benefit extracted from this big data analysis to make predictive decisions. In this case, one should consider building a platform for accumulating all data relevant to the ethanol business such as corn prices, ethanol supply and demand, ethanol production centres, cost of natural gas, sugarcane yields etc. The next step would be using these data sets intelligently to analyse it and see how different variables affect ethanol margins.
This will use sophisticated mathematical modelling combined with intelligent choice of data also known as data dredging. The final step will be to use this analysis in making decisions for capacity planning, capacity expansions, ethanol pricing, so that it will lead to enhancing gross margins in the ethanol business.
Conclusion Data mining and analysis is the new age and commodity business is the old age. The challenge is to place this new age business on top of the old age business and extract the best value. I can only see a big opportunity.
The writer is based in New York and is the founder and Managing Director of Opalcrest (www.opalcrest.com). Prior to founding Opalcrest, he was a commodities trader in London and New York. He can be reached at pravin@opalcrest.com