Refueling profitability in the oil sector with machine learning and analytics

When consumers look at the price on the pump in forecourts, they are not aware that it is determined by a large number of factors invisible to them, other than the crude oil price. Fuel producers (refiners or blenders) and dealers have to manage multiple sources of data daily, including prices, stock availability, logistics constraints, and complex contract structures that make the final price a highly scientific decision. 

COMPETING ON THIN MARGINS

Large and medium integrated oil companies are facing stiff competition from smaller players who take advantage of the market conditions and profit by optimizing margins around contracts and logistics. In response, most of the integrated companies have invested in advanced ERP systems that execute complex pricing algorithms to calculate and set prices per retail site, and even update them several times a day to respond to the competition profitably, although, this does not limit their losses through competition in contract lifting manipulation.

THE PROBLEM WITH SUPPLY AGREEMENTS

Oil companies tend to have supply agreements between themselves to support their operations that supply to an extensive retail network and serve large commercial customers (dealers, manufacturers, fleet companies, etc.). Their ability to absorb the risk by setting up these contracts gives them the advantage to operate at scale. Making them vulnerable to exploitation, at the same time, due to the lack of real-time information and/or the ability to process the data to control operations. Most operations take place from the third-party-owned terminals that can neither supply the information nor do they have the ability to process the data. This effectively removes the option of managing volume allocation per customer.

PRICE FLUCTUATION DETERMINES CONSUMER BEHAVIOR

One of the key factors to ensure enough fuel supply for the market is the ability to control the volumes lifted at terminals (depots). These are stocked based on the forecasts provided by their customers which in turn are based on the agreements with their customers. A key element that forecasts fail to include is the price fluctuation that determines daily operations and price determines customer behavior. 

WHAT IS GAMING BEHAVIOR?

Let’s assume a major oil company X operates with both contract customers and in the shipment market (i.e., they are allowed to lift fuel from various terminals using the agreement set by X and customers can also directly place orders with X). Based on the analysis of both contract lifting and shipments, customers display gaming behavior. Two short examples can illustrate this. If the contract pricing term is based on a monthly average, contract customers tend to lift more than average (over-lifting) when the contract prices are favorable. Similarly, if the shipment customers have a monthly average contract with another supplier, they may order more from the company when the prices are favorable. In both cases, unless the supply manager receives early warning of this behavior, the company stands to miss the opportunity for better negotiation terms on the contract for that day. 

HOW CAN MACHINE LEARNING AND ANALYTICS HELP?

The above situation can be avoided by employing machine learning together with analytics. Using historical data widely available in ERPs, machine learning algorithms can build profiles for each customer which will enable them to build behavioral models that can predict gaming. Analytics can determine better pricing terms for contract negotiations. Further, these models can process real-time data to monitor volume consumption using the information provided by terminals. Technologies borrowed from the banking industry can offer real-time credit assessment and highlight risk factors for contract negotiation on the spot, resulting in better margins. 

The additional margin calculated for company X above was estimated to between $3-7 per tonne of fuel, enough to generate millions of dollars of additional profit.

In summary, industry 4.0 and the internet of things combined with machine learning and analytics have the potential to provide a competitive advantage to organizations.

About

Panagiotis Tsiakis is a former Associate Partner, Infosys Consulting (until August 2020) and currently Director - Information Technology at HELLENIC PETROLEUM

https://www.linkedin.com/in/panagiotis-tsiakis-76ab231/

Source:

https://www.infosysbpm.com/blogs/digital-business-services/Pages/enhancing-customer-experience.aspx