How dynamic pricing algorithms can increase gross profit by +10%Data science is helping Brazilian Oil & Gas sector to overcome challenges imposed by the economic crisis that affects the country.
Guilherme ZuanazziBlockedUnblockFollowFollowingJun 21The Oil & Gas sector in Brazil has been facing challenges due to the economic crisis and the technological revolution.
These are the main pain points:1.
The fuel resale segment is in a context of restrained margins by both sides of the supply chain:• The upstream segment reduces gas stations margins through increased fuel purchase costs.
• Consumers are facing limited family budget and becoming more price sensitive, turning it difficult to resellers to pass on cost increases.
2.
Mobile apps (such as Waze) are providing consumers with real-time price information, allowing customers to find out which is the nearest and cheapest gas station in the region before refueling.
3.
Petrobras, the semi-public Brazilian petroleum multinational, has implemented a new pricing policy which allows prices to move accordingly to fluctuations in international commodity prices.
This new policy, which begun in 2016 and has been intensified in the following years, brought a new dynamic of fuel purchasing costs.
Resellers do not know how to deal with it, since this sector has been consolidated in Brazil during decades of high-regulated prices.
These points together result in a greater pressure to margins in a sector whose margins were historically low.
Considering the highly competitive and dynamic environment, resellers must make daily decisions on pump prices according to multiple variables.
Some of these include: the cost of fuel purchase, taxes, competitors’ positioning, vehicle flow, day of the week and, the most important, the willingness to pay of consumers.
Willingness to pay is expressed in the concept of elasticity.
This concept seeks to answer mathematically what is the percentage of increase in demand if there is a price reduction in X percent?.And what is the drop percentage in demand if there is a price increase?Answering these questions using science is a must to successfully optimize pricing.
This is the goal of dynamic pricing algorithms.
By leveraging large databases it is possible to identify and isolate the effects of elasticity.
We can then simulate the demand reaction for different price and market scenarios, and optimize price decisions, capturing margin or volume, depending on the business strategic goals.
The daily cycle consists in the following steps: (1) modeling, (2) simulation and (3) optimization.
As demand reactions occur, the algorithm updates the coefficients used for calculation, learning accordingly to the success rate of its own forecasts and suggestions.
Different machine learning techniques can be used for step (1) modeling, such as: Ridge Regression, ARIMAX, Kalman Filter and Neural Nets.
The success in this step all depends in asking the right questions.
For example, ARIMAX technique can be very powerful for demand forecasting as a function of prices (“X” being a set of exogenous price-related variables).
However, if the question we want to answer is “What is elasticity effect’s value?”, a technique like Ridge Regression might be preferable.
Even if the forecast of demand is less assertive in this model, the algorithm will be able to better isolate the elasticity effect’s coefficient.
It was after several months of development and implementation that Aprix, a pricing startup, developed a pioneer artificial intelligence for dynamic fuel pricing.
The enterprise works with the major fuel resellers in Brazil.
When comparing gas stations that use the algorithm VS gas stations which continued using the traditional pricing method (based on Excel spreadsheets), the group using the algorithm has achieved an average increase of +9.
6% on gross profit.
The technology enables gas stations to survive in the new highly competitive context, turning the threat into an opportunity.
Although the Oil & Gas industry is still composed of traditional enterprises, there is no doubt that the future of pricing is based on AI and that it is gaining more and more space around the world.
Dynamic pricing algorithms are already used in fuel retail, mainly in the UK and the United States.
Aprix is the one who is building this future in Brazil.
Faced with this trend, the question we ask every day in Aprix is the following:What are the next sectors that will use dynamic pricing algorithms to increase profitability?.