How to know how many litters of diesel, and gasoline I’m going to sell using machine learning

Decision-making about which establishments to open, where, and how to generate new customers has never been an easy task in retail network planning. There are so many factors of high commercial value to take into account that reaching the best and most optimal results depends on an in-depth analysis.

Locatium proposes a solution, and for this, it has developed a planning and optimization model for the entire retail network, both online and offline. We will be talking about this and much more in this article.

It is no secret that fuel retail companies are constantly exposed to location problems, and this is multiplied when there is a very extensive retail network. This, in terms of sales, translates into little or no relationship with customers, both real and potential, and therefore less income. Despite the fact that sales are the main indicator, there are other factors such as cannibalization, location of competitors, points of interest, etc., that would also be affected. 

Locatium, based on its large real-time data repository, is able to take these and many more factors/variables into account and design a machine learning model that generates the best locations and an increase in sales between 15% to 30%. But what is the most important data, and how can I use it to arrive at such promising results? Keep reading, and you will find out everything.

In this sense, the mobility of the vehicles becomes a key factor since it is the main indicator of the possible number of customers of a new service station. However, there are other factors that are also of great importance in the planning of the retail network. There are points of interest that are of great commercial value for service stations, such as locations of competitors, nearby partners, convenience stores, car washes etc. (Visit our Use Cases).

The best combination of all these elements can detect the weak points of the location, as well as provide recommendations and suggestions on possible locations, with a high degree of accuracy. So much so that we can predict the number of liters that can be sold at each service station, even going so far as to differentiate between types of fuel.

But, at this point, how can we really predict the number of liters that are going to be sold? To the analysis between the variables presented, we add the sales data of our client, which turns each analysis into a unique analysis. All of this information (billions of pieces of data) is used to train an AI model that we have developed internally, which is capable of learning existing features. Then the result of said learning is extrapolated to the possible zones.

No doubt, such results depend on the quality of the data, but that has never been an obstacle for Locatium. We have global data coverage, high accuracy/granularity, and most importantly, real-time data. Here is an example of our data repository, which you can expand at this link

  • Tracking billions of mobile devices (which we use for dwell analysis, home location identification, visitation analytics, etc.)
  • Tracking billions of vehicles (which we use for traffic analysis on the road network, vehicle home location identification, etc.)
  • Consumer Data (purchasing power, consumer profiles by category, etc.)
  • Advanced Demographics (gender, income, education, nationality, etc.)
  • Real estate insights.
  • Credit Card insights.
  • Social Network Insights.
  • Points of interest, including the location of competitors.

Thanks to the way we deal with the data, we are capable of perfectly arriving at different use cases, which we adapt and personalize to each client, such as site selection, cannibalization analysis, competition intelligence, white space analytics, capacity planning, portfolio optimization, and so on, that you can see on our website.

Now you know that we can predict the number of liters that your company will sell in the new locations, and with a high level of accuracy and geographic granularity. If you want to see a demonstration of our work, as well as learn about our visualization tool, do not forget to contact us. Feel free to book the time that works best for you on our calendar tool



Contact us for a practical demonstration of its functionalities.