Proposal of Machine Learning Algorithms to Improve Sales Forecast in Schneider Electric
Tutor / Supervisor
Student
Ibled, Charles Franck
Document type
Master thesis
Date
2020
rights
Open Access
Publisher
Universitat Politècnica de Catalunya
UPCommons
Abstract
In the context of my previous internship in the Global Sales Operations department of Schneider Electric, I participated in the development and creation of a pipeline management playbook that aims to spread good practices about pipeline management in the company. One of the challenge is to compute a consistent sales forecast that will be used to allocate resources and undertaking a series of actions to ensure that the sales target will be reached. My objective is to propose a method that improve the Sales Forecast accuracy and provide business insights and business visibility. I started by analysing the current Sales forecast methodologies used by Schneider Electric. From my observations, I decided to use the data available in a different way. I created an Hybrid Method that combine Machine Learning Algorithms and classic forecast analysis techniques. The Machine Learning algorithms predict the outcome of the current pipeline. It predicts if the opportunities will be Won or lost and when it will be qualified as so. The classic forecast analysis techniques predict the level of opportunity creation over the year and if these opportunities should be closed within the year and with which success rate. I defined this second analysis as the Pipeline dynamic analysis. By combining these two analysis, I can predict accurately the outcome of the current pipeline and predict what opportunities will be created and won during the year which is the goal of the Sales Forecast. The key point that can be improved in my method is how I predict the time needed for an opportunity to be qualified as won or lost. In order to implement the method and validate it, some analysis should be automatise in a Business Intelligence tool to be able to analyse more data. The conclusion of this paper is that the classification algorithm that have been developed are able to predict the outcome of the current pipeline with an accuracy superior at 90 % which represents a real opportunity for the company. Moreover, I highlighted the role of the pipeline dynamic analysis that I have not seen in the current forecast methodology and I proposed a method to capture it.
