Sales Forecasting Self Service Analytics

About Client

The client is one of the top 100 industrial groups in the world with a heritage of 350 years. They have consistently demonstrated their ability to invent products that improve quality of life. They manufacture and market solar control glass, fire resistant glass and other various types of float glasses.

Background

Organizations often use different sales forecasting methods to predict long-term and short-term demand. However, predicting sales or demand with high accuracy is no easy feat. The process is complex as a number of variables influence demand. There is also a need to build and test multiple models to come up with the most reliable forecast.

Challenge

  • The client wanted to plan production capacity for windshield glasses based on prediction of car sales for the near-future. Cars are expensive and the sales forecast depends on a wide range of macro and micro variables.
  • There used to be delays in the decision making process due to involvement of the IT team at every step due to continuous model building.

Our Solution

  • Data, including flat files were collated from disparate data sources
  • Data cleansing was done after combining several datasets together
  • R & Shiny dashboard was used to show the effect of predictor variables on car sales using correlation plot
  • Feature engineering was applied based on the correlation plot to identify important variables. Some of the variables that were considered included the GDP, Loan percentage, Petrol-Diesel Price etc.
  • The framework included allowing any business user to run several algorithms including Holt Winter, Arima, GAM, etc. for better decision making.

 

Outcome

  • Forecast of car sales based on variables identified helped the client predict production capacity for windshield glasses.
Ad Hoc Reporting
Capabilities Allowed Business Users To Create Custom Reports On Their Own, Thus, Reducing Dependency On IT Time And Cost Of Report Development.

An accuracy of

82.6%
was achieved within the first 6 months of implementation. The use of Machine Learning ensures that the accuracy increases over time. Self-Service Analytics for business users enabled the client to speed up their decision-making process with minimalistic intervention from the IT team