Optimize asset maintenance through predictive costing

Our client wants to better planned the cost of the rail maintenance and renewal works. The objective is to give an accurate estimation of future costs based on thousands of past jobs, with dozens of criteria. This model should also help project manager to optimize the conditions under which the work is carried out, like night-time, time of the year, with alternating train circulation etc.

Scope : 6 years of historical data  – 8 b€ worth of works analyzed – over 250m lines of data with dozen of attributes


  • Develop machine learning regressions models to predict costs.
  • Develop interactive visualizations. These has been key to understand the main cost drivers and embark the client team in a more data-driven approach.
  • Quantify the complex impact of each cost driver based on 6 years of historical data


  • Implemented an advanced cost prediction model based on historical data and industry expertise.
  • Delivered visual apps to explore historical data and cost correlations