THE PROBLEM
Client’s finance function makes monthly predictions for the current year and the coming year for 11000+ Opex time series on a monthly basis. However, the predictions are done manually, making the process laborious, time consuming and error prone. Hence the finance team wanted to implement a machine learning solution to generate these forecasts automatically with a high degree of accuracy

INXITE OUT APPROACH
People Cost Model
Developed model to predict wages for different workgroups of employees for 450+ cost centers involving many functions at several locations in several countries and currencies. To ensure business explain ability and handle business constraints such as allocation of wages to employee groups based on their seniority with certain minimum predefined gaps in-between and upper/lower bounds, a custom algorithm was developed that was robust against issues such as business process driven irregularities and data-insufficiencies typical in such scenarios.
Travel and Expenses Model
Developed a model to predict the travel expenses for each cost center based on:
- Historical travel patterns and the corresponding distribution mix
- Seasonality of travel
- Head count of the cost center and its distribution by work group.
Scope Change Model
Developed mechanism to handle business events like creation, migration, merging and splitting of cost centers
RESULT
Achieved an aggregated percentage error of ~1% (critical metric for the business) across all cost centers