Demand Planning
Development of a historical sales base, purged from outliers and events.
New Demand Planning approach
Baseline
Development of a historical sales base, purged from outliers and events
Consolidation of events data, in standardized and replicable format, so that they can be leveraged in the Generation phase
Generation
Definition of forecast horizon (short- vs long-term)
Introduction of automatic methods, reducing the manual effort and increasing robustness
Time series models combined with causal models (machine learning), grasping historical trends alongside event inputs, enabling more accurate sales forecasts
Validation
Implementation of a strategy to validate all generated forecasts, highlighting forecasts with low accuracy or deviating from the target
Analysis of similar products (brand, packaging, …) historical sales and forecast to enhance the validation process
Monitoring
Accuracy metrics definition, transversal to the whole organization
Monitoring dashboard with views customized to the needs of the stakeholders, allowing different types of analyses
Identification of KPI targets to assess improvement opportunities
Boost the forecast
Generation
Time Series Models
The forecasting model uses times series analysis that are able to capture trends and seasonalities (e.g. various exponential smoothing and autoregressive and moving average models).
Aggregation
Different levels of aggregation are automatically tested and evaluated in the history, namely both product and sales channel hierarchy.
Causal Models
The model will further increase reliability by applying causal models (e.g., machine learning) to relate trends and seasonality to exogenous variables by including events info and other data.