Challenge

A global company specializing in label solutions and packaging services. The goal was to implement continuous improvement of demand forecasts for 100,000+ items aggregated at different levels at a yearly and monthly level for next 24 months existing on the SCM platform. ROBUSTNESS ISSUE - Every periodic data load on the platform disturbed the tweaked forecasts.MANUAL / NON_MODULAR - The SCM1 platform used by the client lacks the functionality to automate the forecast tweaking process.Hence, a need for automated forecast improvement solution was realised to make the demand planning process efficient and enable faster decision making.

Approach

Automated Forecast Tweaking. Clustering based models were built to cluster the forecasting model hyperparameters at different hierarchy levels. H-PARAM GROUPS - The clustering model can be dynamically tweaked to group the hyperparameters as per the business requirements. CLOUD INTEGRATION - The model was then Integrated to a cloud-based platform to automate the forecast improvement. DYNAMIC ACCURACY CHECK - The developed solution dynamically checks forecast accuracies at multiple hierarchy levels – brand/ category/location on a monthly basis after every data load.

Outcome

Reduced the no. of items requiring manual tweaking each week by 40%. Standardized approach to manual tweaks, minimizing subjectivity. Reduced total time taken by 2 weeks. Forecasts available for business use by Week 3, instead of Week 5. Improved Supply Chain Planning on account of above.