In an ever more competitive market, our client in the agricultural thread market has been following a mass-customization strategy by tailoring the package of its products for each client.
The budgeting process, previously used to define production needs, is no longer enough to address such a degree of customization. Moreover, clients’ preferences change throughout the year, and there is no decision support system to review the initial predictions. Hence, stock levels have been rising to accommodate these fulfillment challenges.
Our solution encompasses a forecasting algorithm and a tactical production planning model.
The forecasting algorithm relies on an ensemble of a time-series forecast with a machine learning method capable of accounting for the budget and advance demand information.
The monthly update of the forecasts guarantees a more proactive revision of the sales predictions. The tactical production planning model works hierarchically. Information such as sales volume, client concentration, and demand uncertainty is first used to make a yearly categorization of the products as Make-to-Order (MTO) or Make-to-Stock (MTS), with further guidelines whether an MTS product should be produced to the final packaging or as a semi-finished product.
Then, a rolling optimization algorithm is monthly employed to prescribe the most profitable production plan within a 12 months’ horizon.
By adopting our holistic solution, our clients may expect a reduction of stocks by 17,5% and a decrease in machine utilization by 11% without jeopardizing customers’ service level.
The anticipation given by the yearly planning horizon leads to an increase in the production lot sizes. From a qualitative perspective, our decision support system aggregates all the relevant information in a flexible and accessible interface.
A more sophisticated and data-driven approach allows the team to allocate its time to more value-adding activities in the process.
By: Daniel Pereira
