Thesis Archive

A cost minimization model for a multi-component product closed loop supply chain considering big data dimensions (2021)

Pamela Nichole D. Chuateco
Carla Natalia Isabel Y. Del Rosario
Ysabel Dominique L. Reyes

Abstract:

The integration of forward and reverse flows in a closed loop supply chain has led to more uncertainties such as order forecasting, information distortion, demand forecasting, and varying return and return quality behavior. The rise of digitization in supply chain management holds reasonable cause to believe that the adoption of Big Data Analytics (BDA) into the CLSC would be a viable solution. Big Data, through its dimensions of Variety, Velocity, and Volume, has the potential to increase the accuracy of decision-making in procurement, create more precise forecasts in demand, increase the probability and quality of returns. Despite the abundant gains that may be realized, current literature and mathematical modelling on BDA capable CLSCs are scarce and barely tackle the potential effect on the operational decisions a manufacturer must make when facing the aforementioned uncertainties.

To address this gap, a mixed integer nonlinear programming mathematical model for a multicomponent product in a CLSC with BDA integration was constructed from the manufacturer’s point of view. The system considers uncertainties across Procurement, Demand Management, Collection, and Recovery, such as order allocation, demand forecasting, return uncertainty, and recovery failure. The unique feature of this study includes the operationalization of the three dimensions of Big Data, which take on intensity values among low (0.00-0.34), medium (0.35-0.66), or high categories (0.67-1.00). These intensity levels are further integrated into per activity effects, which are observed across the aforementioned uncertainties.

Due to the integrative nature of Big Data wherein the dimensions of Variety, Velocity, and Volume coexist, a sensitivity analysis was conducted across seven key combinations (MMM, MLH, MLM, HLM, HLH, HML, HMM) that demonstrated behaviors with the most variation. It was found that the systems with BDA prioritized recycled raw materials and met order targets across 6/7, excluding MLH. This revealed that the system benefits from the order flexibility provided by the procurement effect only when the reverse flow activity effects are at par or supersede the average. Further, regardless of linear or seasonal, it was found that the uplift to actual market demand is proportional to the demand effect. Across all systems, the demand forecast was at par with the actual market, showing that BDA maximizes demand potential for optimality. In terms of collection, BDA was able to uplift return behavior, with more growth experienced in the customer segment with low initial return probability. This shows that BDA is able to transform underutilized areas to enhance the chances of return. When the collection effect realized through BDA is larger, the total amount of incentives disbursed to consumers decreases, showing how the data investment reduces variable costs in the long run. Lastly, higher recovery effects maximize the amount of units recycled and remanufactured as opposed to disposed, showing that in CLSCs that aim to reduce material costs through takeback may benefit largely from BDA integration. All seven systems realize the investment payback period in a maximum of four years, showing the financial rewards of the set-up. Across the system combinations, the HMM system holds the best collective performance across CLSC operations and cost savings.

For future research on closed loop supply chain systems with the presence of Big Data Analytics, it is recommended that the scope is expanded into other functions that carry uncertainty, such as logistics and production planning. Further, it is encouraged that researchers innovate methods that create new methods to operationalize the quantitative impact BDA brings. Future researchers are encouraged to explore the relationship of BDA with dynamic demand, return, and collection, as well as expand the objective function to model more trade-offs.

Adviser:

Dennis E. Cruz
[email protected]

Panel Chair:

Jayne Lois G. San Juan
[email protected]

Panel Member:

Willy F. Zalatar
[email protected]