Systems Optimization

Businesses are constantly optimizing, particularly in the design and operation of complex systems. The engineering and business combination in the LGO curriculum uniquely prepares students to work on systems optimization problems in a variety of industries. These problems all use large-scale analytics models to produce results.

Saving Lives by Increasing Patient Throughput

Wendi Rieb (LGO ’15)

Company: Massachusetts General Hospital, Cancer Infusion Unit
Location: Boston, MA

Problem: Massachusetts General Hospital’s Cancer Infusion Unit has about 40,000 patient visits a year (145 a day). They come for chemotherapy, blood transfusions, infused antibiotics, and hydration treatments. There are long wait times for patients and a shortage of nursing staff and resources, yet throughout any given day, only about 55% of chairs are used because visits are concentrated from 10 a.m. to 2 p.m. At other times during the day, the unit is underutilized. Meanwhile, MGH is growing at 2.5% annually. It is critical for the hospital to use existing facilities more efficiently. Wendi increased the Cancer Center’s throughput by smoothing the Infusion Unit’s intraday utilization.

rieb imageApproach: Wendi used retrospective systems optimization modeling and prospective simulation modeling techniques. Using her algorithm, she proposed new scheduling guidelines to generate more predictable and balanced patient flow, which optimized the scheduling process for infusion-only visits and practice-to-chemotherapy appointments.

Impact: Wendi’s algorithm demonstrated the potential to recover 20 chairs—33% of capacity—at mid-day while smoothing throughput throughout the day. In total, the scheduling technique allows for 10,000 additional chemotherapy treatments annually. The algorithm also suggested that the optimal state could be achieved with minimal adjustments to staffing and no adjustment to the Oncology Practice. The algorithm also respects the existing primary nursing model and treatment-specific limitations.

Wendi’s system allows the unit to absorb growth and relieve existing resource shortages.  The Cancer Center could consolidate infusion operations currently housed in other facilities into the unit to generate more operational efficiency.

Ordering Policy and Supply Chain Responsiveness

Ana María Ortiz García (LGO ’16)
Company: Verizon
Location: Basking Ridge, NJ

Problem: Verizon purchases Customer Premise Equipment (CPE) from various suppliers. They also create long-term equipment purchase forecasts. Supplier contracts establish ordering policies that restrict how much forecasted order quantities (OQ) can be adjusted, despite known or anticipated changes in demand. Historically, this policy creates greater variability on inventory levels and strains production plans. Verizon asked Ana to improve the supply chain’s responsiveness by changing only the ordering policy.

Garcia imageApproach: Ana used historical demand data to investigate how alternate policies would impact key metrics. This approach was applied to one pilot product, and can be used as a framework to evaluate contract terms.

Impact: For each simulation, all policies were ranked across metrics to create a visual representation of policy performance. Ana recommended an alternative approach that improves the prioritized metrics. For the piloted product, changing the ordering policy reduced inventory by $3.4M, saved $500,00 per year and reduced OQ variability by half.  These results can be extended to other products for additional benefit to both suppliers and Verizon.

Productivity Optimization of an Automated Material Handling System

Willow Primack (LGO ’15)

Company: Amazon
Location: Los Angeles, CA

Problem: With the expansion of its fulfillment center network, Amazon developed material-handling robotics systems that use conveyance, sensors, and software to work with associates. These systems improve productivity and labor efficiency and standardize operations between fulfillment centers. Amazon asked Willow to examine an automated material-handling system at a fulfillment center. Her project is a case study on queuing theory and simulation techniques for modern warehouse systems.

primack-chartApproach: Amazon’s warehouse system is organized around three central stations: item induction, processing, and batch processing. The system is laid out in parallel lines linked by a central sorter that routes items to different stations. The central sorter is programmed to balance work uniformly between production lines, which have different capacities due to varying skills of the associates.

Willow’s model matched product arrival rates to each line’s individual capacity using utilization ratios. She analyzed data on item arrival and processing at different stations to characterize the how much time was needed for each activity. She then conducted an experiment using a Monte Carlo simulation model.

Impact: Willow’s initial results suggested that productivity could improve by 5-10% if Amazon only changed their software logic. This would be a savings of several million dollars per year across Amazon’s fulfillment center network. The adaptive system response will allow exceptionally skilled associates to work at full pace. The project’s approach to system characterization and modeling will also contribute to the company’s continuous improvement process.