Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2018. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged student-submitted from PDF version of thesis. / Includes bibliographical references (pages 79-83). / Transportation costs constitute an important part of total logistics costs and have a dramatic impact on all kinds of decisions across the supply chain. Accurate estimation of transportation costs can help shippers make better decisions when planning transportation budgets and can help carriers estimate future cash flows. This study develops a forecasting model that predicts both contract and spot rates for truckload transportation on individual lanes for the next seven days. This study considers several input variables, including lagged values of spot and contract rates, rates on adjacent routes and volumes. The architectural approach to short-term forecasting is a neural network based on Nonlinear Autoregressive Models with eXogenous input (NARX) models. NARX models are powerful when modelling complex, nonlinear and dynamic systems, especially time series. Traditional time series models, including autoregressive integrated moving average (ARIMA), are also used and results from different models are compared. Results show that the NAR model provides better short-term forecasting performance for spot rates than the ARIMA model, while the ARIMA model performs slightly better for contract rates. However, for a longer-term forecast, the NARX model provides better results for contract rates. The results from this study can be applied to industrial players for their own transportation rate forecasting. These results provide guidelines for both shippers and carriers regarding what model to use, when to update the model with new information, and what forecasting error can be normally expected from the model. / by Xiwen Bai. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 45-46). / Logistics providers process millions of packages daily and collect an incredible amount of data from these shipments. As new sensors are added to more and more packages, companies will now have increasingly fast access to even more data. However, how will logistics companies leverage this idea of big data to generate the most business value for their customers? Using a qualitative approach by interviewing current users of real-time monitoring devices, we were able to understand how customers perceive the value added by this technology. Moreover, we scoured a significant amount of literature on sensors, the logistics industry, and upcoming technological breakthroughs. We quickly discovered that customers do not perform extensive quantitative analysis to determine the trade-offs and financial benefit of using real-time sensors in their shipping processes. Additionally, we found that customers are unwilling to analyze this big data themselves, but instead want their logistics provider to interpret the data to provide value-added services. Therefore, logistics providers should leverage all of the data they collect, instead of simply creating value when shipments become exceptions, e.g. out of temperature range. We propose using smart contracts on a permissioned blockchain to automate business processes and reduce frictions within the shipping parties and other intermediaries. / by Tarun Tiwari and Anthony Toteda. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references (page 27). / Changing regulatory requirements continues to be an increasingly complex issue in the medical device industry. Regulations place stress on regional supply chains across the world. Most recently, the European Parliament issued the Medical Device Regulation (EU) 2017/745 instituting new compliance framework for all devices manufactured, sold, and/or distributed in the European Union. The new framework requires the implementation of unique device identifiers and more stringent conformity assessment procedures. In addition, many device classification types have changed, post-market clinical surveillance has been instituted, and traceability through a centralized IT database is now mandated. While the the act aims to improve patient safety and efficacy across the medical device industry, it poses huge impacts across both the physical and informational flows in supply chains. This research evaluates the regulatory impact across supply chain operations using predictive modeling and machine learning. The model determines how various activities and events in manufacturing and sourcing environments contribute to supply constraints when modified to accommodate new regulatory requirements. The model also determines how product attributes contribute to performance variability. By taking a proactive approach to assess the impacts of regulatory changes, firms can optimize supply chain flows to reduce cost, lead-time, and service level risks. / by Melissa Medina. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 113-116). / Supply chain visibility is critical for businesses to manage their operational risks. Availability of high quality and timely data regarding shipments is a precursor for supply chain visibility. This thesis analyses the errors that occur in shipment data for a freight forwarder. In this study, two types of errors are analyzed: system errors, arising from violations of business rules defined in the software system, and operational errors, which violate business rules or requirements defined outside the software. We consolidated multifarious shipment data from multiple sources and identified the relationship between errors and the shipment attributes such as source or destination country. Data errors can be costly, both from a human rework perspective as well as from the perspective of increased risk due to supply chain visibility loss. Therefore, the results of this thesis will enable companies to focus their efforts and resources on the most promising error avoidance initiatives for shipment data entry and tracking. We use several descriptive analytical techniques, ranging from basic data exploration guided by plots and charts to multidimensional visualizations, to identify the relationship between error occurrences and shipment attributes. Further, we look at classification models to categorize data entries that have a high error probability, given certain attributes of a shipment. We employ clustering techniques (K-means clustering) to group shipments that have similar properties, thereby allowing us to extrapolate behaviors of erroneous data records to future records. Finally, we develop predictive models using Naive-Bayes classifiers and Neural Networks to predict the likelihood of errors in a record. The results of the error analysis in the shipment data are discussed for a freight forwarder. A similar approach can be employed for supply chains of any organization that engages in physical movement of goods, in order to manage the quality of the shipment data inputs, thereby managing their supply chain risks more effectively. / by Rohini Prasad and Gerta Malaj. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, First author, 2017. / Thesis: M. Eng. in Logistics, Massachusetts Institute of Technology, Supply Chain Management Program, Second author, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 60-61). / Returnable Transport Items (RTI) are a critical component of domestic and international trade. The large variability in the geographic supply and demand of goods shipped using RTI impacts the items overall availability at different locations within a network. This research focuses on improving our partner firm's RTI inventory supply in the United States and Canada by developing a one-month-ahead forecasting model to predict the net monthly international flows. To develop the model, six years of historical time series data was decomposed into key elements: level, trend, and seasonality. The results of the decomposition method were used to narrow the forecasting models considered to state space seasonal exponential, SARIMA, state space Holt-Winters, and multivariate regression methods. These four methods were then used to predict the pallet flows using two different approaches. In the first approach, two separate forecasting models were developed, one for the United States-to-Canada flows and the other for the Canada-to-United States flows. The derived Canada-to-United States value was then subtracted from the corresponding United States-to-Canada forecast to calculate the predicted net international movement. In the second approach, we forecasted the net pallet flows between the two countries utilizing only historical values of net international movements. Ultimately, 36 unique models were created using both approaches. The naive forecasting method served as a performance benchmark to the developed models. The performances of the 36 models were then compared using multiplicative and mean composite scores, both of which were based on three accuracy metrics: MAPE, MASE and MAD. Our research found that out of the 36 forecasting models, only seven models outperformed the baseline naive forecasting method. These seven forecasting models were further filtered by qualitative metrics such as ease of implementation and software platform dependence. The state space seasonal exponential model was ultimately recommended due to its superior performances on both the quantitative and qualitative metrics. / by Patrick A. Jacobs and Rajdeep Singh Walia. / M. Eng. in Supply Chain Management / M. Eng. in Logistics

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 76-77). / This research analyzed freight performance to determine the groupings of attributes that influence carrier performance. Binary logistic regression and hierarchical clustering were used to identify individual and groupings of freight attributes that impacted performance success in terms of on time delivery, on time pick up, and first tender acceptance rate. From the analysis, three main performance groups of carriers were identified and their subsequent underlying attributes and strategies were analyzed. This research confirmed industry belief that differing strategies and freight profile roles result in different performance, specifically that more focused carriers tend to provide better service than unfocused carriers. Insights for shippers were gleaned from the analysis and comparison of a different shippers' carrier portfolios. From this, diversified portfolios with a higher proportion of more focused carriers were shown to have stronger performance. The significance of this research is that it offers a strategic review of groups of freight attributes that contribute to performance outcome. Within this strategic review, carriers were shown to have different underlying roles within shippers' portfolios which may suggest the need of different ways of measuring their performance. / by Caroline C. Bleggi and Frederick (Qian) Zhou. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (page 79). / The security alarms services market in the United States delivers hardware equipment and services to homeowners and businesses to help monitor and enhance personal property protection. Customer satisfaction via wait time reduction, first call resolution, and cost minimization are key drivers of success to players in this market. Most companies invest heavily in customer service systems including call centers. Our client, AlarmCo, a top provider of property protection, manages an inbound call center that supports a range of questions from customers who call within thirty days from the alarm installation date. Often, security companies fail to utilize strategic solutions when managing inbound customer call traffic and default to reactive measures which unnecessarily increase customer wait times. The key question the team aims to address in this thesis is: "How can we improve the customer service experience for customers of a major security service provider in the United States?" For this thesis, MIT partnered with OnProcess Technology, a managed services provider specializing in complex, global service supply chain operations, to develop a robust framework to preemptively reduce the number of inbound customer calls, and thereby improve customer service. Using ABC segmentation, the team categorized customers by reason code and demographics. To simulate the client's call center queue, the team calculated the key inputs for the queuing model including average wait time, interarrival rates and number of servers. The team then chose and developed the M/M/n stochastic queuing model for the simulation. The M/M/n queue reflects a simple system with parallel servers, arrivals with a Poisson distribution and service times that are exponentially distributed. Next, the customer segmentation was used to develop targeted preemptive solutions. Taking into account feasibility ratings, the team assigned success rates to each solution and adjusted the inbound call data accordingly. By analyzing the outputs of the simulation before and after adjusting the dataset, the team quantified the impact of preemptive solutions on the call center queue. Ultimately, narrowing to twelve strategic preemptive solutions led to the enhancement of the as-is queuing model by reducing average wait time by up to 35%. / by Qiao Chu and Nisha Palvia. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Thesis: M. Eng. in Logistics, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 95-98). / As part of standard business cycles, new technologies continue to emerge that disrupt industries and capture market share from stagnant incumbents. In the trucking industry, Uber for Freight (UFF) is one of these innovative business models. Loosely defined as platforms which seek to more efficiently match shippers' loads with truck drivers, these companies are seeking to 'uberize' freight transport through algorithm-based applications. By eliminating the middleman of a carrier or broker, these startups' value proposition is cost savings and increased efficiency gained through a frictionless interface. While process automation has its upsides, many industry veterans have questioned the potential success of this business model. Furthermore, experts have expressed uncertainty regarding the operational mechanics of an UFF company as well as the true distinction between UFF and a traditional broker. This research seeks to address these questions about the UFF model by first developing a clear description of its players and processes, compiled based on interviews with existing companies in this space. Secondly, this research determines that UFF is best classified as a subdivision of brokers, providing similar services through a different business model that eliminates some degree of human intervention. More than simply automation, UFF provides additional benefits through its rating system and efficient payment processes. As a case study, this research then investigates the applicability of UFF within a specific company. The sponsor company, a large, multinational chemical company, maintains an extensive product offering that reaches customers across almost all industries. These products vary widely in format, hazardous material classification and service level requirements. Based on interviews with sponsor company representatives across functions and geographies, this research examines the challenges and benefits of incorporating UFF into a company's transportation strategy. From these learnings, it was recommended that UFF be implemented gradually, starting on a U.S. lane that transports non-hazardous products with lower service level requirements. If safety and service levels prove satisfactory, the sponsor company can scale accordingly to more complex products or lanes. While UFF has clear benefits and disruptive potential, it must be utilized with the appropriate products and customers; it is not a one-size-fits-all solution. / by Leah Davis and Joseph Lucido. / M. Eng. in Supply Chain Management / M. Eng. in Logistics

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 72-74). / Over the years, supply chain management has continued to change and evolve to become a major component in competitive strategy to enhance organizational productivity and profitability. While considerable research has been done in formulating accurate and robust demand forecasts, many areas for improvement remain in supply chain planning. In particular, many planning parameters (e.g., lead time, waste, yield, run rate, capacity, etc.), which are vital inputs into the planning process, are often not given the consideration they deserve. Oftentimes, the planned values of these parameters were not scientifically derived in the first place, or their actual values may have changed since the planned values' original inception and now differ significantly from its planned value. This research examined one type of planning parameter in particular - lead time, and showed there is room for improvement in how lead time is managed and considered within the current planning process. The research showed that using predictive analytics to predict lead time (predictive lead time) can reduce the deviation between the planned and actual values in the lead time parameter..Moreover, the analyses showed that using predictive lead time can reduce the safety stock cost, the manual labor required in exception management (re-planning), and the manual labor in purchase order management. / by Darryl Yau. / M. Eng. in Supply Chain Management

Thesis: M. Eng. in Supply Chain Management, Massachusetts Institute of Technology, Supply Chain Management Program, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 48-49). / This thesis explores the impact of the Drug Supply Chain Security Act (DSCSA) on various stakeholders in the pharmaceutical supply chain. Specific attention has been dedicated to the impact on manufacturers and distributors/retailers. Although various interpretations of the DSCSA are possible, this thesis takes the perspective of a centralized data model, and tests the feasibility of implementing a centralized database under both data nesting and unit level relational models. This is in contrast to the decentralized system, which is further explored in the partner thesis, Impact of Drug Supply Chain Security Act on US pharmaceutical industry under decentralized information flow (Chang & Mohan, 2017). Both quantitative and qualitative analysis are employed in this thesis. Quantitative analysis was conducted using publically available industry data, from which the impact on overall supply chain costs was modeled. Qualitative analysis consisted of stakeholder interviews, process mapping, and time studies to determine the extent of process changes and what they should look like to conform to DSCSA. After accounting for the current state of implementation, as well as real-world constraints, the findings indicate that the best-practice scenario for Manufacturers and Distributors/Retailers to conform to DSCSA is one using a Centralized data model and nested linked-pedigrees. Although this option is estimated to be 67% costlier than the least expensive scenario, it offers a more robust and secure data model that allows for better long-term scalability. Implementation and cost concerns are also discussed in the conclusion to elaborate on trends and considerations in choosing the appropriate Serialization scenario to pursue. / by Aisha Nabiyeva and David Z.Y. Wu. / M. Eng. in Supply Chain Management