Source: TEXAS A&M UNIVERSITY submitted to NRP
VALUE OF TIME AND TEMPERATURE HISTORY RFID TECHNOLOGY TO LEVERAGE RETURN ON TRACEABILITY INVESTMENTS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
0224569
Grant No.
2011-67023-30102
Cumulative Award Amt.
(N/A)
Proposal No.
2010-04810
Multistate No.
(N/A)
Project Start Date
Jan 15, 2011
Project End Date
Aug 31, 2013
Grant Year
2011
Program Code
[A1611]- Foundational Program: Economics of Markets and Development
Recipient Organization
TEXAS A&M UNIVERSITY
750 AGRONOMY RD STE 2701
COLLEGE STATION,TX 77843-0001
Performing Department
Agri Economics
Non Technical Summary
A threat to the competitiveness of the perishable produce industry is how to maintain product availability without spoilage. Waste is growing and costs an average supermarket approximately $450,000 per year, with perishables accounting for 56% of total store shrink. Spoilage throughout the supply chain is estimated to be as much as 10% of all perishable goods. Most perishable items like greens, meats, or dairy, have non-deterministic, random lifetimes. These random lifetimes depend on the time it takes for products to flow through the supply chain, as well as the product's temperature history and other environmental factors. These factors are generally unknown and highly variable, leading to considerable uncertainty with regard to the timing of product expiration and setting an appropriate sell-by date. The condition of many perishables cannot be ascertained by simple visual inspection. Thus, there is the danger that the product has expired before the sell-by date, and an expired product may be sold to the customer. The results of selling an expired product may range from the cost of a product return and loss of customer goodwill, to health and safety costs, public loss of reputation, and even litigation costs in case of a large-scale health scare. These complications and risks associated with perishable grocery products make effective inventory management challenging. In essence, the inventory control policy needs to balance the cost associated with waste (from throwing away good product), and the cost associated with selling expired product. This research fills the need for a developed inventory control model to demonstrate economic value of information contained in RFID (radio frequency identification) chips, a novel technology increasingly being adopted in the food industry. The outcome will be a mathematical model identifying optimal re-order policy and costs or benefits of acquiring additional information on time and temperature history through RFID tags. The expected impacts include enhanced efficiency in operations of the food supply chain, through the in-depth understanding of the value proposition of information in the hands of primary producers, logistics managers, or food retailers.
Animal Health Component
90%
Research Effort Categories
Basic
(N/A)
Applied
90%
Developmental
10%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
6025010301085%
6045010301015%
Goals / Objectives
Goals and Objectives: The long-term goal is to extend the fundamental knowledge on the value of information that stems from technological innovation in the food products supply chain. This goal will be accomplished by completing the following objectives: (1) to identify optimal stocking levels and sell-by dates for perishable products having random lifetimes; (2) to analytically and mathematically determine the economic value of having data from environmental condition sensors in RFID (radio-frequency identification) tags ; and (3) to examine the value to primary producers of information systems that integrate time and temperature history data with information about production processes in accordance with international standards for food safety. Expected outputs include: (1) an inventory control system to utilize the time and temperature history data; (2) insights and models that are immediately applicable by companies in the perishable food products industry; and (3) improvements in foundational economic knowledge relevant to the competitiveness of businesses in the farm-to-consumer value chain. These models and insights will also provide societal benefits by enabling companies and governmental authorities to reduce public exposure to health and safety risks that arise from perishable product spoilage in supply chains. While our main focus is on perishables, specifically dairy products, the findings are expected to generalize to a variety of perishable products in which quality is significantly affected by time, temperature, and other environmental conditions in the supply chain. Some agribusiness sectors may not be well-represented by the model. For example, the total cost of inventory in some product lines and business outlets may not justify technology adoption even if the RFID - time and temperature history management system is optimal from a cost-minimization point of view. Nevertheless, the relevance of our results for operations research and for managerial applications is compelling because the foundational knowledge in stochastic inventory control has widespread applications in the food and agribusiness sectors. This project will produce a firm-level model and as such, cannot reflect all the ways in which performance depends on exogenous conditions in the market. Demand is parameterized rather than linked with a working price-dependent demand relationship or a game-theoretic structure. This specification is appropriate for the project's cost orientation and will fulfill our intent to provide decision support for companies choosing to invest in RFID or other technologies to enable traceability goals. Timeline: The period of performance for this research is January 15, 2011, through August 15, 2012. The research team members will work collaboratively with regular bi-monthly meetings and concurrent effort during summers. The project findings will be presented at the INFORMS annual meeting (mid November 2011, in Charlotte, NC). The entire project is expected to be completed in 19 months (early 2011 - August 15, 2012).
Project Methods
The researchers will create a series of mathematical models of the perishable inventory control problem for the case when products have random lifetimes, a given sell-by date, and product expiration is not observable. The base model represents the case when there is no time and temperature history data available. The first task is to complete the base model with representative data for fresh produce and a manufactured dairy food. The base mathematical model represents a supplier that sells a single perishable product. The product lifetime is random, and possesses a maximum lifetime of M periods. The replenishment problem is an infinite-horizon stochastic dynamic program (MDP) formulation where the objective is to find the supplier's allowable shelf life (T) (sell-by date) and optimal reorder policy (q) so that its average per period expected cost is minimized. In this model, T is a decision variable, therefore it is possible for expired product to be sold and unexpired products to be discarded. If an expired unit is sold, a hazard cost k is incurred. Any units remaining in inventory at the end of T periods, expired or not, are discarded (outdated) at a cost per unit c. Next, we will incorporate time and temperature history data into the model. This step necessitates linking temperatures to spoilage rates for the product. In the scientific literature, there are several models, most predicated on the Arrhenius law. The parameters for this step will be evaluated and validated through industry contacts. In the third task, we will complete the analytical representation of time and temperature history information along with its value in terms of costs avoided. The project intersects several research streams that include value of information, perishable inventory, cold chain management, food safety, temperature monitoring, RFID, and shelf life prediction. These streams cut across a wide array of disciplines that include agribusiness, computing technology, industrial engineering, technology management, food science, microbiology, and horticulture. From this perspective, our study represents a multidisciplinary contribution to the research state of the art that uniquely ties together multiple fields of research. It is expected that the analytical dynamic programming model will be too large to practically implement. Therefore, emphasis will be placed on developing well-performing heuristic policies. Simulation studies will be conducted, based on real world products and supply chains. The results will provide an assessment of the value of information and the conditions in which time and temperature history information is most valuable.

Progress 01/15/11 to 08/31/13

Outputs
Target Audience: Audiences for the research during the entire project duration includes researchers in the fields of business logistics (supply chain management)and practitioners infood distribution companies. For this reporting period (Jan.-Aug. 2013), the investigators were focused primarily on the academic audiences, readers of academic peer-reviewed journals where the publications are being reviewed and have been revised. The peer - review process at a highly-regarded management journal (Management Science) was extensive but ultimately unsuccessful. After correction of certain modeling issues, revision of the paper's rationale and motivation, and evaluation by the editor, the articles are now under consideration at a different journal which is likewise aimed at the academic peers who have an interest in the specification and methodologies employed in the modeling. A secondary target audience is reached by the efforts of the graduate student under the direction ofDr. Salin during in this reporting period. This target audience holdsmanagement positions and government oversight roles at the ministry of agriculture in a developing country. A more simple inventory model focusing on pricing strategies was completed during the final year of the project as part of the student's masters thesis. This supply chain model inventory and pricing model is a simple simulation of fluid milk distribution where the economic decision is toset a price, given unsold supply and time to expiration. The work was completed and presented to the student's graduate committee, and has been made available through the Texas A&M universitylibrary, accessibleto the entire university community. Changes/Problems: There were no major changes in the approach to the research. One of the Co-PIs had a change in appointment as follows: Gary Gaukler, faculty member at Peter F. Drucker and Masatoshi Ito Graduate School of Management, Claremont Graduate University, Claremont, California. Email: gary.gaukler@cgu.edu What opportunities for training and professional development has the project provided? A masters student received specialized knowledge in economic analysis and inventory control and it is expected that he will bring this experience to his current position in his home country of Iraq. His research was not directly funded by the project but there was extensive cross-collaboration through the role of the project PI as chair of his thesis committee. In association with this research, the PI presented at a food industry forum which has expandedher cross-disciplinary knowledge in food sciences and food preservation. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? 1. Achievement of decision frameworkfor setting expiration. We disentangled the managerial issue of setting expiration dates on perishable goods from the value associated with the full-information of a technology such as an RFID system that is working perfectly. This is important to the discipline of stochastic inventory control. The findings confirm that there isno direct analytical solution tothe problem of setting time to expiration, when the critierion is cost effectiveness of inventory management. Instead, heuristics are derived based on simulations. 2. Construction ofrepresentative solutions in a realistic context. The project outputs includeexample simulations for(1) fluid milk, which has arelatively short cold chain managed by retail firms; (2)a global supply chain for tomatoes; and (3) a fresh product with a very short lifetime (fish).The specific value propositions were found to be highly context dependent on the physical properties of the growth of spoilage organisms. Each of these specific examples was modeled with parameters obtained from published literature or obtained from industry sources. We find thatthe total cost of inventory in some product lines and business outlets may not justify technology adoption even if the RFID - time and temperature history management system is optimal undera cost-minimization objective. The results from the project intersect several research streams that include value of information, perishable inventory, cold chain management, food safety, temperature monitoring, RFID, and shelf life prediction. These streams cut across a wide array of disciplines that include agribusiness, computing technology, industrial engineering, technology management, food science, microbiology, and horticulture. From this perspective, our study represents a multidisciplinary contribution to the research state of the art that uniquely ties together multiple fields of research.

Publications

  • Type: Theses/Dissertations Status: Published Year Published: 2013 Citation: Ayoub, Wisam Hanna. Inventory Management of Perishable Goods Under Demand Variability, Masters thesis, Texas A&M University, College Station, Texas.


Progress 01/15/12 to 08/31/12

Outputs
OUTPUTS: The objectives of this project are: (1) to identify optimal stocking levels and sell-by dates for perishable products having random lifetimes; (2) to analytically and mathematically determine the economic value of having data from environmental condition sensors in RFID (radio-frequency identification) tags ; and (3) to examine the value to primary producers of information systems that integrate time and temperature history data with information about production processes in accordance with international standards for food safety. With regard to objective (1), a general model structure has been developed, along with examples for tomatoes, fish, and fluid milk. The tomato and fish cases have been fully developed and outcomes simulated across a reasonable range of realistic world situations. The inventory model was completed, a heuristic model was developed for quantitative analysis, and simulations conducted. Articles were prepared in format for academic journals. One article was submitted to Management Science: Operations Management. The paper entitled Determining Expiration Dates was reviewed and re-submitted to the journal editor on Oct. 10, 2012. Objectives (2) and (3) relate to the use of technology that records condition information as inputs into a decision about remaining shelf-life. The full-information model has been developed and is the final stages of preparation for submission to an academic journal. A third inventory model focusing on pricing strategies has been developed and is in the preliminary stages. This inventory and pricing model is a simple simulation of fluid milk distribution where the pricing options are the focus. The work is part of a masters-level graduate student research program at Texas A&M (funded by sources other than USDA-AFRI). The expected output of the masters thesis project will contain less detail on tracking the aging of the good than the previously mentioned research products from this study. This limitation is not a major concern in application to fresh milk distribution in the USA, because the distribution system is very short and perishing is rare. However, with fresh produce and other goods with less predictable / controllable supply, the limitation of this spreadsheet model may be of sufficient concern to merit an alternative approach. It is expected that the fluid milk model parameters will be appropriate to model a pasteurized fresh milk supply chain (calibrated to conditions in the USA) to be contrasted with a supply chain for UHT (ultra-high-temperature treated) milk. The latter is the dominant milk product in many developing and emerging economy markets. Thus, our expected results will have relevance to the development of cold chains in markets worldwide, as consumers transition away from raw milk and its attendant food safety risks. PARTICIPANTS: Individuals working on the project include 3 investigators: Salin, Gaukler, and Ketzenberg. Salin is supervising Wisam Hanna, a graduate student who is working on issues related to this project although he receives no financial support from AFRI. TARGET AUDIENCES: Preliminary findings were disseminated to an industry audience at the annual convention of the International Association of Refrigerated Warehouses and the World Food Logistics Organization on April 23, 2012. The presentation led to followup interaction and useful discussions with industry audiences. PROJECT MODIFICATIONS: A no-cost extension of the project was approved August 15, 2012.

Impacts
In the course of thorough peer review at the journal, the treatment of perishability states in the model was critiqued and as a result, the research was improved in terms of the treatment of perishability in the model. Specifically, the perished (or not) state of the product is no longer part of the state space in the dynamic model. This refinement corrects the confusion created by having an observable but not actionable variable on the state space. The revised model is of a retailer who must issue on a FIFO (first-in, first-out) basis - one of our key modeling assumptions. The reviewers approved of the use of food science modeling of perishability, which indicates achievement of the project's purpose of integrating business operations research with food science practice. The two cases that we examine, fish and tomatoes, have important differences in the relationship of distribution costs to perishability. Differences arise because the shelf-life of tomatoes is negatively affected by both too-low temperature and too-high temperature, but fish are only adversely affected at too-high temperatures. The impact of the project in terms of an innovative contribution to the inventory control literature is that units of inventory of the same age class may perish at different times. Further, the research provides a heuristic policy to determine the timing and quantity of orders, using a power approximation which is an improvement over the grid search method. Using a validation data set of 486 different experiments, our heuristic policy identifies the optimal expiration policy in all but one case. Thus, the outcomes during this period constitute notable improvements in foundational knowledge in stochastic inventory control appropriate for food retailing. The insights from the paper that reviewers appreciated include the finding that small changes in ambient temperature can have a significant effect on shelf life and thus retail expiration dates. A 3 degree increase in average ambient temperature upon catch of the fish leads to a one day shorter expiration date in our study. Even more dramatically, a 16 hour increase in the length of the ambient supply chain phase results in expiration dates at retail that are three days shorter. We observe that delays in the temperature-controlled phase have a smaller impact on costs and expiration dates than delays in the ambient stage, but distribution bottlenecks, even with temperature control, can nonetheless increase inventory costs substantially because the product is in the controlled stage relatively longer than in the ambient stage. The finding can be shared with primary producers to stress their role in end-to-end control of the distribution systems and is a necessary pre-requisite to contracts that share value from retail or processing operations to primary producers.

Publications

  • No publications reported this period


Progress 01/15/11 to 01/14/12

Outputs
OUTPUTS: The outputs from the project during the first year include: A developed inventory control system that utilizes information on the time and temperature history of perishable foods in two real world supply chains (tomato and fish). This accomplishes objective (1). The forms of the model are (1) an infinite-horizon stochastic program model; (2) a heuristic policy validated against the infinite horizon model; (3) a set of simulations that operationalize the policy. Further, a second project output is a detailed analysis of the performance of the inventory management system accomplished with simulation studies. The inventory model and related simulations are outputs contributing to foundational economic knowledge relevant to the competitiveness of the food industry. As such, they are part of the achievement of objective (3). Dissemination of the results is in the preliminary stages. There have been ongoing communications with industry and scientific experts regarding the development of the models and data inputs. Further, two journal articles have been drafted in preparation for dissemination, entitled: (1) Determining Expiration Dates for Perishables; and (2) Dynamic Expiration Dates for Perishables. PARTICIPANTS: Individuals working on the project include 3 investigators: Salin, Gaukler, and Ketzenberg. Informal collaborators on the Scientific Advisory Council of the World Food Logistics Organization (WFLO) contributed background information. Further, consultations occurred with attendants at the Produce Marketing Association Conference and the WFLO convention in April 2011; the Food Logistics Conference in June 2011; and Dematic Material Handling Conference in September 2011. TARGET AUDIENCES: Nothing significant to report during this reporting period. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
The results from the study are the first to adopt a cost-minimization objective and identify the impact of product spoilage and the cost of failing to meet consumer demands when shelf-life is specified with a science-based probability distribution. The retail decision-maker manages inventory facing randomness in shelf-life of a form allowing for uneven deterioration with respect to time. That is, older product might have a longer shelf life than newly arriving goods that experienced abuse in the distribution chain. The techniques of a discrete stochastic specification of consumer demand extends the operations research methodologies to the food inventory management field. Further, distinct levels of information available to the decisionmaker, analyzed separately to estimate the value of information, reflect the importance of technological progress in Radio-Frequency Identification (RFID) and Wireless Sensor Networks (WSN) in global food supply chain management. By the fulfillment of our intent to provide decision support for companies considering investments in RFID or other technologies to enable traceability goals, the managerial goals of the project are nearing achievement. The immediate impact of the knowledge about inventory management systems is to extend the field of operations research through integration of the scientific knowledge base in food science. The ultimate outcome will inform managerial decisions in the food retailing and wholesaling businesses. The foundational knowledge in stochastic inventory control has widespread potential applications in the food and agribusiness sectors, and more so if the value of full information is large as the preliminary findings indicate. Ultimately, society will benefit from the cost efficiency that will result from improvements in the management of the risks of perishable product spoilage. Resources were used for salary support for the investigators during the first project year. Activities undertaken include interactions with industry, modeling, writing, and planning for formal academic peer review. Members of the research team attended 4 industry conferences in 2011 and obtained feedback on shelf life studies that form the basis of the outcomes. Only one of the conference trips was funded with project resources; the others were part of industry-sponsored activities by the investigators. The project travel budget will be allocated to an academic meeting in the operations research - inventory management specialty field for high-level peer review in early 2012.

Publications

  • No publications reported this period