City Logistics Research: A Transatlantic Perspective (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

9Demand Patterns and Trends Christopher Caplice, Center for Transportation and Logistics, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA Jean-Louis Routhier, Laboratoire d’Economie des Transports de l’Institut des Sciences de l’Homme, Lyon, France Miguel Jaller, Center for Infrastructure, Transportation, and the Environment, Rensselaer Polytechnic Institute, Troy, New York, USA Robert Chumley, Retail–Business Innovation, 7-Eleven, Inc., Dallas, Texas, USA Laetitia Dablanc, French Institute of Science and Technology for Transport, Development, and Networks, Paris, France IntroductIon Christopher Caplice Chris Caplice opened this session by explaining that “demand” means different things to different people. To shippers, demand means products; to freight trans- porters, demand means the demand for trucks to get the product into the store or to the final consumer. The views are correlated, but they are not the same. Policies will dictate freight demand but not product demand. The first two presentations of this session focused on freight demand generation from a researcher perspective: that is, how to estimate and forecast demand and what drives demand. Jean-Louis Routhier presented freight survey results from Europe, and Miguel Jaller presented results from the United States. The third presentation, by Robert Chumley of 7-Eleven, provided the private-sector perspective and how urban product demand is changing. A discussion followed each of the two parts of the ses- sion, and Laetitia Dablanc concluded with a synthesis and summary of the session. French cItIes’ urbAn FreIght surveys Jean-Louis Routhier Jean-Louis Routhier described a set of full-scale urban freight surveys (UFSs) of French cities. Funding for the studies came from the French Ministry of Transport as well as from each city surveyed. The aim of these UFSs was to build a model (the FRETURB model) to simulate the existing urban freight situation in these cities. The first set of surveys was conducted from 1994 to 1996 in three cities of different sizes (Bordeaux, Marseilles, and Dijon); the second set is in progress now (2011 to 2014) in Paris and Bordeaux. The aim of these surveys is twofold: on the one hand, they provide understanding of the behaviors and orga- nizational aspects of urban pickups and deliveries (i.e., they simulate the existing situation). On the other hand, they feed into a tool to make diagnoses of urban logistics without the need for collecting large amounts of data, thus reducing costs. The surveys are a decision aid and can be used for short-, medium-, and long-term forecasts. Survey Description Routhier next described the details of how the UFS worked. The first task was to find a relevant unit of observation. Could it be the commodity being moved, the vehicle on the road, or the transport company? In fact, it was none of these. The best unit was the delivery (or pickup) operation serving an establishment, using a vehicle. At that point it is possible to observe (and survey) the formation of vehicle flows and their impact on the urban environment accord- ing to the transport system, the logistic strategy of the firms, and the establishment’s environment.

10 city logistics research: a transatlantic perspective A UFS consists of three complementary, nested surveys: 1. A survey of business establishments that describes the activity and size of the establishment and identifies all the freight delivery operations in relation to the char- acteristics of establishments and their environment; 2. A survey of truck drivers that describes the trips and the conditions of carrying out deliveries and pick- ups; and 3. A survey of truck companies that describes the logistics organization of for-hire trucking companies. The three surveys worked together. The establishment survey included site visits and reviewed log books for a description of pickups and deliveries. The driver survey included driver logs and Global Positioning System (GPS) data to identify routes and stops. In the carrier survey, it was important that the drivers surveyed were the same drivers who delivered to the establishments surveyed. The sampling frame was the comprehensive register of establishments by activity category. The establishments sample comprised 1,500 establishment questionnaires cov- ering 6,000 deliveries and pickups over 1 week and 8,000 product types (by weight, packaging, and so forth). The trucker sample was composed of 1,000 driver question- naires (6,000 deliveries and pickups), and the carrier sample included 100 major truck companies and large wholesalers. Forty-five activity categories were used, such as “phar- macies,” “bookshops,” or “hardware stores.” Each cat- egory was considered as a homogeneous group of traffic generators and logistic organizations. Some categories were more important than others. For example, “cafés, hotels, and restaurants” represented nearly 5% of total operations (five times more than large grocery stores and twice as much as warehouses). The survey was costly (€1.2 million) and complex due to difficulties in contacting the business owners and poorly motivated respondents. As a result, get- ting responses required a preliminary information and promotion campaign, employing a contactor to ensure recruitment of experienced pollsters, and paying incen- tives for each survey completed. The large size of the sample of establishments and the quality of the responses (thanks to face-to-face inter- views) guaranteed a good estimation of the number of movements generated by each category. The stratifica- tion of the activities was sufficiently detailed to obtain an accurate estimation of freight and freight trip generation. Survey Results The main result of the surveys was a detailed description of the current situation: standard data and indicators. The researchers found consistent and stable relationships and used them to feed a traffic generation model. These sta- ble relationships included the following: one delivery and pickup each week, per job; 75% of deliveries and pickups were carried out by rounds; 80% of deliveries took less than 10 minutes, but pickups lasted 30 minutes on aver- age; and peak hours were 9:00 to 11:00 a.m., which differs from car traffic peak hours of 7:00 to 9:00 a.m. Routhier said that by expanding the driver survey, it is possible to estimate truck traffic on the city roads and to estimate the flows of vehicles within and through the different zones of a city. The surveys also found consistent and stable relation- ships between cities, such as between activity, work- force, and movement generation; between stop duration and size of delivery round; and between distance covered between stops and the overall size of the delivery round. These relationships, observed in all city surveys (small and large cities) were quite similar. All relationships were translated as equations in the FRETURB model. Because the relationships are not city specific, the model is efficient and robust even if it is implemented in cities without surveys. FRETURB Model Routhier explained that inputs for the FRETURB model are the database of the existing establishments in a given city and the city spatial zoning. Therefore, it is easy to feed the model with these inputs. The model consists of four modules. Module 1 estimates the number of deliver- ies and pickups. Module 2 calculates road occupancy due to delivery stops. Module 3 simulates road occupancy by running vehicles. Module 4 distributes the results over 24 hours. FRETURB is a compromise between the simplifica- tions needed to be able to model urban complexity and a comprehensive description of the reality. Suggestions for U.S.-EU Collaborative Research Routhier offered three suggestions for U.S.-EU collab- orative research: 1. To compare urban freight objectives strategies of cities in the United States and the EU (do they really want to know freight demand?); 2. To compare the different methods of data-based modeling oriented toward helping make public policy decisions; and 3. To test the operational applicability of models like FRETURB in the United States and vice versa, which would require comparing data sources and collection processes as well as different establishments (which are

11demand patterns and trends inputs of the model) and choosing a land use data area to test the implementation. u.s. cItIes’ urbAn FreIght surveys Miguel Jaller As a U.S. counterpart to Routhier’s talk, Miguel Jaller’s presentation focused on U.S. cities’ urban freight surveys (UFSs). He began by explaining that development of freight demand models is difficult due to lack of knowl- edge, models, and data. Researchers are still in the pro- cess of understanding freight at different levels. There are not enough data, and there are multiple models that provide different outputs. The freight system is complex, with multiple agents (shippers, receivers, carriers, third- party logistics providers, consumers, freight forwarders), all of whom have their own requirements and impose their own constraints on the system. Each agent has only a partial view of the system, and there are multiple inter- actions and links between them. There are many metrics to measure freight, and a variety of functions (long haul, consolidation) can be performed using different delivery modes and vehicles (bikes, vans, truck, rail, barge). The freight system can also be viewed at many levels of geog- raphy: neighborhood, state, region, and so forth. Because the numerous agents have only a partial view of the system, no single agent can provide a complete picture of it, which would involve knowing a host of metrics, such as amount of cargo, number of loaded vehicle trips, number of empty vehicle trips, number and frequency of deliveries, commodity type, shipment size, cargo value, and land use patterns. Jaller displayed a diagram of the multiplicity of met- rics, noting that the flow of freight of trucks (vehicle traf- fic) differs from the flow of the goods (commodity flow). Data Needs and Sources Jaller listed a dozen techniques for modeling three cat- egories of flow units for urban freight models that focus on either trip interchanges, tour-based models, or both. The point, he said, is that there is debate about which modeling focus is best. Jaller also described the different data required by different modeling techniques. He showed a table with seven data categories (freight generation, delivery tours, agent economic characteristics, agent spatial distribu- tion, network characteristics, special-purpose models, and other economic data) on one axis and six modeling techniques on the other axis. From the table, researchers could see that some of the models require just a little data, while others require a lot. Therefore, if researchers lack the resources to gather a lot of data or do not have the needed level of data, the quality of the model will suffer. As examples, Jaller discussed primary data sources (Commodity Flow Survey [CFS] data, zip code business patterns, surveys, interviews, and travel diaries), as well as secondary sources (Global Positioning System [GPS] data and experts). Jaller focused the discussion on the topic of freight demand synthesis. This technique, which can help fill gaps in the data with good estimates, can reduce data collection costs but may introduce an error. Jaller listed the data gaps he identified. The gaps occur because some of the information is not publicly available or because it concentrates on a certain location, and it is not clear if the data translate to other locations. He concluded that most of the data needed must be collected from scratch. Data collection methods vary widely in cost and response rates, and collection techniques depend on the sampling frame. Data collection methods can focus on origin–destination, en route intercepts, or various loca- tions along the supply chain. GPS data are also helpful, but are limited in that they only provide speed, time, and location; they cannot provide data on trip purpose, com- modity type, or shipment size. GPS data should be con- sidered complementary to more traditional freight data collection procedures, but they do not provide the full picture that can be derived from surveys. Furthermore, commercially available GPS data can be biased and dif- ficult to convert into a representative sample. Generation of Demand and Generation of Traffic Next, Jaller moved to the topic of freight demand gen- eration and noted that there are two perspectives: gen- eration of demand (FG) and generation of traffic (FTG). FG is an economic manifestation of the production– consumption processes, and FTG is the result of logisti- cal decisions. Reviewing more than 60 reports that contain FG and FTG references, Jaller found more than 150 case studies and many different models discussed. Many models were based only on a handful of observations, which shows the need to develop better models and collect freight data. The references discussed many more FTG models than FG ones. Jaller mentioned several issues to be considered in FG- FTG modeling. First, it is important to pay attention to the classification system, whether it is economic based (e.g., North American Industry Classification System [NAICS] or Standard Industrial Classification [SIC]) or land use based (e.g., Standard Land Use Coding Manual [SLUCM] or Land Based Classification Standards [LBCS]). Other issues include the level of aggregation, aggregation proce- dures, and the modeling technique used.

12 city logistics research: a transatlantic perspective In numerous case studies, FTG is usually constant regardless of the size of the establishment and is not pro- portional to the number of employees per establishment. It appears that larger establishments received larger shipments in larger trucks, but the number of deliveries tends to remain constant. Researchers must be cautious when using employment as the only independent vari- able because it will lead to errors. Next, Jaller described the advantages of using CFS data. CFS has more than 4.3 million records of ship- ments. CFS microdata can be used to estimate FG models as a function of establishment characteristics, as well as models at different levels of geography or industry segment. Other data sets that can be used are the Census of Manufacturers (CMF), the Longi- tudinal Business Database (LBD), and the Standard Statistical Establishment List (SSEL). He mentioned that, for the first time, a research team at Rensselaer Polytechnic Institute was granted access to the micro- data for a period of 5 years to conduct freight demand modeling. Jaller summarized the main conclusions of his pre- sentation, namely that freight is a complex system and researchers need to collect data from all economic agents. Data collection methodologies provide different types of information and levels of detail, and GPS data cannot provide all the information required. Freight demand modeling techniques may require different data, and a combination of data collection methods is required. Finally, freight demand models must distinguish between the total amount of freight generated and the number of freight trips required to move it. Recommendations for Research Jaller ended with recommendations for four research directions: 1. Develop innovative data collection methodologies and technologies, 2. Improve freight data synthesis, 3. Develop complementary models to take advantage of the CFS, and 4. Take advantage of administrative records. QuestIons And AnsWers WIth JeAn-louIs routhIer And MIguel JAller Caplice reiterated that Jaller’s presentation was a sur- vey of different studies, while Routhier’s was a deep dive into one survey that led to the FRETURB model to help cities make forecasts. Caplice posed the first ques- tion of the discussion by asking Routhier how appli- cable the FRETURB model is to other cities, both in the EU as well as in the United States. Routhier said that the model is applicable to other cities because cities do not need to collect as much data to implement it. They need to know the location of establishments in the city, but such data are easy to obtain; thus, implementation of the model in other cities is possible. The problem of applicability to the United States, however, arises from the fact that cities in the United States are larger and less dense, which means that logistics operations in U.S. cities are not the same as in Europe. The FRETURB model has been implemented with success in Switzerland, Belgium, Spain, and Italy. It is possible to use FRETURB effectively in European cities, but he was not sure if it would work in the United States. Jaller noted research was needed on the hypothesis that operations in Europe are the same as in the United States. There have been more restrictions in the EU on operations in city centers, such as low-emission zones, so different establishments have adapted to these con- straints. These adaptations have affected the number of deliveries that establishments get each week. In addition, in the densest city areas, establishments hold less inven- tory due to the higher price of land. So, if cities are more dense in the EU, they would hold less inventory; in less dense areas, establishments can have larger stock rooms. Finally, density affects the whole supply chain because of the implications for vehicle capacity. Carriers have to use small vans rather than large trucks when entering city centers, which alters FTG models because two smaller vans might go out rather than one big truck. Anne V. Goodchild noted that Routhier’s model col- lected a lot of data, but it is still based on using averages by classification. For a certain classification, the model uses one value to represent trip generation. Would it be possible to use a range of values rather than a single value? Routhier answered that about 6,000 establishments were surveyed; thus, it was possible to build stratification categories of activity that are relatively homogenous by their logistics activity. To the extent that demand is the same in one city as another, it is possible to generalize. Having 45 types of activity makes the model robust. Per- haps there could be 50 or 60 types of activities, but even if there is some dissimilarity within a category, that dis- similarity is less than the difference between categories. Caplice asked Jaller to comment on Goodchild’s ques- tion, because Jaller used a single value. Jaller replied that he used multiclassification analysis and models for the entire pool of employment and found differences in the trips generated in different levels. The number of freight trips may start increasing as employ-

13demand patterns and trends ment increases, then it drops, and then it starts increasing again. The drop occurs as the carrier moves to a larger vehicle. Jaller said he did not find differences among industries in different regions. Kazuya Kawamura asked whether, when using the FRETURB model in cities where surveys have not been conducted, such as Spain, small-scale surveys were con- ducted to see if freight was similar, and if validation sur- veys were done to adjust the model. Routhier answered that they have not done additional surveys due to lack of funds. The FRETURB model is intended to help clients, who in this case are the public authorities who receive the FRETURB results. Clients have been happy with the results. The model is consis- tent with local observations. The FRETURB model is a comprehensive description of logistics in an urban area, and that is what public authorities need. Barbara Lenz noted that urban stores have little room for storage and asked if the models take into account the storage strategies used by stores. She added that in Germany, different stores have different storage strate- gies and that some pharmacies have almost no storage space. Jaller answered that although the database contains many models, forecasting for grocery stores in the city center that get deliveries is not possible because there is not enough information. In the United States, business- pattern data on a countywide level and often down to a microlevel of individual land parcels are available. One can use models from New York and collect a small sam- ple to apply it to a different area. Ken Button voiced concern over big, complicated mod- els. He told the story of modeling for Bay Area Rapid Transit (BART) in San Francisco, which used the same type of data to which Jaller referred. The modelers fore- casted that 15% of residents would switch from driv- ing their cars to using BART. A different model simply surveyed 400 residents and forecasted that only 5.6% would switch. The eventual outcome was that 5.5% of residents switched to using BART once it was built. The second forecaster did not collect tons of data; he used a simple model and simple data. Sometimes the problem is taking the view of a systems engineer. The people who run companies are humans and make decisions as humans. So perhaps instead of complicated models, researchers ought to look at how a small number of operators work, the incentives of the business, structure of the business, and how much emphasis is put on efficiency versus inventory, and use that information to build a model. Button ended with an example from the Netherlands regarding solving deliv- ery problems to the Schiphol airport. The model Jaller proposed would not have helped the airport decide how much to deliver by truck compared with attempting a completely new solution: shifting the freight to barge. Jaller replied that it is not possible to go to a munici- pal planning organization (MPO) and simply say, “trust me.” MPOs want to see numbers. True, there have been mistakes. For example, a few years ago it was common practice to ban car traffic by plate number, but now that practice is known to be ineffective. Nonetheless, mod- eling is useful and can drive behavior change. Before, carriers had the power to do what they wanted, but now consumers want trucks to be green, so that drives behav- ior. If locals do not want large trucks operating at night, that will influence behavior, as well. The question is how to identify the best objectives. Routhier responded to the point about BART by saying that passengers have two main alternatives: car or public transport, perhaps bike. But companies have hundreds of thousands of variables like different types of vehicles, storage, and so forth. These factors were not taken into account until the French surveys. It is necessary to model only small samples of such activity, but it is necessary to have a comprehensiveness of activity and to make stable categories. Forty-five categories is not complex; it is a reduction of the complexity, and it is a compromise given the money available to carry out the surveys. Alessandro Damiani asked Jaller whether his analysis also covered service trips, such as utilities or electric meter readings. Jaller replied that most analyses do not consider ser- vice trips, but that the Phase 2 survey will include it. Freight delivery vehicles are a big issue in Manhattan. They are considered commercial vehicles and share the same space as service vehicles, but they have different purposes and structures. Nonetheless, they are a big pro- portion of the vehicles coming into the area. Rosário Macário commented that freight traffic differs from passenger traffic: it is not a public service, so the city will not pay for data collection on it. Another chal- lenge to data collection is that the stakeholders in this process (authorities and private entrepreneurs) want the status quo. Authorities do not want liabilities and entre- preneurs do not want regulation, so who will pay for the research? She suggested the need for a more pragmatic, business-oriented approach that lets entrepreneurs cre- ate business models. She asked what the capacity of the models was to generate data. Jaller replied that the models have been used in dif- ferent cities in the United States and validated via small samples. What is needed more than data collection is iden- tification of the main drivers of economic activity; models could be developed from those drivers. He has used mod- els to generate traffic flows and behavioral microsimula-

14 city logistics research: a transatlantic perspective tions that are useful, and they generate additional data. His research focuses on how to best use the data. The pri- vate sector will provide researchers with data if confiden- tiality is protected. Companies have incentives to improve their data, Jaller said. In New York City, many companies pay $1,000 a month per truck in parking tickets. If com- panies see a strategy that lets them reduce two delivery tours to one, they have incentive to do it because it will reduce costs. Given their close relationship with society, companies are also motivated to be at the forefront of being sustainable. Routhier said that the FRETURB model is oriented to public decision making. Some freight companies use the model because they can change the initial situation and then simulate different types of behavior to see the result of those changes. Routhier is developing a policy tool on the basis of this model to test different scenarios for decision making. shIFts And trends In urbAn retAIlIng And buyIng behAvIor Robert Chumley Robert Chumley offered a private-sector perspective on urban retailing and buying behavior. He remarked that he was puzzled when he was asked to present at this sym- posium, because he is neither an academic nor a logisti- cian (his title is Vice President of Innovation at 7-Eleven). But then he realized that academics need inputs that are others’ outputs. They need to understand the decisions being made inside private-sector organizations to use as inputs to their models. Chumley began by describing the impact of urban- ization and how it is affecting consumers, retailers, and logistics networks. He noted that, according to the United Nations, for the first time in human history, urban dwellers outnumber rural residents worldwide. By the year 2025, estimates are that there will be 4.6 billion urban dwellers. In the United States, more than 80% of the population resides in urbanized areas. Chumley hypothesized that new consumers are mov- ing into cities, and they behave differently and have new demands. Those demands, in turn, drive retailers to create new retail formats and products which, in turn, drive changes in logistics networks, assets, and destina- tions. Indeed, destinations are increasingly more dis- persed and see lower volumes. Recounting the history of retail, Chumley identified the trends from producer driven (1850s to 1950s), to distributor driven (1960s to 2000), to customer driven (2000 to today). The future will be “anytime, anywhere,” with the consumer firmly and irreversibly in place as the driver of retail activities. The New Urban Consumer Chumley next described the new urban consumer. Older empty nesters are moving into cities from the suburbs, but so are young professionals who are making their first homes in the city, attracted by jobs of the new economy and reasonable housing prices. Cities are characterized by ethnic diversity and an on-the-go culture that encour- ages new lifestyle choices, such as shedding cars in favor of public transport. Such change brings new mobility constraints as well as opportunities. The shopping behavior of urban consumers differs from that of suburbanites. Consumers have cast off their suburban extra refrigerators and freezers and weekly trips to Costco. Instead, in urban environments, people shop much more frequently, choosing local shops and more fresh and organic foods. Having cast off their cars, they rely more on walking and smaller package sizes. Given the higher ethnic diversity in urban areas and the new cultural influences, urban consumers are more inclined to be curious about ethnic food items and new cooking methods. Changes in Urban Retailing Urban retailing is likewise going through profound changes in response to the new consumers. As retailers move into urban spaces, they often have to occupy much smaller footprints. The most significant implication of these smaller footprints is that retailers are considerably limited in terms of the inventory they are able to carry on the sales floor and in the back room. As a result, urban retail shops have increasingly complex supply or demand chains operating within restrictive local ordinances, restricted access, and cumbersome product movement environments. Moreover, most urban retail locations are leased rather than purchased, which affects productiv- ity and profitability. In rapidly growing urban areas, the balance of power has shifted to landlords, who often set terms strongly in their own favor. Finally, urban retail locations require human capital to run. It can be chal- lenging to find qualified talent willing to work for these relatively low-paying (often part-time) jobs in urban settings. Many businesses are adapting their business models to run on less human capital than their suburban counterparts. As a result of new consumers and new urban locations, retailers have to find new store formats and new business models and deal with multiple end points as logistics net- works become more fractured, Chumley said. With smaller-footprint stores, restrictive ordinances, and rapidly evolving consumer purchase patterns, many retailers are developing new store formats. For example, Best Buy is moving into major transit hubs, and one

15demand patterns and trends Starbucks is located in the vault of a historical bank, preserving the historical design of the building. Offer- ing undifferentiated products in a mundane store would result in a loss of business. It is rumored that Starbucks has more than 100 designers working on hundreds of unique, one-off designs for urban spaces. From an individual location perspective, many urban retail establishments are less productive than their coun- terparts in a suburban shopping mall, so innovations are needed. Indeed, the complexities of urban retail are driv- ing innovation in all aspects of the business, including the supply chain. Logistics networks—driven by the unique challenges of traffic patterns, restricted access, local ordi- nances, parking requirements, and loading–unloading difficulties—are turning to a portfolio of delivery meth- ods to ensure the right products get to the right places at the right time for the right cost. For example, Fresh Direct uses a fleet of home-deliv- ery vehicles in New York City. Coca-Cola uses 100% electric vehicles, and even UPS’s familiar brown trucks are being reconceived into smaller vehicles. Retail is evolving from consumers going to the store and bringing their items home to ordering on the web and having items delivered to their home. Eventually, with ubiquitous mobile capabilities, consumers will order from anywhere and want their goods to be deliv- ered anywhere. Wal-Mart is experimenting with multi- channel shopping in which they ask in-store customers if they want to help deliver an item to a neighbor in exchange for a gift card. Traditional paths of the first and last mile are changing. Amazon, which staunchly fought application of sales tax to online purchases, is now giving up its fight in cer- tain states, presumably because it plans to build fulfill- ment centers in those states so that it can offer same-day delivery. Chumley posed a question: Will these actions lead to a race to the bottom, as Wal-Mart offers same- day delivery for $10 regardless of purchase amount, and Amazon counteroffers with $8.99 delivery for same- day shipment plus 99 cents per item? Consumers are demanding same-day delivery but do not want to pay for the convenience. Who will pay for the needed data? Who will pay for the increased complexity? Some interesting multichannel solutions are emerg- ing, Chumley noted. For example, eBay is testing 1-hour delivery in San Francisco. A start-up called Deliv aims to help existing retailers by using crowdsourcing as a way to deliver goods to customers in the same day. TaskRabbit lets consumers offer any task (such as getting groceries) for which they are willing to pay, and individuals bid on what they think the work is worth. Instacart offers same- day grocery delivery in San Francisco, and Google Shop- ping Express is testing 1-hour and same-day delivery. Retailers are scrambling to get data for models, but by the time they get the data, Chumley said, it may be too late. Chumley described another innovative retailing model, that of the “endless aisle” offered by retailer Tesco in Korea. Located in a subway station in Seoul, the aisle is a virtual aisle consisting of photos of 500 prod- ucts with barcodes. Consumers can scan the barcodes on products they want and submit their order in the morn- ing; the order is ready to pick up on their commute home in the evening. The concept is being tested in New York City, as well. These nontraditional distribution networks are adding strain to existing logistics networks. 7-Eleven’s Response to New Urban Consumer Demands Chumley shifted to describing 7-Eleven’s response to these new urban consumer demands. The chain was founded in 1927 as an ice house, but with the inven- tion of the refrigerator 2 years later, the founder pivoted to offer fresh items such as milk (items that would go into the refrigerator) and remain open 7 days a week. Founder Joe Thompson’s motto was “Give them what they want, when and where they want it.” 7-Eleven is now the world’s largest retailer in terms of the number of outlets: 50,200 locations doing more than 14 billion cus- tomer transactions a year with total worldwide sales of more than $85 billion. Over its 86-year history, the com- pany has moved from horse-drawn carriages for local delivery to 100% electric vehicles that can hold both hot and cold food in the same small delivery car. 7-Eleven receives 17 million deliveries a year to its stores in the United States alone. The items get to a store in one of three ways. The first way, a central distribu- tion center (CDC), is a dedicated network that brings fresh items such as baked goods daily to a store. Second is the “wholesaler” way, which is a shared network that delivers twice a week. The third method is direct store delivery (DSD) from multiple suppliers who deliver more than 30 times a week. The typical store receives about 40 deliveries a week. All products are ordered by the store (as opposed to being pushed onto the store by headquar- ters), and store managers determine each store’s product assortment. Chumley next showed photographs of different 7-Eleven urban store formats, saying that the most leading-edge solutions were in Europe and Asia, not the United States. From rolling carts in Thailand to stores that do not have walls and never close in Singapore to a store in a central train station in Europe that is among the most productive in the entire 7-Eleven network, 7-Eleven is experimenting with many formats. In Japan, local franchisees offer delivery to the home within a 1- to 3-kilometer radius, delivering via electric vehicles. Another concept is a “store within a truck” that drives to neighborhoods with elderly populations. Finally, in

16 city logistics research: a transatlantic perspective Taiwan, 7-Eleven offers a catalog of 600,000 items that can be ordered by phone and delivered to the store or any location of the customer’s choosing. 7-Eleven is also experimenting with multichannel business models and has partnered with Amazon in a model in which customers could order items from Ama- zon and have them delivered to a locker in a 7-Eleven store. The customer feedback has been favorable, but the big orange locker boxes in the store are a bit cumber- some, so 7-Eleven is still testing the concept. Chumley concluded by reiterating the central prob- lem that urban retailers face, namely that consumers are becoming more and more demanding and are increas- ingly unwilling to pay for convenience. As a result, networks have become complex and their problems expensive to solve. Chumley offered four topics for further discussion: 1. Changes in consumption patterns and buying behaviors of recently urban consumers; 2. The impact of multichannel commerce on brick and mortar stores (who is winning, and why); 3. Successful multichannel operations (how to man- age a cross-channel portfolio); and 4. Understanding desirable, feasible, and viable home- delivery business models. QuestIons And AnsWers WIth robert chuMley Chelsea (Chip) White asked whether, as 7-Eleven opti- mizes its flow of goods, it also looks at its flow of information. Chumley replied that 7-Eleven can correlate sales with temperature, time of day, and humidity. The issue is to turn the data into consumable information that leads to decision support and taking a different course of action when needed. He said 7-Eleven has more data that it can use, so the company needs to decide which data are the key indicators. No two 7-Eleven stores have the same assortment. The typical store has 2,500 unique items of 9,500 choices, but the broader funnel is 95,000 items. It’s difficult to drive consistent decision making given the huge assortment. Caplice asked whether 7-Eleven works with its DSD sup- pliers to go to CDC. Chumley replied that 7-Eleven is exploring pilots in Los Angeles, but that the DSD vendors believe that DSD is a true value proposition and do not want to let it go. When Chumley worked for Coca-Cola, he believed in the value of DSD, too. He believes that DSD vendors eventually need to understand that channel blurring is taking place and that the distinction does not exist in the customer’s mind or the retailer’s mind. In the Los Ange- les pilot, the only trucks allowed on the 7-Eleven parking lots are those that came from a CDC. Edgar Blanco asked if Chumley saw synergies from Europe or other countries. Chumley replied that in the United States, fresh food is 10% of 7-Eleven’s total business, but in Scandinavia and Asia fresh food is 30% to 50% of the total business. In the United States, CDCs have a range of 200 miles for daily delivery. In Japan, with 15,000 stores, the range is 12 miles, with deliveries of fresh and hot food arriving three times per day. Deliveries are determined by what is ordered, so there is a product mix issue. But there is also the emergence of different strategies of delivery out from the store, so the end point may no longer be just the store. The network has to adapt for both inbound delivery and outbound delivery from the store to a cus- tomer location. Caitlin Rayman said that there had been discussions during the symposium about engaging with the private sector for data and planning. She asked Chumley how researchers could work with 7-Eleven to improve col- laboration and data sharing. Chumley replied that the best course was to contact the Vice President of Logistics for 7-Eleven. He said that in the United States, most deliveries to the stores are made at night. Trucks leave the CDC at 6 p.m., deliver to the store at 3 a.m., and are back at the CDC by 6 a.m. The DSDs come around 4:30 p.m. 7-Eleven wants trucks off its lots during peak hours to make room for customers. Lance Grenzeback asked if 7-Eleven used urban conges- tion data to make changes to its routing decisions. Chumley replied that when he worked at Coca-Cola, daily dynamic routing was done based on traffic and construction updates. The company knew from cases and aggregate delivery time on the truck that a route was, say, 8 hours and 22 minutes. If a driver was in 5% to 10% of that time, the route time was validated. So companies can and do use congestion data. 7-Eleven has static routing (trucks deliver to stores in the same order), so traffic and congestion merely increase the delivery time because the routes are static. But the company could influence route times if it looked at individual routes. Lenz said that small shops cannot afford nighttime deliv- ery because of the staffing required to receive it, and there are no lockers. She asked how 7-Eleven manages it. Chumley answered that 7-Eleven stores are open 24 hours a day, so there is always someone there to take delivery. Because the routes are static, a store knows that a truck will be there between, say, 3:00 and 3:30 a.m.

17demand patterns and trends Chumley added that 80% of 7-Eleven’s stores in the United States are not in urban areas, so noise-level issues are not a concern now but may be in the future. synthesIs And suMMAry Laetitia Dablanc Summarizing the Demand Patterns and Trends session, Laetitia Dablanc said that researchers may have more freight demand data than they think they have. Data collection has been done in many cities, as Routhier and Jaller described, and individual companies know about their customers’ demand and the evolution of that demand, as Chumley discussed. Large parcel and express companies have a global view of hundreds of thousands of operations an hour. But the information is partial, dispersed, and local. Barcelona and Madrid collected local data, but the data were not shared, even between those two cities. It is sur- prising, but data are not being collected even within a single region, Dablanc said. Also, the data are often col- lected for a different purpose, as Browne and Goodchild mentioned in the first session. Another comment during the question-and-answer session that followed Jaller’s presentation was that including the supply chain in engineers’ models is dif- ficult. Browne and Goodchild also made this point, Dablanc noted. As Jaller said, there is a need to over- come the “partial” view of the urban freight system. Researchers have to assemble what exists and also create data from scratch by using primary data sources. As the French UFSs have demonstrated, comprehen- sive surveys that include all urban supply chains can be done. One way to do them is as two coordinated surveys: one surveying the generators of demand and the other surveying delivery operators. Such surveys are costly, as both Routhier and Jaller stated. Jaller estimated that the ideal UFS for New York City would cost $7 mil- lion. However, comprehensive surveys can lead to better modeling. Jaller’s presentation showed that there are many dimensions to urban freight generation analysis. Freight demand can be both an economic manifestation (freight generation) and a logistics one (freight trip generation). Researchers need to be aware of data gaps, and they need to understand behaviors, not just collect traditional data. Few specifically urban case studies exist, and there is a trade-off between the need for comprehensive surveys to resolve modeling issues and the cost of those comprehen- sive surveys. A noteworthy point Routhier made was that in Paris, cafés, hotels, and restaurants generate five times more deliveries than big box retailers. That finding shows that freight demand is not just the product of big players, but also of these small generators of freight, such as local stores and offices. To understand urban demand patterns, researchers need to understand retailers’ internal decision-making pro- cesses as they respond to changes in consumer behavior and adapt their transportation and logistics in response to the trends they see. Urban consumers do not want to pay for deliveries, so retailers have to be smart and efficient in delivering goods at low prices. Urban environments require new small store formats in densely populated areas, so retailers’ transportation and logistics must adapt to this new situation. Another conclusion Dablanc drew from the presenta- tions so far is that in both Europe and the United States, very few private software developers and consultants (such as PTV) are developing urban freight modeling. City authorities are not requesting these models because, in general, urban freight transportation “works” in that we all get our goods. However, the externalities of con- gestion and pollution are motivators for learning more about freight demand. Dablanc ended with focusing on some of the speak- ers’ suggested topics for collaborative research. First, she identified the topic of innovative data collection method- ologies and technologies, such as GPS; the use of existing records, like those of departments of motor vehicles in the United States or light transport registers in the EU; and a better CFS for urban areas. Second, she suggested comparing establishment databases between Europe and the United States. Third, she suggested comparing urban freight behaviors between Europe and the United States. This last topic had three subparts: 1. Assessing the operational applicability of EU mod- els (like FRETURB) in the United States, and vice versa; 2. Examining the changing urban economy in the face of e-commerce (the future for stores, new delivery busi- ness models); and 3. Examining the changing urban economy in the face of increasing urbanization.