Abstract
The rise of mobile food vending in US cities combines urban space and mobility with continuous online communication. Unlike traditional urban spaces that are predictable and known, contemporary vendors use information technology to generate impromptu social settings in unconventional and often underutilized spaces. This unique condition requires new methods that interpret online communication as a critical component in the production of new forms of public life. We suggest qualitative approaches combined with data-driven analyses are necessary when planning for emergent behavior. In Charlotte, NC, we investigate the daily operations, tweet content, and spatial and temporal sequencing of six vendors over an extended period of time. The study illustrates the interrelationship between data, urban space, and time and finds that a significant proportion of tweet content is used to announce vending locations in a time-based pattern and that the spatial construction of events is often independent of traditional urban form.
Introduction
The proliferation of Twitter-based mobile food vendors in cities across the United States has created opportunities for social gatherings to occur spontaneously and in unpredictable locations. Structured upon both physical mobility and continuous online communication, this recent phenomenon poses interesting questions about urban space and information technology. How does information technology inform social practices in urban space and what new spatial and temporal relationships develop as a result? How can we develop an accurate description of emerging activities in cities that combines real-time data with more qualitative forms of urban analysis?
The objective of this research is to interpret emerging relationships among online communication and urban settings. As such, we seek to develop new forms of analysis that accurately describe actual behavior and allow for planning that suits emergent behavior. Tweets, posted on the social media website Twitter, provide a lens to better understand the multiple roles and functions of real-time information as it informs urban settings. We analyze food vending in Charlotte, NC, closely documenting the daily operations, tweet content, and spatial locations of six vendors over an extended period of time. Using an integrated approach of qualitative and quantitative techniques, we show how online communication creates new spatial and temporal relationships in cities and provides new opportuntiy for innovative analyses.
Our hypothesis is that contemporary urban form must account for the transformation and dislocation caused by the rapid proliferation of portable, omnipresent information technology. While using methods of topic modeling and frequency analysis of tweet content, onsite documentation, participant observation, and interviewing, we reveal unique relationships between data, space, and time. We analyze this condition by identifying contemporary food vending as events that have both a temporal and a spatial component. We also reveal the ways events are supported by online communication, located with little relationship to traditional urban form, and can be further organized and studied using emerging techniques in visual analytics.
The field of urban design focuses on the spatial growth of the city and the design of public spaces using qualitative readings of the urban context. These methods lack a way to anlayze temporary and spontaneous social settings that are supported by the rapid nature of online communication. Furthermore, urban activities that are both digital and spatial can alter daily patterns of use in cities and prompt planners and urban officials to accommodate and respond to new design settings. As such, there is a pertinent need to develop an integrated approach for analysis and planning that critically examines online communication in relation to spatial settings in order to challenge established methods and suggest alternative frameworks.
Contemporary food vending
For centuries, street food vending has been a cultural practice in many countries supporting local economies and serving food shortages. In the United States however, street vendors emerged with the migration of immigrants to New York City in the 1700s and 1800s, primarily serving working-class tenement neighborhoods. Over time, vendors have shown to provide social and economic benefits to urban communities, and yet, urban renewal efforts, driven by public health concerns and cultural and racial stigmas, suggest otherwise. Although a complete history of food vending is outside the scope of this discussion, historical dynamics show that vendors have faced repeated political resistance and that their growth and decline parallel the economic and regulatory climate of cities. Today, new wave food trucks negotiate many of the same historical challenges, but the demand for vendors has dramatically grown among urban populations, suggesting a new shift in public perception.
At a basic level, new wave vendors are defined by their high-quality and often high-priced food served in stylishly branded and fully equipped catering trucks to patrons desiring a new type of outdoor food experience. Information technology allows vendors to communicate and exchange real-time information as well as to mobilize their business activating urban areas with a predictable crowd. Unlike prior forms of vending that relied on word-of-mouth communication, physical proximity to customers, or a repeated presence, technology affords vendors the ability to notify large numbers of people well in advance. This dynamic allows vendors to create a demand virtually, building a select population that is well versed in contemporary food culture.
The rapid growth of these food vendors can be linked to a variety of converging factors. Specifically, the economic downturn of 2008 and the subsequent decrease in consumer spending and consumer confidence shifted demand away from the broader food-services sector and toward affordable food options (Brennan, 2014). The increasing use and accessibility of social media platforms, smart phones, and global positioning applications allow vendors new ways to market their business and announce their locations. Aspiring chefs and culinary students, who enjoy experimenting with culturally diverse cuisines and who are equipped with the knowledge of the restaurant industry, seek similar success as the popular Los Angeles–based Kogi BBQ food trucks, known for serving Korean–Mexican tacos. Furthermore, young culinary entrepreneurs and established chefs, who may have difficulty maintaining their restaurants, find starting a mobile food business more financially feasible and flexible. 1 Last, the lack of prior mobile vending precedents and loose municipal ordinances initally allowed vendors to easily navigate urban areas. Combined, these forces supported the national growth of the food truck industry whose revenue nearly doubled from 6% to 18% in 2008 (Brennan, 2014). In September 2010, the US Small Business Administration created a webpage with “Tips for Starting your Own Street Food Business,” and many local governments for the first time started addressing their operations whether viewed as a community asset or competition to established restaurants. Although the exact number of mobile food vendors in the United States is undocumented, it is estimated that the food truck industry is worth US$800 m and estimated to increase to US$985 m by 2019 (Brennan, 2014).
The speed and usability of information technology to promote, connect, and expand vendor operations across cities amplified the growth of the vending industry. Vendors, customers, and advocacy groups use social media platforms (e.g. Twitter, Facebook), smart phone applications that offer real-time tracking of trucks (e.g. TruxMap, Food Truck Fiesta, Foursquare, Road Stoves GPS, and Truck Spotting), smart phone payment applications (e.g. Intuit’s GoPayment and Square), photography and video platforms (e.g. Instagram, Vine), as well as blog, business, and food review websites (e.g. MobiMunch, Yelp, FoodTrucksIn, and Urbanspoon). These tools, which together create a media ecology, compliment the nimble and flexible business models of vendors by providing an instant way to build a demand.
Twitter initially became the most popular among vendors and still is today. Structured on a micro-blogging framework that allows sending of 140 character messages or “tweets,” vendors use Twitter as a free mass-marketing tool to communicate to a localized audience their latest or future locations, daily menu items, or if they are out of service. Vendors also use Twitter to choose locations on their daily route and to understand the locations of their fellow vendors. Customers find this real-time information helpful in locating their favorite trucks and menu items. In an industry reliant on mobility, Twitter provides a virtual infrastructure to communicate real-time information and assures vendors a sufficient customer base in a variety of public and private locations, such as office parks, college campuses, alleyways, empty parking lots, plazas, community parks, and tourist areas.
Food vendors do not roam freely; they continuously negotiate evolving spatial regulations. Some public adversaries argue that vendors congest streets, weaken business for brick and mortar establishments, contribute to crime, and use unsafe food practices and therefore should be contained or removed entirely. In most US cities, the restaurant industry succeeds in establishing spatial boundaries around their businesses. However, in Portland, where small business incubation is established, vendors are known to cultivate community in neighborhoods, provide access to food in areas with few options, promote entrepreneurship, and create jobs in a down economy (Kapell et al., 2008).
The food vending landscape in Charlotte
Unlike west coast US cities that have hundreds of vendors, Charlotte’s slower growth of new wave food vending can be attributed to strict spatial regulations that exclude vendors from street parking in the downtown business district as well as the need to educate customers about new food experiences and truck sanitation. “Opposed to roach coaches that sold the prepackaged stuff, I had to educate them on how gourmet food trucks actually serve quality food” mentioned one vendor (Cirsan, 2013, personal communication). In 2008, Charlotte officials, concerned with crime, traffic congestion, and noise associated with the rising number of vendors, sought to revamp its vending ordinances with strict controls on public right of ways. This policy first concerned established Latino vendors who had been serving at construction work places, strip malls, and convenient store parking lots for years. With an unsuccessful petition, the Latino vending community began to decline just as the new wave vendors began to expand, following carefully made business plans and aggressively seeking new locations to arrange property agreements.
In 2011, a privately owned lot in the up and coming Historic South End District was sited for food trucks by Charlotte’s Downtown Association. The first weeks were experimental and slow hosting just four trucks, but grew rapidly in the spring months of 2012. “Food Truck Friday” now hosts 20 trucks each serving between 300 and 400 tickets in 4 hours (Portillo, 2014). Today, 125 mobile food units and 56 single operator pushcarts have registered health permits in Charlotte-Mecklenburg County (Mecklenburg County Health Department, 2013), of which approximately 40 are new wave mobile vendors (FoodTrucksIn, 2014) and 20 operate on any given day.
In 2014, Charlotte’s Planning Department proposed another set of regulations that if implemented will impose significant new spatial restrictions that limit operating within 100 feet of a restaurant, bar, or nightclub. “I don’t think they are intentionally trying to harm food trucks, but I do think they don’t understand what we do,” said one food truck owner (Portillo, 2014: 1). As the industry continues to grow, Charlotte’s vendors, along with many other cities, will face new regulatory hurdles. Despite these challenges, the national industry is expected to grow with the number of trucks increasing an annualized 9.4% in the next 5 years (Brennan, 2014).
Literature review
Scholarship on vending first investigated the politics of urban informality, micro-economies, and cultural and symbolic capital in developing countries (Bhat and Waghray, 2000; Tinker, 1997). Aside from industry reports and numerous news articles, scholarship on the contemporary vending involves conflicts over vendor rights, regulatory pressures, cultural stigmas, and litigation (Hernandez-Lopez, 2012; Linnekin et al., 2011; Norman et al., 2011). Literature addressing the ramifications of information technology and mobile food vending is limited to Caldwell’s (2012) work that identifies patrons’ ability to oscillate between the virtual realm of tweeting and actual vending locations as central to this new food experience and the work of Wessel (2012) that found cognitive and behavioral differences among food truck patrons who use social media and those who do not. Broadly, the topic of mobile food vending can be situated in a variety of social, cultural, and economic discourses. We aim to address the theoretical and empirical approaches in urban design and media studies.
Recent urban design approaches illustrate the temporary and flexible nature of activating urban space. A variety of related discourses such as tactical urbanism (Lydon and Garcia, 2014), do-it-yourself urban design (Douglas, 2013), guerilla urbanism (Hou, 2010), spontaneous interventions (Venice Biennale, 2012), opportunistic urbanism (Ramirez-Lovering, 2008), and temporary cities (Bishop and Williams, 2012) reference case studies of pop-up retail and cafes, guerilla bike lanes, flash mob gatherings, chair bombing, as well as food and craft vending. Douglas (2013) suggests these new forms of human-centered urban design are reactions to formal, tightly controlled, and regulated urban planning and a byproduct of the uneven spatial development produced by global finance and urban renewal. Social theorists Henri Lefebvre (1974) and Michel de Certeau (1984) influence this discourse by addressing space as a social construction whereby citizens exercise agency in a capitalistic society. While informal urban tactics take on many forms and in a variety of urban contexts in developing or developed nations, they are based on creating urban infrastructure through collaboration and using attainable design tactics that circumvent formal, lengthy planning processes. This topic is particularly relevant for situating mobile food vending as an outgrowth of prior restrictive modes of planning and a flexible urban practice where vendors appropriate underused spaces to operate their business.
Urban space and information technology
While flexible urban space is one paradigm that explains food vending, we suggest vendors’ intimate exchange with urban space and information technology has larger implications for traditional approaches of urban analysis.
Melvin Webber’s (1964) established claim that electronic communication produces aspatial interactions is applicable both to the introduction of the telephone and to the proliferation of the digital communication. This insight initially led theorists to question the relevancy of city space, suggesting that reduced face-to-face contact, home employment, and infrequent automobile travel diminish the economic value of place (Negroponte, 1995; Pascal, 1987). These technological determinist views cast technology as “an essential and independent agent of change that is separated from the social world” (Graham, 1998: 168). Conversely, theorists propose that technology breeds intensification of urban activity through both electronic and transportation networks (Gaspar and Glaser, 1998; Graham and Marvin, 2001).
A second discourse aims to explain the co-evolution of city development and information technology in a capitalist system, which focuses on cities as global finance centers and communication hubs that are electronically linked, yet face spatial polarization through uneven development (Castells, 2000; Sassen, 2001). Similarly, scholars analyzing regional contexts argue that information technology creates new, specialized populations and social structures within a city, whether through local competitive advantage (Saxenian, 1996) or electronic communication channels (Webber, 1964). More recently, Townsend (2013) recognizes the role of technology firms in civic collaborations where policy makers institutionalize strategic thinking about information technology. He found a disconnect between technological solutions and the urban setting, concluding that the emergence of smart cities and associated policy frameworks need to recognize citizens’ keen sense of urban problems.
Discussions in the 1990s around online communities and social networks explored the roles of social identity, participatory democracy, countercultures, and privacy (Jones, 1998; Rheingold, 1993), but more recently some online communities have developed a direct relationship with physical communities. Websites such as Nextdoor.com, i-Neighbors.org, frontporchforum.com, and Patch.com facilitate communication networks focused on local settings. The unique quality of these platforms is not in their ability to connect users in a global or national framework but rather their use of the neighborhood unit as an organizing mechanism, which reflects the enduring significance of place in building networked relationships.
Studies of social justice movements operating both online and offline explore the online dialogue generated from political unrest and the ways information technology mobilizes people in space (AlSayyad and Guvenc, 2013; Harlow, 2011; Massey and Snyder, 2012). Investigations of the Occupy Movement and the Egyptian Uprising reveal the rapid spread of trending information between activists locally and regionally. In this context of social change, the role of online communication manifests in urban space through expression and dissent.
Urban sensing has emerged since the proliferation of smart phone technology (Campbell et al., 2008; Cuff et al., 2008). Geocoded data from microchip sensors about a variety of urban processes such as traffic flows, air pollution (Resch et al., 2012), population densities (Calabrese et al., 2010), and weather conditions (Eisenman and Campbell, 2006) can be easily collected and analyzed using geographic information system (GIS) platforms or visual analytic programs. Urban sensing helps us to understand invisible and fluid urban systems, allowing data to become a form of public infrastructure that supports decision making among citizens and city officials.
The fast-growing transportation network companies (TNC) or rideshare programs such as Uber, Lyft, and Sidecar use geolocational data from cell phones to unite customers with available drivers providing transportation options to once neglected areas of cities and decreasing traffic congestion. Ride-share companies redefine existing economies such as taxi companies through faster and more efficient services that are supported by smart phone technology.
Twitter’s real-time data have been used to detect natural hazards such as earthquakes (Sakaki et al., 2010), fires, and floods (Vieweg et al., 2010) through content analysis. These studies help to advance situational awareness about natural events taking place in the environment using online communication. For example, work by Sakaki et al. locates and verifies earthquakes occurring in Japan through tweets and instantly emails Twitter members the information. These useful risk management tools address the intersection of communication and space in real-time.
Geographer Stephen Graham (1998) reminds us that technological determinism assumes that technology directly causes social and spatial change in cities. He acknowledges the human construction of space as a method to ground and contextualize uses of technologies. He argues for a recombination perspective anchored in principles of actor–network theory, which suggests agency among humans and technology is a relational process containing multiple, heterogeneous networks. We find this approach useful to explain the variety of ways technologies can have contingent effects and become linked to specific social contexts.
Analysis and discussion
This study uses a variety of methods to examine the mobile food vendors’ online comunication and their shifting operational settings, such as topic modeling and frequency analysis of tweet content, ethnographic interviews and participant observation of vendor operations, and spatial mapping of vendors’ movements and temporal sequencing throughout the city. Six mobile food vendors, of the 20 regularly active in Charlotte, were chosen based on the number of “followers” (>1000) tallied on each vendor’s Twitter account and their repeated presence at Charlotte’s popular “Food Truck Friday.” We collected 1000 tweets from each vendor on 15 November 2012, which included the vendor’s account name, the date and time of each tweet, the number of retweets, and the tweet content. The tweets spanned periods of 4–9 months depending on the vendor’s tweeting frequency. Each vendor was interviewed twice on topics including length of time in the business, operating methods, approach to using Twitter, scheduling procedures for events and locations, and menu items related to location or time. Each vendor was visited at least three times to record the arrangement of the truck and movement of customers through diagrams, photographs, and time-lapse video.
Communicating space: Tweet cluster analysis
Our analysis began by examining the content of each vendor’s 1000 tweets using a simple automatic topic analysis. This effort provides a set of topics that naturally emerge from grouping verbally similar tweets together. Text analysis of tweets is difficult using traditional automated topic analysis due to the short length of documents, heavy use of slang and abbreviations, and noise from URLs, attached images, and automated tweets from applications. More sophisticated methods such as latent semantic analysis (LSA) did not produce better results with our data, likely because the short length of the documents violates the assumptions of these methods. Therefore, we used a simple topic analysis method based on k-means clustering.
We first simplified the tweet data by extracting the most frequent 1000 keywords minus a standard list of stop words, or words to ignore (e.g. overly common words such as the, an, of), augmented with tweet-specific stop words including RT, #, and URL components such as http and t.co (the standard beginning of a URL automatically shortened by Twitter). These 1000 keywords become dimensions in a transformed dataset. Instead of being a string of words, each tweet is a vector of zeros (does not contain the keyword) and ones (contains the keyword). This is known as a “bag of words” technique since the order of words is ignored. This results in a dataset in which each tweet is a point in a high-dimensional space.
We then performed k-means clustering using the Euclidean distance between the vectors that represent the tweets. k-means clustering takes a number of clusters as an argument and classifies data items according to the tightest clusters when the space is separated into that number of groups. We ran the analysis with values of k ranging from 4 to 10 clusters and examined the results by hand. Based on the qualitative results of interviews, the topics produced by the eight-cluster analysis were judged to be the most meaningful. While this is a subjective judgment, there was a great deal of consistency between the analyses, so this choice does not significantly affect our results.
The eight clusters in the final analysis are summarized in the following table (Figure 1). The top 10 features for each cluster are listed, along with the number of tweets in the dataset, which are classified as part of the cluster. The largest cluster, which we labeled “Miscellaneous,” is not a clearly defined topic and contains a number of generic terms found in other clusters. This was a common feature to all of the topic analyses we produced and is likely related to the fact that many tweets are difficult to classify due to their noisiness and the small number of words in the classification. We were able to derive semantically separable clusters due to the fact that our tweets are more restricted in content than most collections of tweet data, since they come from Twitter accounts with the common goal of marketing a food truck. Even so, there are large numbers of tweets in the dataset that fall outside this common semantic space, including mentions of and retweets from accounts not associated with food trucks. The Miscellaneous cluster is largely an artifact of this off-topic tweeting and, as a result, is more affected by the difficulty of meaningfully classifying very short documents based on a bag-of-words approach.
Eight topic clusters derived from k-means clustering analysis.
The other seven clusters reveal more meaningful patterns that relate to the trends we saw in interviews and the onsite analysis. The “Food Truck Trend” cluster, which contains a mix of hashtags, mentions, and words relating to the phrase “Charlotte food trucks,” is related to trending topics about food trucks that emerge around popular events. The “Truck Mentions” cluster, which includes vendor usernames preceded by the @ symbol and is used to mention or reply to another Twitter account holder, suggests an active and ongoing dialogue occurs between vendors and their customers. The “Food Truck Friday” cluster contains a series of terms such as rally, Friday, and Southend that relate to topics about the most popular event in the city. The clusters “Schedule” and “Location” include terms about time and location-based information, which reveal the robust link between data and urban space.
After eliminating the “Miscellaneous” cluster, the remaining seven clusters were used as a framework to compare the total tweets and retweets (i.e. tweet content posted by another user) for each vendor. Figure 2 shows the “Schedule” cluster was tweeted and retweeted most frequently across all the clusters highlighting the importance of advanced planning by vendors. Also, the clusters “Food Truck Trend” and “Food Truck Friday,” that include popular general terms about food vending, are consistently retweeted across all vendors. This reveals tweets travel between Twitter accounts frequently, buildng a dense network of information exchange. Furthermore, tweets that include terms about “Gratitude” generated very high retweeting for the vendors @WingzzaTruck and @roamingforkNC, showing positive dialogue and relationships can extend past the point of sale. Similarly, the owner of @roamingforkNC mentioned, “Rather than focusing on money, I think first about quality and customer service. My customers need to know that they are appreciated and I am thankful for the opportunity to get out there and give them what they expect” (Cirsan, 2013, personal communication). Last, the “Truck Mentions” cluster shows a consistent amount of retweeting suggesting a strong back and forth communication exists between vendors and their customers which may look like, “First @WingzzaTruck stop of the year! They are always so nice and the food is delish!”
Total tweets and retweets organized by vendor and cluster topic.
Event construction through tweets
If a physical time-based event is linked to online communication, the tweets should reveal time and location-based information. We began by organizing vendors’ tweets into event related (spatial) and nonevent related and identified the frequency of locations in the event-related tweets.
In order to determine the tweeting frequency for each vendor, a 2-month timeline (September–October 2012) was constructed showing the days in which events occur, the numbers of event-related tweets, and the number of nonevent-related tweets (Figure 3). The total count of events, event-related tweets, and nonevent-related tweets are identified for each vendors’ timeline. In general, the total number of tweets for any one event ranged from one to nine, the time frame ranged from 3 weeks before and including the event, and event-related tweets most frequently occur on the day of an event or within 1 week prior to an event.
A 2-month timeline of each vendor: the day an event occurs (red lines), the frequency of multiple events occurring on a single day (dark red lines), event-related tweets (green), and other tweets (blue).
First, our timelines reveal multiple events occur on a single day more frequently during the work week. This challenges the presumption that vending is a leisurely weekend destination. Employees now find food trucks to be a common food choice during lunch hours. Subsequently, the owners of @herban_legend, @TheTINKitchen, and @roamingforkNC mentioned they serve lunch 3 days a week at remote business park locations or college campuses.
Second, the nonevent-related tweeting reveals vendors often communicate about a variety of aspects, such as types of food, customer feedback, truck operating issues, and their personal life. A tweet by @roamingforkNC states, “In Costa Rica … Rejuvenating … Relaxing … What’s happening where you are?” The vendor @WingzzaTruck tweeted, “Hey #TeamWingzza if you could add one item to our menu what would it be? Working on some things for you! #CLTFood.” This tweeting proves to be a consistent type of communication throughout the entire 2-month mapping for all vendors, suggesting it is essential to continue dialogue and engage customers in ways that are personal and positive.
Third, some vendors when compared show an inverse relationship between the number of event-related tweets and the number of events. For @herban_legend, who has a total of 68 events and 95 event-related tweets, and @papiquesotruck, who has 41 events and 163 event-related tweets, the amount of tweets bares no relationship to an increase of events. Interestingly, this may be related to @papiquesotruck’s later opening in 2012 and need to attract more followers. The vendor @papiquesotruck mentioned, “My website doesn’t drive my business, social media does. I can tell that my customers use Twitter a lot because we will post a secret menu occasionally and within minutes we’ll have people standing in line” (Stockholm, 2013, personal communication).
Using the vendors’ event-related tweets, we identified how many times a location was mentioned and tagged it as either “one-time” or “repeated.” Most locations were named multiple times for a given vendor (e.g. Food Truck Friday), while others were only mentioned once (e.g. Democratic National Convention). We then tallied the number of each type of event and aggregated them to determine how many days in advance vendors tweeted (Figure 4). We found that one-time events are tweeted about more often and earlier in advance (2.61 times on average; 1.84 days in advance) than repeated events (1.84 times on average; 0.86 days in advance), although they take place less often. This suggests there is high value in online communication when familiarizing customers with new locations. The mobile nature of their practice can serve as a disadvantage compared to restaurants who have predictable and fixed locations. The vendor @papiquesotruck is well aware of this challenge stating, “I like building long-term partnerships with businesses so that my customers know where to find me on a regular basis” (Stockholm, 2013, personal communication).
Average number of one-time event locations and repeated event locations per vendor.
Next, we investigated a single vendor, @herban_legend, to better understand event-related tweeting in relation to events. We generated a spark line for each of the 209 total events that shows corresponding tweets identified with each event (Figure 5). The largest amount of tweets related to a specific event is nine, and a pattern of tweeting emerges well in advance of the day of the event and often multiple times. This asynchronous relationship between the event and online communication is represented in the following: 33% of all events have a single tweet at least 1 day before the event, 61% of events with two or more total tweets have at least one tweet before the day of the event, and 92% of events with three or more total tweets have at least one tweet before the day of the event. These numbers reveal vendors’ routine practice of planning and announcing events ahead of time. Unlike prior forms of vending, new wave vendors are able to construct an event and ensure a predictable customer base well before it physically occurs.
Mapping of event-related tweets across time specific to each event for vendor @herban_legend.
Vendor time–space sequencing
Working toward a visual analytic interface that illustrates the intersection of space, time, and data, we illustrate the vendors’ choice of location and movement over time (Figure 6), first over the course of a year for relative frequency of locations, and second, over the course of 1 week to illustrate their movement between locations. Vendor websites and tweets over 12 months provided location information.
Mapping of vendors’ location frequency for 1 year and movement for 1 week.
In the first map, locations visited three or more times by vendors within the 12-month period are geographically marked using a circle. The diameter of the circle corresponds to the number of times that location repeats among all vendors. If a location is visited eight times, the diameter of the circle is eight pixels. If more than one vendor visits a location, the circle is divided corresponding to the percentage of times that each vendor frequents that particular location. The second map relates to the spatial movements of the vendors. Considering that the data shows vendor activity often reoccurs in a weekly pattern, we chose a random week and plotted arrows to connect the locations in the order visited. We then repeated this process for each vendor.
In addition to visualizing the frequency of space and the flows of movement between locations as part of an analytic program, the vendor locations reflect many of Charlotte’s geographic and economic conditions: a compact downtown core, a decentralized landscape linked by an extensive roadway system, business park developments located on the southern periphery, the university as a node of activity north, and major thoroughfares and areas of commerce throughout. Furthermore, this organization reveals an uneven spatial and temporal distribution. Much like the telephone permitted the decentralization of communication, online communication in this setting supports the mobile expansion of the food vending economy. Mobility allows vendors to travel to meet customer demand. While some may view this as a competitive advantage, vendors face the difficulty of making their business known in unfamiliar locations. Mapping the fluid nature of urban processes is a continual challenge for planners who seek to understand urban use patterns. In these maps, we hope to represent actual behaviors occuring in time and space.
Event spaces
Based on the highest frequency of events, we chose four spaces to compare with traditional gathering places and formerly established Latino vending locations. Using standard urban design techniques, we studied both the immediate site of the trucks (e.g. customers, site amenities, and any other co-located trucks) and their position within a larger urban context (e.g. nearby land and building uses) (Figure 7).
Four locations analyzed for their characteristics of traditional gathering spaces.
The first space is a privately owned vacant dirt parcel situated on the periphery of Charlotte’s central business district, also known as Historic South End. With the help of Charlotte’s Downtown Association, the site became home to “Food Truck Friday” in October 2011. Aside from its slow start, it is now a widely popular activity on Friday evenings in the city. The downtown association mentioned a number of factors were used in determining this location: visibility from pedestrian paths, scale of the space so that it could accommodate food trucks while also intimate enough for patrons, close proximity to public restrooms, available parking, and few nearby food establishments. This site’s context resembles a traditional gathering space with its abundance of open space and close proximity to retail and public transit. Yet, the site is left unused otherwise, lacks shade devices or trees, lacks permanent infrastructure, and has relatively low pedestrian activity other than Friday.
The second location caters to lunchtime patrons multiple days of the week on an active vehicular thoroughfare adjacent to a culinary university and a residential condominium. Key factors in the success of this location are truck visibility to vehicular traffic, accessibility by foot to residential and educational buildings, and nearby infrastructure for seating and shade. However, the space does not serve as a designated gathering space, rather the presence of the food truck parked on a busy street activates nearby spaces.
The third space is a large parking lot located in an office park south of the central business district. Patrons working in nearby offices take food to their cars to eat alone, sometimes in pairs, while others take food to their office building. Given the short lunch hour and the lack of seating options, this location has large influxes of people and then lulls of emptiness leaving patrons little opportunity to socialize. A vendor mentioned that her more dedicated customers order in advance, allowing them to avoid lines. The design elements of this space have no bearing on the vending experience or traditional gathering spaces.
The last location is nestled behind a local brewery in the Historic Arts District northeast of downtown. The paved parking lot accommodates a single truck nightly that is organized by the brewery owner. Given the remote location of the brewery, a majority of customers drive; however, customers tend to occupy the site for long periods of time. The site does offer seating for patrons indoors and outside, but lacks open space, shade, street visibility, and pedestrian and public transit access.
Our findings show that vending locations are not initially conceived with regard to accommodating people formally or informally. Situated on predominately privately owned and underutilized land, these spaces are instantly transformed with vibrant mobile restaurants, makeshift seating, and social activity. Unlike the established Latino vending locations that respond to a laborer’s workday serving at industrial work places, commercial strip malls, and convenient stores, new wave trucks cater to White and African American populations in office parks and large vacant lots to accommodate their pre-notified customers. Furthermore, the resemblance between new wave vending locations and traditional parks and plazas varies among sites. Overall, we suggest traditional urban design methods cannot fully explain the development of vending locations and that information technology affords a virtual dialogue where demand for vendors is generated and organized in advance.
Findings and conclusion
As we began this research, we faced the question of how information technology would affect the use and even the relevance of the city. To that end, we examined contemporary mobile food vending as a new form of urban spatial occupation linked to online communication. Our investigation of this phenomenon revealed the underlying temporal relationships afforded by online communication. This creates a demand for new food experiences in unfamiliar urban spaces by extending the reach of an audience beyond the physical propinquity of a food vendor. We suggest that this urban condition requires new strategies that combine real-time information with more traditional forms of analysis.
Presumably, mobile food vendors might find places in the city that closely follow and reinforce the planned spatial pattern. If this were true, existing urban design methods would be sufficient to explain and plan for their growth. On the other hand, the combination of online communication and food vendors could lead to extreme forms of diffusion. The locational decisions of a truck might reflect the aggregated demand derived from information technology, shifting locations in real-time to meet the greatest number of customers. The truck could move many times over the course of a day, a “just-in-time” restaurant. In this scenario, urban design is irrelevant; the trucks would move frequently and without regard to the nature of the space they occupy.
Our study shows how data, space, and time are a tightly woven network that functions before each food truck event. Specifically, spatial information is announced several times in advance of the scheduled date, creating an audience for a particular location at a particular time. This allows the event to reach more than a spatially adjacent audience, as opposed to prior modes that relied on repeated presence, word-of-mouth communication, and even music played through loudspeakers. We also found that the local spatial arrangement of the trucks was highly variable. Sometimes, the trucks reinforce a more or less predictable arrangement of urban form (e.g. Food Truck Friday), but often, the spatial arrangement responded only to the location of potential customers, especially in peripheral urban spaces (e.g. office parks and university campuses).
Thus, our research led us to suspect that neither data nor urban space considered separately could adequately explain vendors’ behavior. Existing descriptions of space using canonical methods of urban design could not describe the locations of the trucks with any degree of certainty, while at the same time, investigations of the data network using either topic modeling or temporal information was only meaningful with some knowledge of the spatial locations.
At the most primary level, the unit of analysis will become the event, something that combines location, time, and data. The identification of events will help to guide decisions relating to emergent forms of commercial, social, and civic functions (e.g. mobile health clinics, libraries, and retail businesses). This effect will be particularly pronounced with new businesses and social gatherings.
As a result, locations for public activities of all kinds will be less based on planned urban space than on propinquity. Being on “Main Street” will matter less than it used to, but being nearby will still be important. An audience will be able to find any event, but a specific location will be important for gathering a critical mass. Previously marginal urban spaces will become more important and valuable.
Finally, by using mixed methods, this research proposes a hybrid form of urban analysis that is informed both by a qualitative recognition of human urban places and a quantitative understanding of large flows of data. A visual analytic interface, suggested by the direction of our research, can offer an understanding of the underlying processes that create urban places. The defining quality of this interface will be its discursive nature, allowing users to test alternative interpretations while viewing multiple forms for information.
The emphasis on the spatial over the temporal in urban anlayses has become a serious drawback, making the understanding of new forms of urban activity difficult to see or to analyze. Considering only the informational network would be equally one sided. Our analyses of these urban events in Charlotte represent some first tentative steps toward an integration of qualitative and data-driven analyses. As data become more readily available, urban analysts will need to reevaluate traditional methods in order to interpret and plan for urban acitivities supported by online communication.
Footnotes
Acknowledgements
The authors wish to acknowledge students Araceli Bollo, Parris Boyd, Jake Coltrane, Allison Etheridge, Yashika Gulati, and Leslie King whose hard work and enthusiasm were critical to gathering and understanding this information.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.