Optimization of hot dog vendor location for college student convenience


Business site selection has always been high-stakes: the opening of a new business location has extremely large monetary implications. Location can impact margins, response to competition, and effective exploitation of possible market segments (Clarke 2013)(Cliquet 2013)(Ghosh 1983). It is also well-established that convenience is a significant factor in consumer decisions, especially those regarding food (Bonke 1996). In a university setting, decisions such as food vendor placement become particularly important, as daily food is the second highest consumer expenditure for college students (Adams 1997). In the present study, these understandings were incorporated into a decision procedure regarding the position of a hypothetical hot dog vendor on a college campus, in which convenience for students was evaluated using spatial information.

Assigned problem

A map was given of a college campus showing the walking paths and dormitories and approximate distances between the intersections (Figure 1). We were asked to answer questions about the location of a hot dog vender:

  1. Where on campus should you set up your stand?

  2. How does your location change if you set up two stands?

  3. Suppose A and C are female dorms and D, E, and F are male dorms. How would your location change if 30 percent of females and 80 percent of males are likely to eat at your stand?

  4. Suppose the path between B and C and the path between E and D go uphill and that it is twice as hard to walk uphill as downhill. How would your choice change?

Proposed solution

We propose an algorithm that determines the most convenient location as the position that minimizes the distance between the dormitories and the hot dog vendor location.

Map representing the college campus, where A, C, D, E, and F correspond to a dormitory. A and C are female dorms, and D, E, and F are male dorms.


The proposed model assumes that:

  1. The closer the hot dog stand is to the student’s dorm, the more convenient the hot dog stand is.

  2. The more traffic encountered by the hot dog stand, the more business it will receive.

  3. Each dorm has the same number of students.

  4. Every student wants to go to each dorm equally.

  5. Students will tend to choose the shortest path to their desired dorm.

  6. Students are always hungry.

  7. The number of students in each dorm is the same.

  8. The same number of students from each dorm will go out.

  9. The most traffic will be at the intersections rather than along a path.


To find the point with the greatest convenience for students, a factor \(c\) was calculated, called the “convenience factor.” This convenience factor was defined as the overlaps of paths divided by the average distance between the dormitories, as given by the equation: \(c=\frac{t}{a}\). Greater intersection traffic is directly proportional and greater average distance indirectly proportional to the convenience factor. This equation controls for variations in modes of transportations and walking speeds: it determines the ease of traveling to vendor position in an absolute sense.

Defining variables

\(t\) was defined as intersection traffic, or the overlaps of the paths between the dorms, calculated as the number of times in which a point coincided with a path ( \(\overline{AB}\), \(\overline{CD}\), and so forth). \(a\) was defined as average distance from each dorm to that location, calculated as an arithmetic mean: the sum of the shortest distances between the dorms and a point, divided by the number of paths.

Application of model

Baseline vender location

We were asked to determine the best location for the hot dog vendor on the map given. For this baseline, calculations for \(t\), \(a\), and \(c\) are shown in Tables 1, 2, and 3, respectively (Appendix). In this system, as \(E\) had the greatest convenience factor (\(c=1.5\)), \(E\) is recommended as the vendor location which optimizes student convenience.

Multiple vendor locations

If multiple vending locations are added, the efficiency of the system is improved, as students will eat at the nearest vendor location of two. To evaluate this system, the top four performing intersections (\(E\), \(D\), and \(B\)) from the baseline model were conjugated as a permutation: \(E\) with \(D\), \(D\) with \(B\), \(C\) with \(B\), \(E\) with \(B\), and so forth). Intersection traffic was determined as the number of non-redundant events during which a path between dormitories coincided with either position. Event were considered redundant if a path coincided with both positions. In this case, only one event would be counted. Table 4 summarizes these calculations (Appendix).

For the same reason, average distances were calculated as the distances from each dormitory to the corresponding closest vendor location. The resulting \(c\) values are shown in Table 5 (Appendix). In this system, \(B\) and \(D\) (\(c=3.2\)) were found to have the greatest convenience factor, and therefore these points are recommended as the optimal vendor positions.

Gender preference

It was assumed in the question that dormitories corresponding to points \(A\) and \(C\) were female, while \(D\), \(E\), and \(F\) were male dormitories. It was also given that \(80 \%\) of males and \(30 \%\) of females purchase hot dogs. In this scenario, several modifications to the baseline model is called for.

Intersection traffic was calculated as \(t=\sum\nolimits_{} g_i\), where \(g\) is gender value. The gender value is calculated as \(.8\) for a path from a male dormitory to another male dormitory and \(.3\) for female to female. Since traffic going both ways of a path is equal, the value was given as an arithmetic mean of purchase likelihoods: \(.55\) for male to female dormitory paths and vice versa. Since location \(B\) does not correspond to a dormitory, it was not included in these calculations. The gender values for each dormitory and the corresponding \(t\) values are given in Tables 4 and 5 (Appendix). As the average distances did not change, the baseline \(a\) values were obtained from Table 2. The resulting calculated \(c\) values are summarized in Table 6 (Appendix).

It was found that location \(E\) (\(c=.6750\)) was best suited in the new environment. This is intuitive given that it is nearest to a male dormitory. In addition, \(E\) is the point closest to the midpoint of the line \(\overline{DF}\), which approximates an alignment of the male dormitories. This is important because male students in this system are more than twice as likely as female dormitory to purchase hot dogs.


We were also tasked with accounting for a change in topography: how would optimal location change if the paths \(\overline{BC}\) and \(\overline{DE}\) go uphill? We found that a change in topography does not alter \(t\), \(a\), or, accordingly, \(c\) values. Rather, since students are equally likely to travel from one dormitory to the other, the net effect \(\Delta c\) is \(0\).