Our assigned topic is retail, which encompasses the types and locations of shops in Les Halles and its surroundings as well as their relation to streets, metros, and customers. The variable we chose to analyze is store position within Les Halles.
Our initial exploration of Les Halles led us to formulate an assumption: that location may be key for shops’ frequentation (i.e., the number of visitors who visit a shop). The visits we conducted showed us that the mall of Les Halles, consisting in a multilevel square structure, does not offer the same visibility to all the stores it hosts. We identified two main differences that could exist between similar stores:
– Whether the store was located in the corner of a level or in the middle of a hallway;
– Whether the store was located at a directly accessible level (street or metro levels) or an intermediate floor.
By conducting a quantitative analysis of the flow of customers, we hoped to determine whether position (including level and location on a floor) is a key determinant of the flow of customers into a store.
To isolate and test the importance of location, we had to keep all other variables constant. The number of factors/variables determining the frequentation of stores proved this impossible. However, in trying to control variables such as reputation and the type of the store, we came up with a linear regression with different variables.
We grouped stores by category (e.g. fashion, beauty, etc.) and then further divided up these stores by reputation. To grade the reputation, we decided to take a scale from 1 to 5, 5 describing a very renowned brand. We went on the Google Trends website to compare the number of searches for each store in France in 2015 and assigned reputation scores based on the results.
Having dedicated an afternoon to observe and count the number of people going into several stores per 10 minute intervals, we then ran the following linear regression analysis on the computer program STATA :
The first results showed little importance for shops’ location in customers’ turnout, with large standard margin of errors and high p value scores. We acknowledged that the relatively small size of our data sample (we gathered data for 39 stores), the diversity of shops, and the difficulty to control all outside variables may explain this STATA result. However, from purely observing the customers and people watching, we noticed that position did seem to play a role in the number of customers that visit certain stores. The conclusions we drew from the data we gathered proved to be highly relevant on several sub-categories of shops, and on particular stores with interesting characteristics, such as Fnac, which has multiple levels and position entrances.
The following charts provide a comparison of frequentation among different entrances of Fnac, a store with entrances in many positions. Because we are analyzing one store, the store reputation and the type of store are all kept constant.
Fig 1: Comparison of frequentation of customers in Fnac across different store levels and different positions. Stores located on easily accessible 0 or -3 levels saw many more customers than stores located on floors -1 or -2. Stores with entrances in the middle of the floor saw more customers than stores with entrances in corners.
We also analyzed our position variable across similar stores which sell the same products; we first chose to compare shoe stores (all of which have similar reputations) having noticed that the mall hosts a lot of them :
Fig 2. Comparison of frequentation of customers in shoe stores (all with similar reputations) across different store levels and different floor locations. Stores with level code = 1 averaged 27 customers per 10min and level = 0 averaged only 4 customers per 10min. Stores with position code = 1 averaged 20 customers in 10 minutes while position = 0 stores averaged only 3 customers per 10min.
We did the same with fashion stores, this category being the most represented in the mall with a variety of stores with different reputations.
Fig 3: Comparison of frequentation of customers in fashion stores with reputation 1-3 across different store levels and different floor locations. The average frequentation for position = 1 is 24 customers per 10min compared to only 10 customers per 10min for position = 0. For stores level = 1, the average is 32 customers per 10min while it is 8 customers per 10min for level = 0.
Fig 4: Comparison of frequentation of customers in fashion stores with reputation 4-5 across different store levels and different floor locations. For level = 1, the average frequentation is 37 customers per 10min while it is 24 customers per 10min for level = 0. The average frequentation for position = 1 is 8 customers per 10min compared to 38 customers per 10min for position = 0. These results are, however, inconclusive because of the small sample size (we had an extremely limited selection of stores for reputation 4-5 with position = 1).
Even if our quantitative methodology didn’t work out because we couldn’t gather enough data, there is something instinctive about the impact of the position of a store in a mall or a street. Therefore, inspired by the Nakagin Capsule Tower from the Japanese architect Kisho Kurokawa, we thought about making movable shops. By creating “pod shops” that could be moved by a robotic arm for example and be reassembled, the disparity of frequentation between shops because of their position could be lessened. Actually, it could work a bit like the structure designed by the Boston architects Howeler +Yoon and Los Angeles digital designers Squared Design Lab. The only restriction of this model would be that it only applies to above ground buildings.
Source of inspiration: Kurokawa Kisyo, Nakagin Capsule Tower Building (1972)
Cha Cha Yang, Guillaume Zucman, Julio Mendez Cabrera, Julie Hosteing