Source-linked AI summary
An analysis of visitors' behavior in the Louvre Museum: A study using Bluetooth data
Yuji Yoshimura, Stanislav Sobolevsky, Carlo Ratti, Fabien Girardin, Juan Pablo Carrascal, Josep Blat, Roberta Sinatra
TL;DR
Museums need better evidence about visitor movement to address hyper-congestion and protect visiting conditions. This paper analyzes Louvre visitors’ sequential movement and spatial layout using large-scale Bluetooth observations, finding that short- and long-stay visitors use broadly similar numbers and patterns of key locations. These constrained patterns may contribute to uneven visitor distributions and inform more dynamic flow management.
Problem
Museums can become overcrowded when visitor numbers exceed spatial capacity, while existing movement studies often use limited spatial and temporal data.
Method
The paper uses systematic Bluetooth proximity detection to analyze Louvre visitors’ sequential movement, spatial layout, and routes across the museum.
Results
Short- and long-stay visitors visit similar numbers of key locations, while longer-stay visitors tend to explore them more extensively and frequently follow similar paths.
Takeaways & Limitations
The constrained and recurring movement patterns may produce uneven visitor distributions, supporting more dynamic management of museum flow and congestion.
Takeaways & Limitations
Bluetooth records transitions between sensor nodes rather than complete trajectories and cannot directly capture visitors’ introspective factors, wayfinding, or orientation.
Abstract
from arXiv · showhide
Museums often suffer from so-called "hyper-congestion", wherein the number of visitors exceeds the capacity of the physical space of the museum. This can potentially deteriorate the quality of visitor's experience disturbed by other visitors' behaviors and presences. Although this situation can be mitigated by managing visitors' flow between spaces, a detailed analysis of the visitor's movement is required to fully realize and apply a proper solution to the problem. This paper analyzes the visitor's sequential movements, the spatial layout, and the relationship between them in large-scale art museums - Louvre Museum - using anonymized data collected through noninvasive Bluetooth sensors. This enables us to unveil some features of visitor's behavior and spatial impact that shed some light on the mechanism of the museum overcrowding. The analysis reveals that the visiting style of short and long stay visitors are not as significantly different as one could expect. Both types of visitors tend to visit a similar number of key locations in the museum while the longer stay type visitors just tend to do so more extensively. In addition, we reveal that some ways of exploring the museum appear frequently for both types of visitors, although long stay type visitors might be expected to diversify much more given the greater time spent in the museum. We suggest that these similarities/dissimilarities make for an uneven distribution of the quantity of visitors in the museum space. The findings increase the understanding of the unknown behaviors of visitors, which is key to improve the museum's environment and visiting experience.
1. Mesoscopic research of visitors’ sequential movement in Art Museum
Museums face hyper-congestion because visitor numbers can exceed spatial capacity, motivating analysis of sequential movement and spatial layout. This study uses large-scale Bluetooth observations to examine how stay duration and visit order relate to circulation and visiting conditions.
- Hyper-congestion occurs when visitor numbers exceed museum capacity, potentially degrading visiting conditions and visitor experience.
- Prior museum studies commonly examine either broad visitor composition or circulation in limited rooms and areas.
- Existing studies often rely on spatially and temporally limited data, while simulation-based analyses simplify human behavior rather than revealing actual movement.
- The paper analyzes Louvre visitors’ circulation from entrance to exit as a whole mobility network, relating stay duration and visiting sequence to experience.
- Bluetooth proximity detection produces large-scale datasets representing sequential movement, enabling analysis of global behavioral patterns at low spatial resolution.
- The analysis excludes introspective factors such as learning and meaning-making, focusing instead on physical presence and movement between places.
2. Visitor’s sequential movement and analysis framework
The framework combines Bluetooth-based systematic observation with prior approaches to visitor mobility and explicitly defines the method’s scope. It estimates routes and presence while acknowledging limitations in trajectory detail and introspective interpretation.
- Bluetooth proximity detection provides an unobtrusive way to study museum visitors’ mobility without requiring prior registration or device equipment.
- The study identifies stay length as an indicator of interest level and estimates visitor routes between sensors and visitor quantities at locations.
- The method records timestamped transitions between sensor nodes rather than complete device trajectories, so inferred paths depend on the museum’s spatial network.
- The analysis cannot directly address visitors’ expectations, experiences, satisfactions, wayfinding, or orientation.
- Visitor-device representativeness was assessed through a month-long comparison of detected entrance devices with official head counts and ticket sales.
3. Concept Definitions and Data Settings
The dataset was collected during a specific period and processed into a form suitable for analyzing visitor movement and method consistency.
- The study defines sensor locations and dataset components before processing the collected data for analysis.
3-1. Sensors settings in museum and definition of node
Seven sensors were deployed across key Louvre locations along a major visitor trail. Each sensor defines a node through an approximately bounded detection area for tracking Bluetooth-equipped devices.
- Seven sensors cover key locations along a busy Louvre trail leading from the entrance toward the Venus de Milo.
- The sensor network includes Hall E, Gallery Daru D, Venus de Milo V, Salle des Caryatides C, Great Gallery B, Victory of Samothrace S, and Salle des Verres G.
- Each sensor’s detectable node is approximately 20 meters long and 7 meters wide, with coverage varying by museum setting.
- A sensor registers when a Bluetooth device enters its detectable area and continues recording until the emitted signal disappears.
3-3. Collected Sample
The study analyzed 24,452 uniquely detected devices from complete Louvre visits collected across 24 days. The transformed dataset recorded each visitor’s path, check-in and check-out times, and total stay length.
- 24,452 unique devices were selected from data collected over 24 days in 2010.Bluetooth was active on average for 8.2% of visitors’ mobile devices in the Louvre.
- Only visitors starting and finishing at node E were retained so their complete museum stay could be measured.These records indicate that visitors were registered entering, moving inside, and leaving the museum.
- Each logged movement records detected nodes, node-specific stay times, and travel times between corresponding nodes.For example, the sequence E-S-D-E includes stays at each detected node and travel times for E-S, S-D, and D-E.
- The transformed dataset contains one entry per visitor with visit date, museum path, check-in, check-out, and total stay length.
3-4. Partitioning of Visitors
Visitors were partitioned into equally sized groups according to total time spent in the museum, with the first and tenth deciles defining the short- and long-stay groups.
- 24,452 visits were sorted by total museum time and divided into deciles of approximately 2,446 visits each.
- The first decile was defined as short visits, while the tenth decile was defined as long visits.
4. Results
The results analyze visitors’ path sequences and lengths, the frequency of individual paths, and similarities and differences between shorter- and longer-stay visitors.
- The analysis examines path sequence length, path length, and the frequency with which each visitor path appears.Path sequence length counts visited nodes, including repeated visits and excluding node E.
- The statistical analysis identifies visiting patterns and compares the behaviors of longer- and shorter-stay visitors.
4-1. Basic statistics of visitors’ behavior
Visitors’ path sequence length and node visitation patterns show that longer stays produce only modestly more extensive exploration. Short- and long-stay visitors frequently follow similar routes and visit similar numbers of popular locations, with node G as a notable duration-dependent exception.
- Stay duration: More than 30% of visitors stayed 1–2 hours, while only 1.6% stayed more than 8 hours.Visitors staying less than 1 hour represented only one case.
- Path sequence length: 15.2% of visitors visited one node, whereas only 2.9% visited two nodes.The path-sequence distribution is slightly right-skewed, and sequence length alone does not determine mobility area.
- Node visitation: 97% of visitors passed node S, nearly 80% visited nodes D and B, and only 30% visited node G.Node G was therefore the least frequently visited among the reported nodes.
- Node visitation: Node G visitation increased with stay duration, although it remained below 40% for every visitor type.Short-stay visitors were less likely to visit G, while long-stay visitors appeared more attracted to it.
- Similarity of visitors’ behaviors: Visitors staying 1–2 hours visited 4.3 nodes on average, compared with 5.5 nodes for those staying 3–7 hours.The latter stay duration was three times longer but produced only a 28% increase in sequence length; 9–10-hour visitors averaged 6.6 nodes.
- Similarity of visitors’ behaviors: Short- and long-stay visitors followed similar frequent path lengths and visited similar numbers of popular places, with longer stays exploring them more extensively.For shorter paths, the dominant routes included E-S-E, E-S-B-E, and E-D-S-B-E, with no clear difference between stay types.
6. Discussion
Visitors with shorter and longer stays often follow similar routes and visit similar numbers of key locations, while selective path choices concentrate visitors unevenly across museum spaces.
- Uneven spatial distribution of visitors: 13.5% of visitors visited only the Victory of Samothrace, using the Mollien stairs rather than the shortest route through node D.This route is a spatial and temporal detour from entrance E to node S.
- Uneven spatial distribution of visitors: Almost 40% of visitors followed E-D, whereas around 20% followed E-S, with no significant difference between shorter- and longer-stay visitors.Both groups therefore began their museum experience similarly.
- Visitor route patterns: Among visitors visiting at least four nodes, the most frequent path was E-D-S-B-D-V-C-E, covering several iconic exhibits across the museum.The route entered through Denon and exited through Richelieu or Sully, indicating extensive exploration rather than confinement to one area.
- Visitor route patterns: Short- and long-stay visitors visited similar numbers of popular rooms, while longer-stay visitors explored them more extensively than expected.Their path lengths and unique visited nodes were often not strikingly different or were nearly independent of stay duration.
- Uneven spatial distribution of visitors: Visitors selected a limited set of trajectories despite many possible routes, and these choices were almost independent of stay duration.The frequently used E-D-S-B-D trail may concentrate visitors, while rarely observed transitions such as S-G may leave spaces vacant.
7. Conclusion
The study uses large-scale Bluetooth data to characterize Louvre visitors’ mobility styles and their spatial effects. It finds that short- and long-stay visitors often follow similar patterns, helping explain uneven congestion and vacancy across museum spaces.
- Conclusion: Bluetooth-based analysis of large-scale visitor data describes mobility styles and their spatial impacts in the Louvre Museum.The study uses a bottom-up methodology to examine visitors’ activity and behavior.
- Conclusion: Path lengths grew more slowly than stay duration, while the number of unique visited nodes remained almost constant across stay durations.The reported correlation indicates that unique visited nodes were independent of museum stay duration.
- Conclusion: Visitors who visited fewer than four nodes formed similarly sized groups among short- and long-stay visitors and may have focused on only a few iconic artworks.They may also have lacked motivation or information to explore a larger area.
- Conclusion: Frequent exploration patterns appeared in both visitor groups, producing limited path sequences and potentially uneven distributions of visitors across museum spaces.The authors associate these patterns with congestion in some spaces and vacancies in others.
- Implications: Transition rates and node-movement probabilities might support dynamic visitor-flow management and congestion-aware route suggestions.The paper also suggests using audio guides to alter recommended routes according to sensor-measured congestion.