What it can tell us, where it can mislead us, and how to choose the right research method
Used within rigorous research, it can be one of our most valuable tools.
The key is to understand what the data observe, who may be missing or systematically less visible, how estimates are produced, whose absence may matter to a decision, and when other methods are needed.
Mobile location data has changed what tourism and recreation managers can see. Where traditional surveys provide intermittent snapshots, MLD offers continuous, high-resolution observation. Suddenly, it is possible to track crowding at specific beaches, identify where day-trippers are coming from, and measure the impact of an event on a whole region—all without deploying a single clipboard.
That capability is genuinely useful. The concern is not that mobile location analytics, or MLA, is somehow unreliable by definition. The concern is that polished maps and precise-looking estimates can encourage stronger conclusions than the underlying data can support.
A device is not automatically a person. A location point is not automatically a visit. An undetected device is not proof that nobody was there. And a large dataset is not automatically representative.
The practical question is therefore not “Is MLA good or bad?” It is: “What can this particular dataset validly tell us about this particular research question, in this particular place?”
Mobile location data (MLD) is an umbrella term, not a single technology. At least three source families commonly appear in tourism and recreation work:
MLD is the underlying digital trace data. MLA is the analytical process that turns those traces into inferred visits, visitor classifications, maps, flows, trends, and estimates. Many of the most important limitations arise during classification and modelling, not only during location measurement.
A device is not automatically a person. A location point is not automatically a visit. An undetected device is not proof that nobody was there. And a large dataset is not automatically representative.
The practical question is not whether MLA works perfectly, but how to use it responsibly.
These sources do not observe the same people, at the same frequency, with the same precision, or under the same inclusion and privacy conditions. That distinction matters. For example, declining an app permission can affect app-based location data, but it should not be generalized to MNO data, which have a different observation mechanism.
Tourism and recreation research has persistent data gaps. Accommodation statistics miss many day visitors and some forms of lodging. Surveys can be costly, periodic, and affected by sampling and nonresponse error. Trail counters measure passages at specific points but usually cannot identify origins or broader movement patterns. Interviews provide depth, but not population-level estimates.
MLA can fill part of that gap. It is particularly strong when the question concerns broad movement, timing, origin-destination relationships, spatial concentration, or change over time. Recent tourism research continues to expand these applications, including work on tourist tracking, urban spatial choice, protected-area monitoring, and official tourism statistics (Hardy & Shoval, 2025; OECD, 2025a; Costa et al., 2026; Zamengo et al., 2026).
The right conclusion is not to use less mobile data. It is to use it for the questions it is well suited to answer, and to avoid asking it to stand in for information that it never directly observed.
Imagine a mountain biking or hiking trail running parallel to a highway. A geofence around the trail may include legitimate trail users, but it may also include people driving along the highway, stopping at a viewpoint, travelling on a forest service road, visiting a nearby property, or using an adjacent recreation site. Good providers can apply speed, dwell-time, trajectory, repeat-observation, and other filters. Those methods can help considerably.
But this is not only a positioning-accuracy problem. It is also a semantic classification problem. Even a fairly precise coordinate may not reveal why somebody was in an overlapping space or which activity they were undertaking. The same challenge appears when back roads cross trail networks, trailheads sit beside neighbourhoods, ferry terminals overlap commercial areas, or multiple attractions occupy the same compact area.
Municipal lines, park polygons, tourism regions, and trail buffers look exact in GIS. Visitor behaviour does not. A visitor may sleep outside a boundary and spend most of the day inside it. A road may cross repeatedly into and out of a polygon. A small change in a geofence can materially change the number of devices included in a compact community. For MNO data, network cells also do not align neatly with tourism boundaries. This is why smaller and more complex geographies generally require greater care in interpretation.
Forest canopy, terrain, and connectivity are source-specific issues. For app-based data that depend on GNSS-derived positions, trees, terrain, cliffs, buildings, and signal reflection can reduce positioning quality. GPS.gov notes that smartphone accuracy is typically much better under open sky and worsens near trees and structures, while a 2026 smartphone study found materially larger positioning errors under foliage than without foliage (Japertas & Mažeika, 2026).
One technical point is important: no cellular service does not automatically mean no GNSS/GPS position. A phone can calculate a satellite-based position without an active mobile-data connection. But the full commercial data pipeline may still depend on the application, background permissions, caching, upload timing, network availability, and vendor processing. MNO data behave differently again because the location observations come from network interactions. “Coverage” therefore needs to be defined, not assumed.
| Coverage type | What it asks |
|---|---|
| Positioning | Can the device establish a sufficiently accurate location in this physical environment? |
| Network / transmission | Can relevant observations reach the system collecting or processing them, immediately or later? |
| Panel | Does the provider observe enough eligible devices or SIMs in this geography and time period? |
| Population | Do the observable devices adequately represent the people whose behaviour we want to understand? |
Most managers do not see raw observations. They see the end of a processing chain. A typical path may look like this:
From trace to estimate
Person or traveller → device / SIM → location observation → usable record → inferred visit → resident / visitor / pass-through classification → weighting or expansion → dashboard estimate
The most important limitation may not be technical at all. Mobile datasets represent observable devices or subscriptions, not automatically the full visitor population. Depending on the source, some people have a much higher probability of appearing in the data than others.
For app-based panels, inclusion may depend on carrying an eligible smartphone, using participating apps, granting location permissions, allowing background collection, and generating enough usable observations. For MNO data, coverage may be broader, but it can still depend on which operators are included, roaming arrangements, network events, multi-SIM ownership, shared devices, and the method used to expand subscribers to the population. Crowdsourced recreation apps introduce another layer of self-selection because users choose both the platform and whether to record an activity.
This can create unequal statistical visibility. Children may participate without carrying an individual device. A family of five may generate only one or two observable devices. Some privacy-conscious visitors may limit app-based tracking. Some people leave phones behind during recreation. Others may be highly observable because they carry several devices or generate many location events. Devices, observed devices, visits, visitors, and people are therefore different units of analysis.
Why this matters beyond statistical accuracy. Destination data increasingly inform investment, access, infrastructure, services, marketing, and monitoring. If a place or population appears lightly used partly because it is less observable, and managers treat that signal as complete, a measurement blind spot can become a planning blind spot. The risk is not that MLA automatically creates inequity. It is that unequal observability can shape which needs are most visible in the evidence used to set priorities.
There is a deeper limitation: people who do not make a trip usually generate no mobility trace at the destination. Someone may be absent because of transportation, cost, accessibility, safety, information, cultural fit, or other participation barriers. MLA can describe observed movement, but it usually cannot measure unmet or suppressed demand or explain why a person chose not to visit. If the management question is “Who are we not reaching, and why?”, surveys, interviews, community engagement, and targeted consultation become essential.
Privacy creates a related issue, especially for app-based data. A person who limits location permissions or background tracking may become less observable to the provider. That is a legitimate privacy choice, not an analytical defect. Destination planning should therefore avoid treating greater digital traceability as greater importance: people should not have to be highly trackable for their needs to count.
The direction of bias should not be assumed. Recent research has found systematic representation differences associated with demographic, socioeconomic, geographic, and urbanization characteristics in some mobile datasets (Jardel & Delamater, 2024; Mu et al., 2024; Wang et al., 2025). Other work has found strong representativeness in particular app-based datasets after careful home-location estimation (Sinclair et al., 2023). The methodological lesson is more useful than a blanket warning: test representativeness for the dataset, geography, period, and population you actually have.
A 2026 protected-area study illustrates the same principle in a recreation context. Costa et al. (2026) found that mobile fitness-app records at Portugal’s Paiva Walkways correlated with ticket counts and were useful for spatiotemporal monitoring, but the app data did not rise proportionally with ticket counts during higher-pressure periods. The data were useful as a proxy, but not a direct replacement for the established count system.
Methodological principle
Data absence is not automatically visitor absence. A low device count can reflect low visitation, but it can also reflect weak panel coverage, privacy suppression, sparse observations, difficult geography, source-specific inclusion rules, or a mismatch between the observable population and the population of interest. For questions about access and inclusion, people who never make the trip may be the population managers most need to understand.
The availability of millions of records does not remove the basic requirements of valid research. The research question still comes first. Before accepting an MLA result, a defensible study should be able to explain at least five things:
1. The construct: What are we actually trying to measure: trail users, visitors, trips, visits, nights, pass-through movement, or something else?
2. The target population: Who should the result represent? All visitors, tourists, residents, day users, overnight visitors, or a particular segment?
3. The observation mechanism: Who has the opportunity to appear in the dataset, and under what source-specific conditions?
4. The operational rules: How are visits, repeat visits, dwell time, residents, workers, pass-through movement, and boundaries defined?
5. The validation strategy: What independent evidence can test the estimate or interpretation?
This is also why surveys remain important. A well-designed survey is not “older technology” competing with MLA. It is a designed measurement system capable of collecting variables that mobility traces generally cannot observe directly: motivations, visitor characteristics, expenditure, satisfaction, perceived crowding, barriers, accessibility needs, values, attitudes, and management preferences. Surveys have their own sampling, recall, nonresponse, and measurement errors, so they also require rigorous design. The point is complementarity, not methodological hierarchy.
Start with the management question, not the technology. The table below is a practical starting point. "Best starting evidence" does not mean "only method." The appropriate design depends on the consequence of the decision, the population, geography, budget, and required level of confidence.
| Management question | Best starting evidence | Useful complements | Role for MLA |
|---|---|---|---|
| Where are visitors moving? | MLA / GPS-based movement data | Local knowledge; survey for interpretation | Primary |
| When are places busiest? | Counters and/or MLA | Observation; operational data | Primary / strong |
| How many people use a site? | Calibrated counters or validated estimates | Observation; survey; MLA | Useful if validated |
| Where do visitors come from? | MLA plus visitor survey | Accommodation / administrative data | Strong |
| Who are the visitors? | Carefully designed survey | Administrative data; interviews | Supporting |
| Why did they visit? | Survey and/or interviews | Focus groups for shared themes | Limited |
| How much did they spend? | Visitor survey; transaction / business data | Interviews with operators | Limited |
| Were they satisfied or crowded? | Survey; interviews | Observation; focus groups | Context only |
| What barriers or access issues exist? | Survey; interviews; community engagement | Expert / accessibility consultation | Usually limited |
| Who are we not reaching, and why? | Target-population survey; interviews; community engagement | Focus groups; accessibility / expert consultation; secondary demographic data | Very limited |
| What do residents or stakeholders think? | Stakeholder focus groups; interviews; resident survey | Expert consultation; secondary data | Context only |
| Why does an unexpected pattern appear? | Mixed methods | MLA + survey + local / expert interpretation | Often very useful |
| What management action is appropriate? | Integrated evidence | Stakeholder input; expert consultation; monitoring | One input |
Use this as a quick diagnostic before buying data, designing a study, or making a strong claim from an existing dashboard.
1. Is the main question about where, when, movement, origin, or change over time?
If yes, MLA may be a strong fit. If the core question is motivation, experience, spending, barriers, or attitudes, add human-centred research.
2. Is the study geography reasonably distinguishable from highways, homes, overlapping sites, and other uses?
If no, MLA may still be useful, but validation and local spatial interpretation become more important.
3. Do you understand the data source, observed panel, visit rules, weighting, and major exclusions?
If no, treat results as signals rather than strong population estimates until the methodology is clearer.
4. Do you need to understand people, including those who may not appear in the movement data?
If yes, add surveys, interviews, focus groups, or other designed research. MLA observes people who generate usable traces; research with people can explain both observed behaviour and non-participation.
5. Will the result influence a consequential decision such as infrastructure, access, funding, capacity, or equity?
If yes, raise the evidentiary standard and triangulate with independent sources.
Best when the question concerns broad movement, timing, origin-destination patterns, or relative change; geography is suitable; the source and panel are understood; and the decision does not require variables the data cannot observe.
Best when the mobility question fits MLA but the site is spatially complex, the panel is thin, or the classification rules introduce uncertainty. Add counters, observation, administrative data, or local expertise.
Best when you need both movement and information about the people generating that movement. Common combinations are MLA + survey, MLA + counters + survey, or MLA + survey + interviews.
Best when the central question is about motivations, satisfaction, spending, barriers, accessibility, resident perceptions, stakeholder relationships, conflict, or management preferences. Use surveys and qualitative research, with MLA as context where helpful.
What is the original source of the data: MNO/network events, app-based locations, crowdsourced activity data, or a hybrid?
Who has the opportunity to appear in the dataset, what conditions determine inclusion, and which populations may be less observable in ways that matter to this decision?
How many devices or subscriptions are actually observed in my study geography before weighting or expansion?
How frequently are locations observed, and how uneven is that frequency across devices?
What is the realistic spatial accuracy and source-specific limitation in the environment I am studying?
How is a visit defined, including dwell time, repeat visits, residents, workers, and pass-through movement?
How are highways, roads, overlapping geofences, small polygons, and boundary effects handled?
What is observed versus inferred or modelled, including weighting, expansion, imputation, suppression, and classification?
Has the panel, source, privacy environment, algorithm, or methodology changed between the periods I am comparing?
How has the product been validated against independent evidence in settings similar to mine, and what complementary method is still needed?
State the source type, study period, geography, and unit being reported: observed device, inferred visit, estimated visitor, or another modelled quantity.
Describe how visits and visitor categories were operationalized, to the extent the methodology is available.
Avoid false precision. A modelled estimate reported to the nearest individual can communicate more certainty than the model warrants.
Document known panel, privacy-suppression, coverage, weighting, and methodological changes, particularly for comparisons over time.
For access, inclusion, or participation questions, state which populations may not appear in the data and use methods capable of reaching non-visitors where needed.
Separate description from explanation. MLA may show a movement pattern without telling you why it occurred.
For consequential conclusions, show where independent evidence converges or conflicts. Disagreement between methods is often a finding worth investigating, not a nuisance to hide.
Mobile location analytics has opened an important window into visitor mobility. Tourism and recreation managers should use that window. But we should not mistake the view through one window for the entire destination.
For some questions and geographies, MLA can provide exceptionally valuable evidence. For others, a trail counter, well-designed visitor survey, intercept study, transaction dataset, focus group, interview, or expert consultation may be more appropriate. Often the strongest design combines several sources deliberately rather than adding methods only after problems appear.
The practical standard is straightforward: define the question first; understand how the data were generated; distinguish observation from inference; examine who is represented, who may be less visible, and who may be absent because they are not participating at all; validate important findings; match the research design to the consequence of the decision; and report uncertainty honestly.
Mobile location analytics is not a replacement for rigorous tourism research. Used within rigorous tourism research, it can be one of our most valuable tools.
Costa, J., Gomes, A., Stokes, M., & Saraiva, M. (2026). Enhancing monitoring in protected areas: Evaluating mobile apps as a tourism proxy. Tourism Geographies, 28(3), 865-886. https://doi.org/10.1080/14616688.2026.2671422
Creswell, J. W., & Plano Clark, V. L. (2025). Designing and conducting mixed methods research (4th ed.). SAGE Publications. https://us.sagepub.com/en-us/nam/designing-and-conducting-mixed-methods-research/book269004
Dattilo, B., Di Torrice, M., & Sabato, M. (2026). From signals to statistics: Measuring same-day visits with MNO data. An Italian use case. Statistical Journal of the IAOS. Advance online publication. https://doi.org/10.1177/18747655261480867
Gong, S., Saadi, I., Teller, J., & Cools, M. (2025). A comparative analysis of mobile phone data and travel surveys in understanding travel behaviour. Transportation. Advance online publication. https://doi.org/10.1007/s11116-025-10708-4
Grassini, L., & Dugheri, G. (2021). Mobile phone data and tourism statistics: A broken promise? National Accounting Review, 3(1), 50-68. https://doi.org/10.3934/NAR.2021002
Hardy, A., & Shoval, N. (2025). 25 years of tourist tracking: A geographical perspective. Tourism Geographies, 27(3-4), 851-862. https://doi.org/10.1080/14616688.2025.2462222
Japertas, S., & Mažeika, M. (2026). The influence of tree canopy on GPS positioning accuracy. Scientific Reports. Advance online publication. https://doi.org/10.1038/s41598-026-68024-5
Jardel, H. V., & Delamater, P. L. (2024). Uncovering representation bias in large-scale cellular phone-based data: A case study in North Carolina. Geographical Analysis, 56(4), 723-745. https://doi.org/10.1111/gean.12399
Li, Y., Yao, E., Yang, Y., & Li, B. (2025). Understanding tourism travel behavior by combining revealed preference survey and mobile phone data. Transportation Research Part A: Policy and Practice, 194, 104408. https://doi.org/10.1016/j.tra.2025.104408
Mu, X., Zhang, X., Yeh, A. G.-O., & Wang, J. (2024). Evaluating the representativeness of mobile big data: A comparative analysis between China's mobile big data and census data at the county level. Applied Geography, 166, 103260. https://doi.org/10.1016/j.apgeog.2024.103260
OECD. (2025a). Using alternative data sources and tools to measure and monitor tourism. OECD Tourism Papers, No. 2025/11. OECD Publishing. https://doi.org/10.1787/09ad7240-en
OECD. (2025b). Using mobility data to measure domestic tourism movement in Australia. In Using alternative data sources and tools to measure and monitor tourism. OECD. https://www.oecd.org/en/publications/using-alternative-data-sources-and-tools-to-measure-and-monitor-tourism_2a655ab8-en/using-mobility-data-to-measure-domestic-tourism-movement-in-australia_1c57dd79-en.html
Reif, J., & Schmücker, D. (2020). Exploring new ways of visitor tracking using big data sources: Opportunities and limits of passive mobile data for tourism. Journal of Destination Marketing & Management, 18, 100481. https://doi.org/10.1016/j.jdmm.2020.100481
Saluveer, E., Raun, J., Tiru, M., Altin, L., Kroon, J., Snitsarenko, T., Aasa, A., & Silm, S. (2020). Methodological framework for producing national tourism statistics from mobile positioning data. Annals of Tourism Research, 81, 102895. https://doi.org/10.1016/j.annals.2020.102895
Sinclair, M., Maadi, S., Zhao, Q., Hong, J., Ghermandi, A., & Bailey, N. (2023). Assessing the socio-demographic representativeness of mobile phone application data. Applied Geography, 158, 102997. https://doi.org/10.1016/j.apgeog.2023.102997
Wang, Y., Guan, X., Ugurel, E., Chen, C., Huang, S., & Wang, Q. R. (2025). Exploring biases in travel behavior patterns in big passively generated mobile data from 11 U.S. cities. Journal of Transport Geography, 123, 104108. https://doi.org/10.1016/j.jtrangeo.2024.104108
Zamengo, B., Rossi, E., Zaramella, A., Mancino, L., Cappellari, D., & Di Sorte, D. (2026). Integrating Mobile Network Operator data into tourism statistics: Operationalizing the usual environment and evidence from Venice. Statistical Journal of the IAOS. Advance online publication. https://doi.org/10.1177/18747655261491187
Dr. Farhad Moghimehfar is the BC Regional Innovation Chair in Tourism & Sustainable Rural Development at Vancouver Island University and founder of Destintel Innovations Inc. His work focuses on tourism, outdoor recreation, rural development, visitor-use research, and decision-support systems that combine rigorous research with practical destination intelligence.