Data and Parameters
Which data providers are used?
MobileScapes uses a combination of cellular network data and 4 combined sources of permission-based, de-identified mobile movement data from apps. Cellular network data is blended with app data for trend analysis. These mobile data are sourced from trusted suppliers who meet data quality and privacy compliance standards.
How is cellular network data used?
Cellular networks improve the stability and sample size of our data, increasing the accuracy in weekly, monthly, and yearly trending analyses. Cellular network data is provided to EA in an aggregated and de-identified format. No Personally Identifiable Information (PII) is included.
Our Cell Network Coefficient is applied to data weekly at the geofence level.
What is the spatial coverage of cell towers?
Cell towers have national coverage and subscriber data represents approximately 1/3 of the Canadian population.
What time period can I get data for?
Data pulls are available from January 1st, 2021. We only carry data for the current year and the three preceding years. For example, in 2025, we will retain data from January 2022 through 2025.
Data are updated daily and posted every 10 days.
A single extract will be limited to a maximum of 2 years of data.
Why is new data available after 10 days instead of right away?
Our commitment is to deliver the highest quality and integrity of data in order to maintain the accuracy and reliability of our estimates. Users will experience a 10-day lag for MobileScapes data pulls.
For instance, data from October 8, 2024 will be available on October 18, 2024. This adjustment ensures you receive complete and reliable data for your analysis.
Unweighted vs. weighted data
Weighted data will be the default. Weighting is to Household Population 15+. Users will have a setting option to retrieve unweighted data. While there is an option for unweighted counts, it is not recommended unless used for Profile or Distance Decay purposes. The trend analysis will be significantly different from the weighted counts.
Weighting allows you to produce population estimates of visitors to geofenced areas. In addition to total visitors, weighted estimates are also created for visits to produce foot traffic estimates.
Is there any way to discern the margin of error in relation to weighted estimates?
Mobile movement data are based on what we refer to as “Big Data”. The concept of a margin of error—that researchers are familiar with in sample surveys—does not apply to this type of data. The frequency of pings and number of devices with location services turned on require weighting and normalization.
EA data scientists have spent several years in R&D to understand the coverage and bias—both spatially and temporally—and they have developed routines using best practices to remove outliers, understand the bias and weight data to general population. It’s not possible to provide a statistical margin of error such as would be associated with a sample survey.
In several real-world tests, our normalized and weighted data provide results closer in line with reality than what resulted from using raw ping data. The introduction of cellular network data improves the accuracy of our estimates through better coverage and representation, less volatility, and enhanced reflection of trends.
Ping observations from permission-based apps
We capture a single ping per day, and that ping is the first ping of the day. For example, if we see pings at 10:00 AM and 2:00 PM on the same day, we only count one ping for the day and for daypart purposes it will be attributed to the 9AM – noon daypart.
We take the first ping because we want to display when the device first arrived in the geofence.
When multiple geofences are aggregated in a single extract, it only captures the first observation in the first geofence visited.
Multiple geofences can be viewed independently in a single extract. When selecting multiple geofences in ENVISION, there is an “Individual” option available.
Are the data privacy-compliant?
Yes. EA MobileScapes data is modelled from permission-based data collected by our trusted suppliers, using location-enabled apps. Our suppliers de-identify the data it is sent to Environics Analytics. The data are combined with cellular network data for trend analysis and then used as inputs to models that normalize and weight to the general population. EA’s expert methodologists use best practices, external controls, and advanced modelling techniques to address bias and small sample size issues.
EA’s Privacy by Design ISO 31700-1 certification validates that our company fully integrates privacy and data protection principles into the design and development of products, services, and systems from the outset.
Geofences
How can I create geofences for my extracts?
Geofences can be hand-drawn on a map in the MobileScapes Data Extract tool, or you can import spatial area polygons (when importing, turn on the Geofence toggle). We have also created an EA Geofence Library, a layer of pre-defined geofences, that can be accessed within the ENVISION MobileScapes Data Extract tool.
Geofence size limitation
The maximum geofence size limit is 5 million square feet. This limit only applies to your hand-drawn or imported geofences. The limit will not apply to EA geofences.
Multiple EA geofences can be selected for a single extract. For user-defined geofences, you may use one geofence per extract.
EA Geofence Library
EA has undertaken a large-scale effort to build a vast layer of polygons for Canadian commercial, institutional and recreational locations. The polygons can be used as geofences for destination-based mobile movement extracts both within ENVISION and for project deliverables. In ENVISION, the library can be accessed using the Geofence Explorer within the MobileScapes Data Extract tool. The geofences are not available for sale outside of the MobileScapes products.
EA relies on our industry knowledge, a large team of analysts and many relationships with partners to identify and classify locations. All data used to validate our process and fill in gaps is used with permission.
What is the difference between shared and unique geofences?
Unique: Locations that can be spatially distinguished from other locations have their own geofence, which can include multi-story or single-story buildings.
Shared: Geofences of locations within a multi-story or multi-use location (i.e. commercial and residential) are categorized as shared.
How do you deal with multi-floor geofences?
MobileScapes cannot distinguish between different levels within multi-tiered buildings. The MobileScapes Extract outputs will capture data from all levels within the desired geofence due to the absence of altitude capture.
How will the data be backfilled if a geofence undergoes a banner change? For example, what would happen to the estimates if Tim Horton's changed to a Starbucks midway through 2023?
The count of visits is based on the geofence boundary lines, not what banner it belongs to. The banner name in the geofence library will reflect any identified changes, but visits will be pulled for the geofence going back to Jan 2021, even though it belonged to another banner at some point in the past.
How does the integration of cellular network data affect geofence counts?
Due to the many characteristics of geofences, such as time of day, day of the week, and typical stay duration, we create a signature for each one. This helps us assign a likelihood that a device on the cell network visited that location. We also use permission-based device data to proportionally assign weights between different locations within the same cellular area, in addition to the previously described attributes. This ensures we are only using cell data relevant to that geofence.
For example, for a geofence that includes a highway, using cellular network data weeds out traffic that just passes through (i.e. highway drivers), because the devices would not meet the minimum stay duration within the geofence.
Why are the results from my custom geofence different from the licensed geofence even though it has the same spatial footprint?
Cellular network data is integrated with ping data weekly at the geofence level. The modelling process for this data uses attributes from licensed geofences to better understand typical visit durations, times of day, and category-specific information. This ensures that only the most relevant cellular data related to the geofence are used for adjustments. Custom geofences lack the detailed attributes of licensed geofences, so default values must be applied. This can result in slight differences in estimates, even if the spatial footprint is identical to a licensed geofence. Therefore, we recommend using licensed geofences whenever possible to ensure accuracy.
Can I evaluate visits to a location that has just recently opened or changed its primary use?
Evaluating brand new locations that were previously open land or have changed their primary use presents some challenges. Our cell network models are designed to ensure that cell tower data is appropriate for each specific geofence. However, in these cases, the cell data might not accurately reflect the geofence’s current primary use.
Outputs
What mobile movement insights are delivered with each extract?
Daily Visits and Unique Visitors will be returned along with the Common Evening Location (CEL), Common Daytime Location (CDL), days, dayparts and months.
Visit metrics can be represented as an “estimated Household Population 15+” weight. They can be thought of as total foot traffic for the geofence over the selected time period.
Visitor metrics can also be represented as an “estimated Household Population 15+” weight. They filter out multiple visits to give a true indication of the unique population identified within the geofence over the selected time period.
International devices
International devices will not appear when pulling a Canadian mobile movement extract within ENVISION.
Methodology for assigning Common Evening Location (CEL) and Common Daytime Location (CDL)
Common Evening Location (CEL) is where a device is most commonly observed between 6PM and 8AM and the Common Daytime Location (CDL) is where a device is most commonly observed between 9AM and 5PM.
A common location is defined by a 100m x 100m cell (third decimal latitude and longitude point) which has the highest score within a specific time window. Daily Common Location would use a time window of 24 hours starting at 6pm local time of the previous date, and Monthly Common Location would use a period of 6 weeks prior to the month's end.
The score used to rank individual cells for Common Location is calculated as the number of unique hours a device is seen in a cell, multiplied by the number of unique days that the device was observed. For example, if a device was observed each hour between 9am and 5pm over the course of 15 days during the respective monthly date range, then this device would receive a score of (8 unique hours) X (15 unique days) = 120.
CEL and CDL are then assigned to a dissemination area (DA) which contains the highest scoring cell, and the postal code is assigned per each device using a combination of Privacy by Design rules and pseudo-households.
Can I understand which visitors are workers or residents in the geofence?
No. EA adheres to strict Privacy by Design guidelines, meaning we cannot reliably estimate which visitors are workers or residents in the geofence.
Why are there records with a weight of 0 in my extract?
Records with weights of 0 represent devices that do not have a sufficient number of observations in a day to be considered a “good quality device.”. We have maintained these records in the extract to allow you to understand the movement of these low-quality devices for any additional ad-hoc analysis you wish to complete.
Why are there records without a postal assignment?
This may be blank due to a device not having a sufficient historical CEL/CDL assignment in the database. These devices were not observed frequently enough within the CEL or CDL assignment window to appropriately assign the device to a location.
How can I calculate weighted visits per weighted visitor?
You can divide weighted visits by weighted visitors per record. Due to the bias within the underlying data, it will have a slight positive bias to it. This is due to visitors who only visited once as they may be undercounted.
Will I receive a notification if the sample size for a MobileScapes extract is low?
Yes, if the number of unique devices in a MobileScapes extract is less than 450, you will be notified via an onscreen alert in ENVISION. A total record count below 450 does not meet our minimum recommended sample size for reliable segment profiling. If you receive the alert, please consider increasing the date/time range for the affected geofence(s), or aggregating multiple geofences to increase the record count. Otherwise, consider limiting your profiles to the Target Group, Social Group or Lifestage Group level.
To discuss the implications of using geofences with low sample sizes, please contact your Environics Analytics account representative or support@environicsanalytics.com.
How can I prevent low sample warnings?
Low sample warnings appear when either a location does not get a lot of foot traffic, when the area being analyzed is too small, or when the timeframe chosen is too short. Increasing the area, you are looking at or increasing the timeframe will often reduce the likelihood of these warnings.
Utilizing MobileScapes for single-event analysis, such as monitoring foot traffic during specific sporting events at stadiums or venues on particular dates and times, is discouraged. This is mainly because longer date ranges yield richer data sets, enhancing the accuracy of our estimates, while shorter durations result in smaller sample sizes.
How are credits consumed?
Aggregate: If multiple geofences are selected for an aggregated extract, each group of 20 geofences counts as 1 query. Example: 100 geofences count as five queries. (Note: Queries will always round up; for example, 21 geofences in an aggregate extract will use two queries).
Individual: This option allows you to save time by not having to run a separate extract for each unique geofence. If multiple geofences are selected for an individual extract, each geofence counts as 1 query. Example: 100 geofences count as 100 queries.
Why should I avoid using just numbers, the "&" sign, or "/" backslash, when naming extracts?
Sticking to letters, underscores, and dashes is highly recommended when naming extracts. Users might encounter an issue caused by extra spacing or special characters in the extract name, which hinder running reports related to the extract, such as the MobileScapes Data Extract Summary Tool report. The issue can only be resolved by re-running the data extract with the recommended naming convention. Each time you run an extract, it will be counted against your total queries.
Usage Scenarios
Where can the mobile movement extract be found in ENVISION?
Mobile movement extracts are found in the “Customers” section of My Data.
When selecting multiple geofences during an extract process, why choose to aggregate them into a single polygon?
When you aggregate multiple geofences, ENVISION will automatically deduplicate the file so you can analyze true unique visitors If your use case is to understand the true unique observed population with no overlap or duplication of devices, this is the best option to choose.
When selecting multiple geofences during an extract process, why treat them as individual geofences?
This option allows you to save time from having to run a separate extract for each unique geofence. However, when analyzing these types of extracts, such as creating a profile or running a MobileScapes Trend report, we recommend you use the filter to select a single geofence at a time, and then run your analysis. If you create a profile or run a MobileScapes Trend report using multiple geofences from a single extract, you risk introducing duplication into your results due to possible double counting of devices seen in multiple geofences.
When I select multiple geofences during an extract process, and then select the individual option, can I create a profile or MobileScapes Trend report for the entire extract?
In this scenario we recommend using a filter to select a single geofence for each profile you create, or each MobileScapes Trend report you run. This will prevent double counting of devices that may be seen in multiple geofences. If you create a profile or run a MobileScapes Trend report using multiple geofences from a single extract, you risk introducing duplication into your results.
If I want to understand unique visitors only for a specific day, daypart or month when running a weighted extract, what is the best way to achieve that?
In this scenario, we recommend selecting the “aggregate geofence” extract option, then selecting the specific day, daypart or month during the extract process.
If you wish to understand visits for a specific day, daypart or month, you could run an entire timeframe for an “aggregate geofence” extract, and then when creating a profile of the extract in ENVISION, select the specific day, daypart or month fields for the aggregated geofence. You will get a profile of visits in this case.
What if my location is open during the night, can I select hours that span between two days? For example, 10 PM to 2 AM Friday/Saturday.
When creating your data extract, you have the flexibility to choose any time period that falls within a single day. For instance, if you select a time frame of 4PM-10PM, the extract will include all observations made during that time on every day you've chosen.
However, it's important to note that you can't select a time range that spans across two different days. For example, if you try to create an extract of just four hours from 10PM to 2AM that covers both Day 1 and Day 2, it won't be possible. Instead, you will need to keep your selection within a 24-hour period.
I have noticed a difference in visit counts between FootFall and MobileScapes. Which one should I use?
Results between the two should be similar, however slight differences in methodologies may result in some differences in the reported visit count between FootFall and MobileScapes for the same geofence and timeframe.