Geopath Activity Patterns
- Kevin Narine
How are patterns of activity understood at a point of interest?
The Standard Data Visits product provides counts of, well, visits to a place. This is the “foot traffic” or visitation at a location calculated through the sampling of mobile device activity of the U.S. population.
Definitions
A visit is defined as an individual’s arrival and associated dwell time at a place. Below are a few examples of what a visit can be:
a person returns after lunch, arriving at their office where they work for the rest of the day
a person stops at a CVS on their way home from work
an Uber driver pulls up to a hotel and waits for their passenger to come out
a person lands in Charlotte for a three day conference
a person drops their child off at school (two visits, one with a 8min dwell time for the parent and one with a 6hr dwell time for the teenager)
A place is a defined polygon generally outlining a location. For Geopath Geopath Understanding Places/Points of Interest are audited and attributes collected allowing the understanding of the circulation of audience within places.
The dwell time is defined as the length of stay in minutes.
A person type is defined by who is visiting and how long they stay for a specific place.
Resident: An individual who’s home location is in a Census Block Group that touches the place. For a restaurant or retail business, these are people that live “in the neighborhood” (approx. 400 households). Due to privacy commitments, Standard Data will not be more specific than the neighborhood. To be clear, residents of a specific place, i.e. lifestyle centers, will not be identified vs. visitor traffic.
Worker: An individual who is seen 3 or more days per week at the place for more than 3 hours per day.
Visitor: An individual who doesn’t work at the place or live in the neighborhood.
Visits versus Occupancy
For visitor counts to a location where people arrive and spend at least several hours (sporting event or the office), the arrival counts can sometimes feel a bit skewed. For example, it may appear that no one is visiting the office at 3pm because not many people arrive at that time. Therefore, measures of visits and occupancy are both reported:
Occupancy: The average number of individuals at the place over the course of each hour, which will show an intuitive view for how many people were there across the whole event or time period.
Visits: A visit is a unique arrival plus a dwell time. For a large event, you may see a spike of visits at the location early and then seemingly very few throughout the event.
The figure below is a comparison of visits vs. occupancy for Super Bowl Sunday 2019 at Mercedes-Benz Stadium in Atlanta, GA.
Venue Circulation Calculation
Using the refined Geopath Mobile Device Panel, Venue Circulation is calculated using the process below.
Place Tagging
The refined panel location data is spatially joined to a database of audited Geopath Understanding Places/Points of Interest providing meaning to dwell time in specific locations. The places database is rapidly expanding as new members are joining with place-based inventory on a monthly basis.
Visits Calculation
Directly associating panel device locations to nearby places is just a first step in the understanding the total visitation at a specific location.
Once a location is associated to a place, we determine which panel device locations actually stopped to visit the place versus simply passing by. For those who did stop, we determine how long they were there. We analyze the locations before and after a device is associated with a place and we calculate the possible amount of time available for a visit and its associated dwell time. This is accomplished by calculating an area of possibility for each minute before and after a device is associated with a place. This area of possibility is compared to the footprint of the place itself to calculate at what point we can be confident a device has begun a visit and when it has left.
Next, it is necessary to determine how many devices in the Geopath Mobile Device Panel were at the location but NOT observed directly. This is primarily a function of the dwell time in a place (the longer people stay the more likely we are to document their activity directly) and the device activity quantified by how many minutes the device is visible throughout the hour (the more often a device is seen, the more likely it is that we are able to directly document their activity).
One way to think of this is that we expect to see devices in a drive through at a fast food restaurant for 6 minutes. For devices seen every few minutes, there is a high likelihood that we observed most of the devices with similar activity levels. For devices seen only a few times during the hour, we know that it's likely we are missing similar devices.
There are a number of metrics that are used to establish the confidence in number of devices not observed at a location:How active the individual mobile device is as compared to the expected dwell time at the place.
How active is the individual mobile device as compared to the devices generally in the region.
How active are devices at this location compared to other locations. Consider a gas station versus a drugstore. The dwell times are very similar, but devices at a gas station (with time spent sitting in the car or pumping gas) are much more active than devices at a drug or convenience store where people are rushing in to get what they need and then get back on the road.
How active are devices while moving versus ones that are dwelling at a location for an extended period of time.
First Party Data Collection and Auditing
It is recognized that given the pedestrian centric nature of venues where most place-based inventory exists, there has been significant advancement and investment in first party data collection tools, either by the network operator or the venue operator. Geopath as an auditing body of the industry allows for this type of data to be submitted as an input into venue circulation analysis. Current technologies include door counters, transaction, check-in data, video recognition, sensor based and others.
Validation
Geopath is dedicated to providing quality data that is thoroughly validated based on variety of external data sources. Granularity in our core data and processes enables us to validate our products against all kinds of fine and coarse ground truth data across all place types.
First Party Ground Truth Data
First party ground truth data is used through out the development of the methodologies to calibrate and validate venue circulation. In addition, this data is used on a continual basis to ensure reliable measures week over week.
The figures below shows a few example of the reported visits to first party data.
In the top example, monthly reports of validated pedestrian detectors reporting circulation on a weekly basis are shown (orange) against the visits (blue).
In the lower example, transaction counts are shown against the number of visits. While this is aligned, we expect some systematically visits will be reported higher than transactions as there are generally more than one person per transaction as well as visits that do not result in a transaction. This is especially prominent at locations with high numbers of pick-up, drop-off, and delivery visits like hotels.
Airport Passenger Counts and International Visitors
In addition to the processing for Visits to Places described above, Airport places go through an additional level of processing as they exist in a unique closed system of connected airports nationwide. More than 300 of the largest airports across the U.S. are individually profiled and then devices within these places are linked as they move through the network. The lack of device activity while traveling in the air, provides a very clear view of device departures and arrivals at any given airport and simple chaining of airports as people move from one airport to the next. By capturing these movement changes allows us to further segment “Visitors” into counts of passengers versus visitors that were at the airport but not traveling. Additionally, the passengers are broken down into counts for Arrivals, Departures, and Transfers.
When using the airport passenger counts we can validate against the total boardings and alightings that are published by Bureau of Transportation Statistics (BTS) as well as other independent sources. Today, most international visitors are dropped from the panel of devices, so when comparing the calculated passengers we are too high when only comparing domestic flights but are systematically too low when considering total international and domestic. The work to create an additional category of visitors quantifying international visitation is underway now and is being validated using the airports and immigration port of entry data. In the mean time, the BTS counts are used to factor up the observed passengers to the total passenger counts.
Transit Station Ridership Counts and Underground Stations
Similar to the airport passenger counts, first-party ridership data has been gathered from 85 transit systems, including more than 2,500 transit stations and have been used to calibrate and validate the visitation to all transit station Places.
This data is key to reporting the Visits to underground stations that are not able to be measured directly. Any signal that can be measured is compared to any historic passenger counts and factors are simply defined to rectify the obscured observations to the actuals where available. Additionally, to continue to report visitation in “real-time,” the visitation trends from stations that are being directly observed are used to additionally factor the final number of visits reported at the stations which cannot be directly observed. Through time as new data are procured, new releases of the measures will update any visitation to the best understanding available.
Dwell Times
Figure below shows the distribution of the average dwell times for selected place types. Our processes accurately capture various types of visits with different dwell times. For instance, note the longer stays at the hospitals and hotels (with higher variation due to different characteristics of workers, patients, visitors, etc.).