Geopath Mobile Device Trip Frequency Analysis
- Kevin Narine
Overview
Rather than using self reported travel surveys, Geopath is now able to observe trip making behavior utilizing the Geopath Mobile Device Panel to calculate the frequency of trips by the same device over the same roadway in the same direction throughout time. These methods create an empirically derived frequency pattern to apply to individual pieces of inventory for calculation of audience reach and frequency.
The observed frequency post-COVID is lower than observed pre-COVID periods. Additionally, observed frequency of trips passing out-of-home inventory (per and post-COVID) is much lower than frequency benchmarks from legacy self-reported travel surveys.
Additional benefits:
Initial validation of observed trip volumes of the mobile device panel correlates well with continuous counting stations from California performance monitoring system.
The routing engine is capable of creating a dataset of exposed mobile devices.
Observed device frequency analysis adds the ability to project reach and frequency out to 365 days.
Method
Route all trips of qualified panel devices over a 50 day period through the street network.
Utilizing every individual road segments within viewsheds of audited inventory, summarize the total number of passes by panel devices within X days and the number of unique devices that passed through that road segment.
Filter for road segments which have a significant sample of devices over the total number of days (n=350+)
NOTE: the sample size threshold for individual segments is very conservative and once we’ve identified stable patterns in the data, we’ll reduce that threshold to explore additional precision.
Validation
Roughly 83k devices observed over the 50 day period qualified for the frequency analysis in Los Angeles. This is 57% of the average weekly panel of ~145k devices in LA. The panel makes up 15% of all observed devices in a week.
California PEMS stations were programmatically matched to the HERE road network based upon location, description, and directionality of the station.
Observed trips of device passes per day were compared to daily traffic counts from California PEMS for mainline stations.
Artificial patterns in the traffic volumes for PEMS stations were observed in some station days. These volume patterns were associated with imputed traffic data from PEMS. The comparison set was then limited to only days with 100% observed counts.
R-squared value for class 1 roadways is 0.60 and class 2 roadways is 0.39
Summary:
The alignments with the California PEMS data is positive and expected to improve with a review of PEMS and HERE map segments assignments. The correlation of the device passes and daily vehicular volume suggests that the routed trips is a good indicator of daily trip activity across roadways. The analysis of these device trips is the foundation for quantifying the expected frequency distribution of circulation passing by out of home inventory.
Next Steps:
Additional work is underway to review the PEMS and HERE maps segment assignment to ensure accuracy of trips to PEMS stations counts. This is expected to greatly improve validation results.
Average Frequency by Road Class (LA)
Road Classes
a road with high volume, maximum speed traffic
a road with high volume, high speed traffic
a road with high volume traffic
a road with high volume traffic at moderate speeds between neighborhoods
a road whose volume and traffic flow are below the level of any other functional class
Class 1- 4 road segments to which inventory is assigned has a very similar frequency pattern over time as all roadways in Los Angeles.
Class 5 road segments to which inventory is assigned have a lower frequency distribution that all class 5 roadways in Los Angeles.
For road segments with >350 unique devices observed over 50 days
Summary:
As expected, highways have a slightly lower frequency over time than arterial roadways. Unexpectedly, class 5 roadways have the lowest frequency, but also differs the greatest from all class 5 roadways in LA. The expectation was that local roads are where people live an therefore must use those roadways to go anywhere. However, this appears to not be the case when looking at all roadways across the area.
The hypothesis is that out-of-home inventory, while assigned to class 5 roadways, isn’t located in close proximity to people’s place of residence, but closer to places of business and commercial purposes. When a person travels from their place of residence to a work or shopping destination, the beginning (“first mile”) and end (“last mile”) of those trips will be on class 5 roadways.
This hypothesis is supported by the alignment of class 5 frequency distributions to those of place-base media place types. (See below).
Average Frequency by Road Class (Nationwide)
When comparing the deep dive into Los Angeles against a sample panel across the entire US, similar frequency patterns emerge, especially on highways. Similar to LA, arterial roadways have a higher frequency than highways.
Average Frequency by Area Type (LA)
Area Types
Core Business District
Urban Business
Urban
Suburban
Rural
When focusing in on class 4 roadways, it is clear that the trip frequency is higher in lower density areas (likely closer to places of residence).
Low average trip frequency to business districts aligns also with place types frequencies of individual POIs. (See section on Place Type Frequency).
Average Frequency by Area Type (National)
Class 1 : road with high volume, maximum speed traffic
Class 2 : a road with high volume, high speed traffic
Class 3 : a road with high volume traffic
Class 4 : a road with high volume traffic at moderate speeds between neighborhoods
Class 5 : a road whose volume and traffic flow are below the level of any other functional class
Summary:
In most cases, core and urban business districts have a lower average frequency than more suburban and rural areas.
Next Steps:
Analysis of inventory broken out by population persons residing within the h3 and surrounding cells “residents” to confirm that trip frequency for residential heavy roadways is as expected and aligns with home-based trips from Geopath Government Transportation Statistics on Trips .
Average Frequency by Place Type
Year over Year Frequency Comparison
Geopath has released many products that highlight the change of trip volumes from pre-covid and post-covid periods. The same comparison can be done when looking at trip frequency.
Summary:
The average number of trips has decreased post-COVID as has the frequency of trip over the same roadways where inventory is located.
Projecting Average Period Frequency
It is possible to calculate a trendline curve from the observed frequency over time. The exponential curves built from 50 days of observed data have a R² ~ 0.99
Summary:
Utilizing the observed frequency from the mobile panel enables Geopath to create accurate projections of frequency over periods extending to an entire year.
Comparison to Legacy Survey Data
Legacy methods of surveying travel of individuals reported a weekly trip frequency of ~4.2 and a monthly frequency of ~12.
Summary:
Moving to the updated observed frequency will create an decrease of frequency per unit and and increase of reach per unit at the same TRP level.