Overview

Placecast™ 2.2 includes several enhancements and new features. This updated release utilizes an improved methodological approach to refine visit reporting at locations that may have previously included overestimates.

Historically, Motionworks published Validation Summary Documentation by place type. These analyses and quality checks focused on the performance of individual locations. The scoring system takes into account weekly visitation, hourly activity profiles, dwell times, and the trade area footprint of each place. The scores were consistently verified and validated with each monthly release and updated whenever new locations were added.

In 2024, our validation and verification metrics were expanded to include the analysis of estimated visitation metrics by place type groups to ensure reasonableness. This improvement to our validation and verification system identified unrealistic growth across some place types through 2023 and beyond. This unrealistic growth trend was a product of ongoing changes in the mobile device data marketplace that required additional modelling steps to be implemented, some of which, over time, began to include overestimation for visitation. Additional fixes for this overestimation were implemented in Placecast™ 2.1 and include applying market and place type averages for specific places that were identified as outliers. While this method continues to be used, Motionworks acknowledges that the approach impacted certain specific places where the visitation that was identified as an outlier would have been more in line with expectations for those specific places. This can be seen in the place type reports linked above, and for the example of theaters below.

With the release of Placecast™ 2.2, this unrealistic growth pattern has been corrected. For example, for all places published, the average estimated place visits in February 2023 and 2024 in version 2.2 versus 2.1 were reduced by 40.4% and 48.8% respectively, thereby bringing the estimated visits back in line with expectations.

Methodology Updates

One of Motionworks’ differentiators is that our data provides a complete view of the full population's movements and activities, not just a count of devices. We use data from mobile devices, consumer spending, and connected cars to reflect the total population's activity – everywhere, all the time. Motionworks strives to continually refine and improve our methodologies, incorporating the best data inputs available. For Placecast™ 2.2, Motionworks has updated its methodology around panel selection, place analysis and hourly activity analysis, confidence measures, and event analysis.

Panel Definitions

Motionworks catalogs activity sourced from more than 200 million devices across North America each month. As described in the ”Identification of the Panel” documentation, those devices' activity patterns are used to deduplicate, classify, and quantify attributes assigned to each device. This curated set of all understood and cataloged devices is referred to as Motionworks’ TotalSelect™ Panel. The TotalSelect™ Panel drives a selection process that defines two complementary subpanels of devices that will be used to understand specific aspects of population activity, while minimizing bias that is present by geographic, socio-demographic, and occupational classifications. Placecast™ 2.2 utilizes these two complementary panels to provide optimized insights for all analytical needs.

BroadSelect™ Panel: The BroadSelect™ Panel offers a comprehensive set of devices that capture a wide range of mobility patterns. Its strength lies in the sheer volume of sightings, which provides extensive coverage and minimizes sampling bias through its large sample size. This panel is ideal for generating a detailed understanding of cohorts, trade areas, and hourly insights at every individual place for its complete history. The BroadSelect™ Panel excels in providing the most inclusive insights across diverse populations and geographies.
PrimeSelect™ Panel: The PrimeSelect™ Panel, which is a subset of the devices in the BroadSelect™ Panel, features a selection of devices characterized by the high frequency of sightings and persistence through time. This panel's strength is its ability to provide precise and reliable insights into device movement patterns. It excels in accurately capturing speed, dwell times, and travel routes through transportation networks. The PrimeSelect™ Panel is particularly valuable for applications like analyzing special events, detailed travel patterns, or offering a granular understanding of movement trends and behaviors.

Motionworks' TotalSelect™ Panel with its two complementary subpanels, BroadSelect™ and PrimeSelect™, are integral to Placecast™ 2.2, offering unique strengths to provide comprehensive insights into population mobility and behavior.

Peer Groups

In Placecast™ 2.1, Motionworks introduced new quality control methods based on the way a specific place was behaving in relation to other, similar places. These Peer Groups were created by grouping places based on type, urbanicity, and region, with a defined number of places per group. In version 2.2, Peer Groups have been updated to incorporate more granular quality criteria for each place's inclusion. Additionally, machine learning techniques are now used to ensure each place joins the group that best represents its activity.

Hourly Activity

With Placecast™, Motionworks has always been able to identify visitation to a place over the course of a day. In Placecast™ 2.2 a new series of adaptive analytics has been introduced, shifting from the hourly profiles established by the Placecast™ 2.1 Peer Groups to a Bayesian statistical methodology. This allows each individual place to maintain its own expected hourly activity throughout the week instead of just being based on the expected activity of its place type. The Bayesian statistical algorithms dynamically evolve the hourly profile for each place based on the latest observations utilizing the BroadSelect™ Panel, ensuring unprecedented accuracy and adaptability. As the panel fluctuates week over week, the location-specific patterns are preserved while smoothly incorporating emerging trends. This provides the most up-to-date and reliable visitation profiles, driving improved user decision-making.

Confidence Intervals

Placecast™ 2.2 has added confidence intervals to many key metrics by treating each potential visit as a binary choice made by the individuals in our digital population: "visit this place" or "do something else." This methodology offers superior accuracy, especially for locations with lower visitation. The approach provides more reliable and intuitive insights, especially across different types of locations and over time, enabling more informed decision-making.

Event Analytics

The new and improved Placecast™ 2.2 methodology allows Motionworks to provide analytics for select events in custom time frames. This new feature, called Placecast™ Select, can be utilized to assess visitation to venues or regions during events of interest, such as sporting events or concerts in stadiums, visitation to a group of locations during a marketing campaign, or an experiential event at a park for example.

This new functionality has helped refine the generalized Placecast™ 2.2 profiles for event heavy place types and is already a key input into several Motionworks solutions being utilized by clients in television, radio, and CPG.

Improved Estimates

Visitation

Placecast™ estimates visitation by year, month, daytype, and hour for millions of places over time. This is done through a sophisticated data pipeline that processes several population mobility data sets with advanced data science methods to produce estimates nationwide and throughout time.

As described in the previous section, the release of Placecast™ 2.2 introduced significant improvements to our data science methods to improve visitation estimates. The overview of these changes is above.

The two key metrics of Placecast™ are visits and occupancy. As a brief refresher, visits, are persons who arrived in the year, month, daytype, hour being measured, and occupancy, is the average number of persons present during the hour. Below is an overview of some of the improvements of stability and reliability in Placecast™ 2.2 as a result of the new methods.

Visits

5 percent drop in overall volatility for monthly visits across all places from January 2019 to January 2024. This volatility is essentially calculated by taking the standard deviation of the differences between consecutive data points through time.
26 percent drop in overall volatility for hourly visits across all places in 2023.
46 percent drop in overall volatility for hourly visits across all places from January 2019 to January 2024.
45 percent drop in outlier estimates for hourly visits. These are hourly visits significantly outside an acceptable range of their average visits.

Occupancy

25 percent drop in overall volatility for monthly occupancy across all places from January 2019 to January 2024. This volatility is essentially calculated by taking the standard deviation of the differences between consecutive data points through time.
35 percent drop in overall volatility for hourly occupancy across all places in 2023.
34 percent drop in overall volatility for hourly occupancy across all places from January 2019 to January 2024.
50 percent drop in outlier estimates for hourly occupancy. These are hourly occupancies significantly outside an acceptable range of their average occupancies.

Dwell Time

Placecast™ 2.2 introduces a significant improvement to our data science methods to improve the estimation of dwell time in a place, which is defined as the time a person spends visiting (or at) a place. Version 2.2 now reports dwell time at the day-of-week and hour level. In Placecast™ 2.1, dwell time was reported at the weekly level. This additional granularity in estimation allows for more targeted analysis.

Trade Areas

As part of the development of Placecast™ 2.2, Motionworks identified improvements to the process and methodology used for Placecast™ Trade Areas. Much like issues identified in other Placecast™ products, Placecast™ Trade Areas has also been affected by ongoing changes in the mobile device data marketplace. As Placecast™ Trade Areas rely on the direct observation of the mobile device panel, these changes led to an overall decline in the size of the footprint of local trade areas for specific places over time.

Placecast™ Trade Areas 2.2 introduces additional modelling steps that use a combination of observed data, transport modeling methods, and data science techniques to ensure stability and consistency of the size and scope of the trade areas reported for a specific place. These modelling steps rely on observed mobile data to help determine the split between local and non-local visitation to a specific place. Then, based on the previously observed data, Motionworks utilizes transport gravity model techniques and demographic controls to estimate the size and scope of the local visitation's trade area footprint. Additional controls based on a specific place's peer group are utilized to ensure reasonableness of the specific place's trade area distribution.

As a result of these changes, average trade area distances from a given place are more reasonable and the number and type of trade area geographies are stable and consistent through time. Additional reporting on trade area validation checks that helps ensure this stability and consistency are included in the Placecast™ Summary Reports by place type.