Motionworks has improved its Population Intelligence platform, measuring activity patterns in and around defined places. These changes impact place-based media measurement in Out-of-Home (OOH) advertising in the United States in the following ways. The latest estimation methods cut average error by more than half from the earlier estimation methods.

	This is What We Used to Do in 2019 (v1.0)	This is What We Do Now in 2022 (v2.0)	Why It Matters
Dwell time computation	Motionworks counted dwell times using only location observations observed within a place’s polygon. This resulted in dwell times that were always less than the real-world dwell. The first location point inside the polygon is always after the arrival, and the last location point inside the polygon is always before the departure.	New Computation – Motionworks expanded the estimated dwell time to include estimating the actual arrival and departure times using location observations before and after directly interacting with a place’s polygon. Additional sources informing arrival and departure times are also used for common locations to create more accurate dwell and travel times.	Observed location data points are not always recorded at the exact times a person truly enters and exits a polygon. By building this time discrepancy into our computation of dwell times now, dwell times increased to reflect actual behavior more accurately.
Dwell time thresholds by place type	Previously, activity in places (defined with data collection polygons) was logged only for observed dwell times greater than 15 minutes – which may have been too long to capture visitation at some place types. For example, visits to gas stations were improperly classified as drop-offs or passbys.	New Thresholds Motionworks now sets a dwell time minimum by place type by identifying a natural breakpoint in the dwell time frequency distribution. This identifies activity at locations with shorter dwell times, such as gas stations, drug stores, and quick-serve restaurants. Also, overnights and other artificial time boundaries have been removed so that hotels and hospitals have longer dwell times compared to previous.	Visits to places that are of specific place types (i.e., gas stations, hotels) whose dwell time threshold is at the extremes (high or low) will see more significant changes to measured activities. Based on the inclusion of shorter duration activities as “visits,” trips between home and work may now be home-to-other and other-to-work. This increases the number of trips and visits overall.
Datasets	Data sources contributing to the Motionworks solution were primarily raw location data feeds (xyt), offered by aggregators.	New Financial and Documented Place Visits Datasets – Motionworks has added two new feeds that provide verified visits to locations. New FTA Datasets - U.S. Federal Transit Administration’s (FTA) National Transit Database (NTD) has been added to estimate station-level counts monthly.	Motionworks continues to innovate its approach to population intelligence. The additional datatypes that are now supported inputs provide valuable insight into trip-making patterns, panel qualifications, and other important intermediate data stages. These impacts manifest in higher confidence in dwell times, trip rates, visits, and more.
Small and “noisy” locations	Ethnic markets and local bars and restaurants with very small activity were overestimated due to inconsistent observations.	New Method – Motionworks understands the average expected activity for any given hour at a place by looking back over several weeks and then uses the observed visits to understand if visits are running higher or lower than the expectation.	Activity patterns at small and “noisy” places are now more stable through time. The latest estimation methods cut average error by 38 percent from the earlier estimation methods.

Quantifying the Changes by Place Type

For each place type that Motionworks measures, the changes between version 1.0 (what we used to do in 2019) and version 2.0 (what we do now) are presented for each publication date below. Each place type is summarized with a paragraph and two tables: the methodology breakdown and change statistics.

Publication Date (vintage) 2023.03.10

Methodology Breakdown

The first table in each shows the breakdown of the methodology applied across all places within the place type. The number of places measured count places in the current year with version 2.0. The methodologies include:

directly measured places
small and “noisy” places
unfocused places

Directly measured places are those places that are large and consistent enough to measure visits and dwell times on an hour-by-hour basis directly.

Small and “noisy” places, a new methodology in version 2.0, are also directly measured. But due to the nature of activities at the location or due to the small area of their polygon, visits and dwell times are estimated by looking back over several weeks. The previous several weeks inform the expected activity, and that expectation is compared to the observed visits in the current week to understand if visits are running higher or lower than the expectation.

Unfocused places are locations like underground transit stations or business or retail locations that might be on the first floor of a multi-story building or within a larger shopping center. These places are not able to be measured directly, so instead, a model for the place is used that draws from the visitation to similar locations in the area along with the audience profiles and hourly activity patterns of the “parent location,” which are directly measured.

Change Statistics

The second table in each shows the median across all the places within the place type for:

visits per week
dwell time in minutes

Additional columns show the absolute and percent changes from version 1.0 in 2019 and version 2.0 in 2022.