Geopath High-Level Media Measurement Process

Owned by Kevin Narine
May 31, 2023
3 min read

Geopath uses data from various third-party sources in conjunction with its proprietary Visibility Adjustment model to produce measurement for OOH media.

Source data includes:

  1. Information about the OOH inventory itself (how big it is, where it’s placed, the direction it is facing, how many ad units are shown on it, etc) [Geopath proprietary OOH inventory database]

  2. Information about specific roadways and the interconnection of roadways (from Terralytics)

  3. Information about the travels of vehicles along a specific roadway, and for how long (from HERE Maps)

  4. Information about recurring activity patterns from anonymized mobile device panels (from MotionWorks)

  5. Information about population demographics (from government agencies, Motionworks, and Claritas)

 

Simply put, traversal of people and vehicles on a road are adjusted for the likely visibility to see an ad posted on specific OOH inventory, per unit, and the likely audience composition viewing that ad over a day/week/month is computed as an annual average, broken down by hour of the day. The process follows these general milestones:

  1. Geopath processes data for each roadway and pedestrian thoroughfare in the USA to determine the average vehicle circulation and the average pedestrian circulation per hour, each day (including the direction of circulation of vehicular/pedestrian traffic on these thoroughfares). The scope of this circulation includes parking lots and pedestrian malls as well as multi-story buildings (such as airports). Roadways circulation data is sourced from Teralytics and adjusted for seasonality/recency based on MotionWorks' observed mobile device data.

  2. OOH inventory is assigned to one (or many) roadways, pedestrian walkways, or places by Geopath media analysts, so that the circulation of people that travel along these paths can be associated the with inventory

  3. The orientation (as well as share-of-voice) for each ad unit (i.e. ‘spot’) on each piece of OOH inventory are used as inputs to compute the average number of impressions where a single person has the ‘opportunity’ to see each unit of OOH signage (on average, per hour). These ‘opportunity-to-see’ or ‘OTS impressions’ form the basis of an audience’s potential exposure.

  4. Each ‘opportunity to see’ impression count for each spot of each unit of OOH inventory is further adjusted by Geopath’s proprietary ‘visual acuity index’ (VAI) model, which adjusts for the probability of ‘noting’ (i.e. ‘noticing’) the OOH spot. This VAI adjustment transforms OTS impressions, where a single person has the ‘opportunity’ to see each unit of OOH signage (on average, per hour), into audience impressions: impressions of an ad unit where the person is likely to see the ad unit. Spot-level reach is determined by dividing these audience impressions by the average frequency of exposure to the specific audience from a specific geographic area (based on MotionWorks' observed mobile device data data).

  5. Audience composition of each OOH spot is derived from the home origin of every mobile device observed to traverse the path where the OOH inventory is located. Essentially, the location where the mobile device was observed to dwell for the longest period is understood to be the ‘home’ of that person, and the audience characteristics of the census block are then applied to that device.

  6. Measurements for collections of multiple spots are aggregated in a similar way as defined above for audience composition and impression metrics. For reach and frequency metrics, Geopath must account for logical duplication of audiences that traverse subsets of the collection and therefore must make a predictive, probabilistic model of reach and frequency based on the collection presented. To achieve this at scale, Geopath ingests tens of thousands of theoretical ‘packages’ of inventory every year and applies them against a data set of ‘observed’ reach/frequency measurements as a source of ‘truth’. A predictive neural network is built and ‘tuned’ once per year from the outputs of this ‘observed reach and frequency’ data set, within a 5% threshold of error.