Estimated Visits: 2022 brings the future of visitation insights

2022-05-13 OOH Measurement, Visitation Insights, Foot Traffic Data, Data Driven Marketing, Traffic Counts by Address, Retail Foot Traffic

Estimated Visits 2.0—an exciting development in visitation volume reporting— is now being offered to Spotzi clients worldwide in our Spotzi Professional and Spotzi Enterprise dashboards. This visitation report uses a proprietary algorithm to estimate precise traffic counts by address, combining an analysis of each location's unique characteristics with human movement data. 

The Estimated Visits report already offers coverage across the US, Canada, Australia, and New Zealand. This allows users to compare location performance across markets, verticals, and even borders. 11 additional countries are also in the process of being added to this growing roster.

Estimated Visits is an invaluable tool to marketers and business owners alike; an accurate insight into true visitation volumes is crucial to the evaluation of store/business performance, the assessment of real estate value, and the optimization of retail business labour and stocking schedules.

Foot Traffic Data: Previous Methods

Human movement data—sourced from intermittently observable smartphone devices with user consent—is currently being used as the equivalent of foot traffic data. This data provides a valuable starting point for the estimation of true visitor traffic, but fails to be a suitable representation of retail foot traffic alone. This is because:

  • Device data is only intermittently observable; Even when a visitor does carry a smartphone and has opted into sharing their location data, their device’s observability at a particular location can vary greatly depending on its settings and usage.
  • Every location provides its own unique visitation context which also affects device observability.
  • A longer average dwell time (such as at sports games, or in the waiting room of a doctor’s office) provides more opportunity for visitors to use their smartphone devices and contributes to a greater likelihood of device observation at that location. 
  • A shorter average dwell time (such as at a business with little-to-no wait time, or a coffee shop fuelling busy morning commuters) provides much less opportunity for device use during a visit. This can lead to less human movement data being observed at these locations even when true foot traffic is substantial. 
  • The availability of cellular service or WIFI at a location will also impact device observability. Similarly, the probability of a device being observed can change drastically throughout the day and night.

Estimated Visits—The New Standard of Visitation Reporting

The Estimated Visits 2.0 report uses a machine-learning model to analyze real-world data and provide insights into true visitor traffic. This model improves upon existing methods of foot traffic data collection in both accuracy and privacy—with data projection being a far more privacy-friendly method of visitation reporting.

This ML model has been trained to recognize and extract complex patterns from movement and traffic data; It combines these travel behaviour patterns with an observability metric to make foot-traffic estimations. Access to ground-truth data is continuously being used to hone the precision and accuracy of these predictions.

Machine-Learning Visitation Reporting Process

The Estimated Visits model uses a probability measurement, and takes the following points into account when extrapolating true visitation volumes from device movement data:

  • How probable is it for a specific device to be observed at any location based on its users habits and usage history?
  • How probable is it for any device to be observed at a specific location based on that location’s unique characteristics?
  • A device’s location data can be processed in several different ways before it's sent out to our database for use.
  • The probability of observation can change drastically throughout the day, and an observability metric must therefore analyze device observations over a sufficient time span.

Interested in our Visitation Reporting?
Do you want to know more about this report and other offerings in our dashboards? Or do you want to plan a demo? Please contact sales for more information.

Estimated Visits: Algorithm Breakdown

Observability Metric

A proprietary device observability metric is used to quantify the overall probability that a device will be observed at any given location.

  • A device set to ping an observation point every 30 seconds will be rated as having high observability.
  • A device only observable when a user opens a specific and rarely-used app will be rated as having low observability.

Average Dwell Times

The probability of observing any device at its location increases with dwell time. Therefore, the average dwell time must be calculated at every location in order to accurately assess the relationship between its movement data and actual foot traffic. Each location’s average dwell time is paired with every visitor’s device observability ranking (see above) to achieve an unmatched assessment of overall observability.

Balanced Panels

A demographically and geographically-balanced set of panels are created each month based on active device observation and the latest available census data. Movement data is then filtered down to these panelist devices, and they provide the basis of a location’s overall estimated visitation. This ensures a consistently-sized sample of devices and corrects for obstacles in  internet connectivity, as well as geographic and demographic bias.

Ground-Truth Data

In the past, research into foot traffic predictions and estimations was obstructed by the limited availability of real-world data. This lack of ground-truth data access was due to businesses heavily regulating access to store POS data, and exclusively using it for internal foot-traffic calculations.

Today, partnerships with a variety of businesses and data sources have allowed us to obtain this missing data. We can hone our machine-learning models by calibrating each algorithm against purchase data, POS data, and data obtained from beacons/cameras. Our machine-learning models are able to provide the most accurate visitation estimates possible based on ongoing pattern-extractions and ground-truth comparisons.

The Future of Estimated Visits and OOH Audience Measurement

Additional ground-truth data is continuously being sought out and utilized across all current and future available countries. The tireless training of the algorithms behind this product will lead to a continuous rise in estimate accuracy, and will allow clients to gain the most precise insights into visitation volumes across the world.

Upcoming Features Include:

  • A large-geography visitation estimator.
  • Expansions into additional countries.
  • The normalization of Zero-Panel Days (based on historic traffic).
  • A tourism estimator.
  • Model Override features—allowing users to adjust settings for custom or experimental insights.

Interested in using the Estimated Visits report in your outdoor advertising strategy? Learn more about this report and other offerings in our Spotzi Professional and Spotzi Enterprise dashboards.

Contact sales today to book a free demo

Related blog posts

Try Spotzi for free and start finding your best-fit customers