In Depth Documentation

If you’re running an OOH campaign to drive web events, app events, or footfall, then we can provide an analytics dashboard that quantifies how many users that were exposed to your ads later took one of these actions.
​
We call these attributed conversions. We use a three-step process to get to this data:
​
​

First, we aggregate mobile location data from various vendors. This data consists of a device ID, lat/lon, and timestamp. This data streams into our database on a daily basis. Note that it takes 5 days for all the data for one day to arrive. For example, we don’t have the full picture for a given Monday until the Friday of that week.
​
​

We define a geographic polygon, or viewshed that encompasses the area where a given device user is likely to have seen your ad. When a mobile device has a ping in the viewshed, we call that an exposure.
​
​
​

This viewshed varies depending on the type of unit, size, facing direction, and environment. In simple cases, for example an urban kiosk, this may be a 10 meter radius. In more complex cases, for example a highway billboard, this may be a hand-drawn polygon.
​
​

The second primary factor in defining an exposure is dwell time. Dwell time is the duration that a device was in the viewshed. We are able to calculate this because devices ping frequently. We define dwell time as the time difference between the earliest seen ping to the last seen ping while the user was in the viewshed. Varying unit types, sizes, and environments also require varying dwell times.
​
​
​

Lastly, for digital units and programmatically-purchased units we apply an additional filter: a device can only be considered exposed if it was in the viewshed around the time the ad actually played. A list of times your ad played is called play-logs. For digital units, we request play-logs from the media owners. For programmatic units, the SSPs already send us the play-logs via APIs.
​
​

We then use these play-logs to create a time filter – the device must have been in the viewshed within plus or minus a few seconds to a few minutes of the play to be considered an exposure. We call the time window. We use varying time windows based on the number of spots purchased, the loop length, the unit type, and the purchase type (traditional digital vs. programmatic).
​
​
​

The next step is to get all converted device IDs. By all conversions, what this means is all the conversions that exist before we match the data to exposures.
​
Here’s how it works for each event type:
​
​
​

Web Events

Let us know the names of the events you’d like to track, and we’ll provide you with a pixel for each event. You’ll need to place the pixels on your site before the campaign starts.
​
​

The pixel takes web visit data and sends it to our Cross-device Graph provider. We call this the “Device Graph” for short. The Device Graph is a database that links web visit data to a set of devices. For example, if a given user visits your website on their laptop, the Device Graph attempts to return to us the device ID for the user’s phone. (Please note that this data is all anonymized.)

We get all of these device IDs, along with their conversion event, on a weekly basis every Thursday, so dashboards for web conversions are updated every Friday morning.
​
​
​

App Events

We recommend using AppsFlyer or Branch to track app events. Simply instrument the events you care about and then provide us with an export on a weekly or bi-weekly basis of the device IDs for each event.
​
​
​

Footfall Events

For Footfall, please provide us with a list of the locations you’d like to track. We’ll draw a radius or geo-boundary around the locations, then we look in our mobile location data to find device IDs that entered those bounds. We will also define a custom dwell time based on the type of location.
​
For example, dwell time for an art gallery conversion may be 5 minutes, but dwell time for a restaurant may be 30 minutes.
​
​
​

Other Events Lastly, if you have device ID data from another event not listed here, for example Point-of-Sale data, we can ingest that data and provide the same type of reporting.
​
​
​

At this point, we have the devices that were exposed and devices that converted. To get to attributed conversions we look for device IDs that exist in both datasets.
​
​

You can see these attributed conversions in the dashboard as Directly Attributed Conversions.

You’ll also see a figure for Total Estimated Conversions. Here’s why we have this: our mobile location dataset only represents 5 to 10% of the total population. We call this the sample rate.
​
Therefore, in order to get a more accurate estimate of conversions, we upsample direct conversions with a conservative estimate that our device dataset represents 20% of the population. In other words, we apply a 5x upsample rate (100% / 20% = 5) to get from Direct to Total Estimated Conversions.
​
​
​

Limits & Restrictions
​
Standard reporting is currently only available for the U.S. & Canada regions.
​
​

Reporting for Europe and Latin America is available but requires custom pricing and additional lead time. Please let us know if that’s something you’re interested in.
​
​

Data Flow Diagram
​
​

​