People Counting via Body Detection: Most Accurate Method for Measuring Audience Reach?
When gathering any statistical data, accuracy is the key component to success!
That is why video analytics is quickly becoming the trusted method for data aggregation and audience measurement in DOOH, Digital Signage, and OOH Media. One important insight that video analytics can provide is accurate impression data via People Counting Technology. This is a fundamental metric that you will need to measure when deploying any media network. Audience ad reach is largely determined by the physical environment, through the integration of AI based people counting technology, you can easily determine if your screens are placed in the best location receiving the highest traffic and generating consistent levels of impressions. Additionally, having accurate and real-time data about ad performance on your network and providing a type of ‘Google analytics’ service for the offline world can be the right incentive to motivate buyers to invest in your digital ad space.
Traditional People Counting Methods
Throughout the years, the DOOH industry has used a number of different methods to measure the audience reach of out-of-home advertising. While each of these methods generates quality data in its own respect, they all carry some limitations when it comes to attributing views and impressions to specific displays:
Transactional Data from Point of Sales – Sales data in retail venues can give a good indication of how many people have entered the store and made a purchase but it leaves out all individuals that have entered and exited without making a purchase. These are all potential viewers. There is very little attribution possible because PoS data does not guarantee that a customer who bought something at the store has also seen the ad display in the store.
Mobile and Wi-Fi tracking – Mobile device tracking provides accurate ‘big picture’ data but these counts tend to also include people who are not within the viewing distance of a screen. Similarly to transactional data, it’s hard to attribute viewers and impressions to a specific display because of the high coverage and reach of this technology.
Surveys – Surveys can deliver a certain amount of qualitative data however they are often very time consuming and require a lot of manual labour. They also mostly rely on individual testimonies and human recollection which is prone to error.
These techniques are generally very well used for understanding the overall picture, but the data cannot be segmented per campaign or per screen. People Counting powered by AI body detection takes the edge on accuracy while providing viewer data in a more granular format. Let’s take a look at how it works!
What Is People Counting and how does it work?
People counting is the process by which people are counted within a given area, location or space, usually for the purpose of statistical analysis. The process can be carried out using a variety of methods, one of which is body detection. Our product DeepSight SDK (and soon also DeepSight Toolkit) now uses body detection to count all people within the viewable area of a screen. This is how our body-based people counter works:
A camera, usually embedded within a digital screen itself or a separate USB/IP camera pointed towards the flow of the audience, is used for counting the number of people passing by the screen using body detection.
The same camera is used for counting the number of people viewing the screen using face detection
This data is aggregated and typically exported to a Cloud-based dashboard for further analysis and reporting purposes.
The resulting metrics are nr. of OTS (opportunity to see), nr. of impressions, nr. of viewers and their respective dwell and attention times.
Body Based Counting vs Face Based Counting
Both body and face detection are useful methods of People Counting in their own respects. They can detect people and do so accurately and with ease. Body detection has a greater object mass (bodies) to work with therefore it is suitable for counting people further away from the screen. Face detection is very well used for tracking the viewing behaviour of an audience and counting the number of true impressions and true viewers closer to the screen.
With Body Detection, you can expect:
Longer Detection Ranges (up to 10 metres)
Ability to work effectively even in crowded environments with frequent occlusions
Robustness when it comes to people facing away from the camera
Up to 30% more people counted!
With Face Detection, you can expect:
Shorter Detection Ranges (up to 6 metres) hence higher attribution of impressions to a specific screen
Granular measurement of impression and viewer numbers
Accurate dwell and attention time data
Using these two in combination unlocks the real potential of any DOOH network to achieve its maximum audience reach. That is why we continue to utilize both face and body detection, combining them into the same reliable product, allowing for more solid data collection.
More Accurate Audience Reach Measurements for all DOOH Screens
So why should you use body detection for your digital signage network?
Computer vision based audience measurement does not rely on tracking devices carried by individuals, instead, it analyzes individuals in an anonymous way using cameras attached to screens that detect if passersby have viewed the ad content and for how long.
Data is collected in real-time, reflecting a true picture of the audience at the time of analysis so it can be effectively used to offer premium ad space programmatically and alerting advertisers when there is highest traffic.
Digital screens can be placed in very crowded, high traffic areas therefore it is important to be able to cut through the noise and understand the true viewer and impression counts. Using a combination of face detection and body detection media owners and advertisers can get both OTS counts and impression counts collectively. People counting can directly attribute audiences to a specific screen and creative which cannot be achieved using general traffic trends.
What does all this information and real-time statistical data mean for Network Owners? At the very beginning, when onboarding a new media network, the network owner needs to answer the basic questions of;
How many people can I reach?
What type of audience profiles can I reach?
How much am I going to charge advertisers for this space?
Having solid audience data can make this process much easier for any media owner. With accurate impression data based on people counting, media owners can prove how much traffic their screens generate on a daily, weekly, monthly basis. They can easily determine their cost per thousand impressions (CPM) and offer transparent reporting on this to the buyers. Through the integration with an existing SSP, media owners can use impression data to enable their customer to buy ad space programmatically – this time based on impressions instead of location or frequency of ad plays in a loop (see Figure 1.).
Figure 1. The process in which impressions are used as currency for programmatic bidding.
By counting people, media owners can guarantee advertisers that the space they paid for was actually seen by their target audience. This way the advertisers get a very clear understanding of how many people were reached with each ad and how each ad performed in terms of viewing time. Apart from impression data, network owners can also use our audience measurement software to collect demographic data about the viewing audience which can be used by marketers to target their creatives to the right people. Such information is highly valuable for any brand therefore if the network owner can prove that they are able to deliver a highly targeted audience with their screens, brands will also be willing to pay more for such space. Not only because of the targeting capabilities but also for the option to collect post-campaign performance data to understand which creative was most successful, received the most attention, converted most impressions into views and so on. All of this is possible only on an audience measurement enabled network.
As for the verdict? There are many measurement technologies out there but if you are looking for one that provides accurate counts for individual screens as well as the whole network, then body detection is the method to utilize.