Being Transparent about Real-Time Impression Verification
—Ad Tech that Verifies Every Impression Delivered
By Marlow Nickell, Co-founder At Grocery TV
When we started building Grocery TV, we wanted one of our core values to be transparency. Internally, we’re transparent with employees through our open salary and financial policies along with the general rule that information is available by default and an explicit reason must be given to restrict access. When it comes to transparency with our customers, one of the key decisions we made early on was to build advertising technology that verifies every impression we deliver.
Most Digital Out-of-Home (DOOH) companies begin by hiring salespeople and assembling their ad network with off-the-shelf technology. Perhaps out of ignorance, we took the opposite approach and focused on building our own technology stack first— from designing and manufacturing the digital displays up to the ad server that ‘talks’ to the programmatic networks. It wasn’t until our third year in business that we hired our first full-time salesperson, and even today our sales and accounts team is relatively small for a network of our size.
Ultimately, I’m glad we went in this direction, because verified impressions weren’t possible with the technology that was available in the DOOH industry when we started, and it’s still an uncommon practice today.
Since we get a lot of questions from clients about how our impression verification works, I’m going to share a little background on the system. There’s nothing we’re doing that’s particularly complicated as the truly impressive work is done with open-source software that anyone can use. My hope is that this article is helpful not only for our customers but also for other DOOH networks as I think real-time impression verification will help validate our industry and accelerate its growth.
Let’s reflect on DOOH standards
Until about 5 years ago, DOOH was merely a subset of the traditional out-of-home industry and operated in largely the same way.
Work was handled by email and spreadsheets and discussions were held in the language of billboards. The most notable changes were that you could put an ad up faster and you could do flips instead of only one campaign.
What we recognized as a company was that digitizing the entire system, and having the necessary backend infrastructure, allowed you to do something completely different. Rather than grafting digital screens onto the traditional OOH marketplace, you could instead operate as an extension of other digital mediums like display, video, social, and Connected TV.
However, the standards are completely different in the digital advertising industry. Advertisers are looking to purchase direct impressions, rather than a rough estimate, and campaigns use impressions as a fundamental unit of measure. Given that, we believed real-time impression verification would be critical to building strong partnerships with programmatic buyers.
The problem with any kind of OOH signage is that it’s not set up from a technology standpoint to provide that kind of granularity. Most networks use a static value based on 3rd party surveys to estimate traffic for a given campaign. Some networks are starting to use computer-vision systems to improve their traffic estimates, but the off-the-shelf options have high ongoing costs so they’re usually deployed to only a small subset of a network.
We believed there was a huge opportunity to bring one-to-one impressions to the DOOH industry, just like any other type of display ad online. When someone walks up to one of our displays, we run content based after recognizing that a person is present and record the event as a verified impression (if two people are present count two impressions, etc.) and calculate that all day in real-time for every display.
Ad fraud and viewability in digital advertising
When we switched our focus from being a “digital version of OOH” to “OOH that’s available for digital buyers” we realized that OOH is an incredibly premium format. For a typical digital campaign, the percentage of true impressions delivered drops significantly when you factor in widespread issues like fraud, ad blocking, and low viewability.
Conversely, DOOH operators have full control over their displays. Their networks are found in brand-safe locations with 100% viewability, and all of the content is curated by the media owner. Brands and advertisers can be assured that their ad isn’t going to be placed next to an offensive video or a controversial political message.
And advertisers seem to recognize these benefits, given that they already spend 2-3x more per impression for digital out-of-home than digital display and video ads.
Building the ad tech to verify impressions
Until recently, most of our peers told us that verifying impressions in real-time didn’t make sense for our industry. In fact, the common consensus was that programmatic DOOH, which uses impressions as the core unit, was most useful for monetizing remnant inventory.
I’d like to tell you we were visionaries who saw where the industry was headed, but, in fact, we were just stubborn engineers who enjoyed solving technical challenges more than selling our solutions. In many ways, this description still holds true.
So, for rather arbitrary reasons, we decided early on that we wanted to create an experience where you could walk up to a static screen that’s displaying a store’s branding, it detects your presence, and then transitions to running entertainment and advertising content from where it last left off. We also wanted to count how many people were present while this was happening so we knew how each play translated into impressions.
So our two main problems were detecting motion near the displays and counting the number of people present. Fortunately, there’s a number of open-source computer vision libraries that excel at solving these types of problems.
To handle the motion detection, we use a common technique called background subtraction. When motion above a certain threshold is detected, we resume running content and search for people with a standard face detection algorithm. Then, if any faces are detected, we record the number found and keep track of them as they move around near the display. To ensure we only count people near our display (rather than in the next lane over), we set a minimum size threshold for face detection. The final count is then stored with each ad play, allowing us to directly track impressions for each ad delivered. Once a display stops detecting faces, it stops running content and returns to a static image of the store’s logo.
Protecting personal data
It’s incredibly important to both us and our retail partners that we protect the privacy of in-store shoppers, so when we built our impression verification system, we had to do it in a way that avoided any personally identifiable information. In particular, we didn’t want our system storing any images of shoppers.
To meet this requirement, we run our impression verification locally on our displays, which lets us analyze images of a display’s surroundings, gather the impression data we need, and then immediately drop the associated image data. The challenge with this approach is that it means our system has to work within the constraints of our display hardware, which isn’t an easy task, and it’s why we have so much appreciation for our engineering team!
Delivering verified impressions is something we’re extremely proud of at Grocery TV, and I’m so grateful to work alongside the talented people who helped build and continue to develop the technology necessary to make it all function.
As a company, we’re committed to transparency. We’ll continue to reflect that not only with our employees and clients but also through our measurement standards.
About the author:
Marlow Nickell is the co-founder and CEO of Popspots, a retail technology company that aims to build a better retail experience. While initially focusing on improving marketing and merchandising performance in the checkout aisle, their long-term vision is to simplify how brands and retailers market and manage their products throughout the store. Their technology tracks out-of-stocks and product placements, while also serving as a digital advertising space known as Grocery TV.
Help us Help You, by keeping an Independent Voice for OOH.
If you read OOH Today via LinkedIn thank you. To ensure you never miss a post and keep unconstrained speech for the OOH/Billboard Industry, please subscribe. Click here to ⇒Subscribe to OOH Today.