The future of out-of-home advertising at London City Airport
By Kaloyan Pashov
Have you ever wondered why cookies that track your preferences and behaviours on the web are so prevalent? Most websites rely on advertising (or the revenues they create) to survive and companies have been building ever more sophisticated tools to measure campaign effectiveness and maximise advertising outcomes. Vast leaps in computational performance have unlocked business models that would have been impossible before — next time you load a webpage, hundreds of different ad buyers will bid to show an ad specifically to your profile. There is no need for an advertiser to know your exact identity — your name, age, and gender are less indicative of your purchasing habits than your purchasing history for example. Organisations like Google, Amazon and Facebook have been assembling vast stores of data, creating profiles for internet dwellers that maximise metrics such as number of “impressions”, “actions” and “click-through”.
Let’s compare this digital advertising approach to what happens in the physical world. Out-of-home advertising (billboards, media walls, screens in the Underground or shopping malls etc.) is playing catch-up. Procurement of advertising media typically goes through several parties. Lead times can span weeks, from generating new content to eventually seeing it on display. Once new campaigns are launched, there are few ways of measuring engagement with the content and the fidelity of those that do exist are incomparable with traditional digital channels. At best, aggregate and manually gathered data such as the number of passers-by and postcode-based demographics, is used to justify variations in price between locations and season. Physical advertising media doesn’t have the benefit of years of well-understood baseline data online such as click-through rates, making it very difficult to measure (and justify) the return on investment for out-of-home advertising.
To try and tackle this conundrum and begin exploring the future of out-of-home advertising, we approached London City Airport (LCY) with some ideas. LCY are continually looking for innovative improvements to the services they offer and as such were keen to test new technologies that might help inform data-driven sales conversations with brands. Deloitte’s IoT Studio, a rapid prototyping team that specialises in the convergence of physical and digital experience, deployed a solution that used pre-trained Machine Learning algorithms to identify data points on faces and track them within an image captured by a camera above the screen. Sounds creepy, but the method is completely anonymous and GDPR compliant, much like the techniques used on a website. The data generated enabled the team to measure and analyse engagement with the content being displayed, ultimately helping LCY to consider how their advertising space can be used more effectively to help brands identify what content airport visitors find most engaging. The data was mostly in line with expectations, but analysis did reveal several key insights. For example, engagement rates declined in line with increased numbers of people in the airport — passengers just wanted to get through the lounge quickly and to their gates.
In terms of the technology itself, a cloud service was initially proposed to do most of the heavy lifting but was quickly dismissed due to privacy and bandwidth concerns. A Full HD image can be up to 20Mbps per camera.
The team then examined what they had in their kitbag, such as low-power, small footprint computers with cameras. These did not meet the requirements, however, mostly because of low frame rates, but also for the following reasons:
- A high-resolution system was required to capture data points on the faces of people located relatively far away from the screen.
- It was important to capture more than a couple of frames per second: The team’s research indicated that interactions with content tend to be quite short — and the data gathered at LCY confirmed this.
- To minimise development time; optimising a model would have required a significant amount of development time.
The team used a custom-built computer housed in a server room nearby, a standard, off-the-shelf CCTV camera, and a custom-built engagement detection solution.
The results
Over the course of a 6-week trial, nearly 100,000 impressions were analysed — further validation that processing data from video is best handled at the edge.
- Where do people engage from?
One of the first hypotheses to be tested qualitatively was that it is difficult for people to engage with any content due to the layout of the space. This was confirmed by the data — there were hotspots of engagement in places where people could easily turn towards the screen, as well as in areas where people were walking towards the screen. It was not entirely surprising to find out that people who sit with their side to the screen have lower engagement rates.
2. When do people engage?
Occupancy vs. Engagement throughout the day (Week 22)
Do more passengers in the area result in more engagements? Generally, this is how advertising space is sold now — by looking at the total number of people that go through an area. This was found to be correct in absolute terms; the greater the number of people there are in the area, the greater the number of people who will turn to engage with the content. Engagement time stayed relatively constant throughout the day.
However, there is a caveat — while there are more engagements in aggregate, the proportion of people that engage with content decreases, significantly. Looking at busier periods, it was found that around 1/3 fewer people would engage at that time compared to calmer times.
3. What type of content is most engaging?
The Deloitte team created three different types of content to measure engagement. Engagement data from the existing campaigns running was initially captured, providing a baseline. The three subsequent pieces of content uploaded proved more engaging. The most engaging piece of content was a quiz campaign, which allowed airport users to pass time by playing along.
Next steps
New and emerging technologies are disrupting existing business models and ways of working, challenging the norm and making us all think about how we operate. Out-of-home advertising is one area set to be disrupted in the coming months, new technology enabling us to gather rich data for the first time in an area previously plagued by a data scarcity. While a variety of players have offered point-solutions in this space for a few years now, they have tended to be expensive and cumbersome due to technical maturity and connectivity costs. This has put off companies from investing in these emerging technologies and so consequently the out-of-home advertising market rarely changed.
As our research shows, old presumptions aren’t always true though. Take the findings around footfall in comparison to the engagement rate. The rate decreased the busier it got, which goes against the logic of traditional marketing pricing. What can media owners learn about the best way to engage target audiences? Perhaps we are about to see a disruption of the existing brand engagement model. As technical maturity and connectivity costs become increasingly less challenging, a host of innovation opportunities will open which have the potential to shake up the entire sphere of marketing.
