In the heyday of TV commercial advertising of the 1960s, marketing decisions involved sending out surveys and shelling out for primetime ad placements. The digital age has given us big data and made it far easier for brands to produce engaging, high-quality videos delivered cross-device. However, when it comes to understanding visual content and its impact on audiences, brand insight has failed to keep up.
The marketing ecosystem is multi-channel, multi-platform, cross-device, and technology-driven. It’s chaotic, and the industry is still trying to make sense of all this big data and tie it back to the big picture. As video takes over the online world, marketers are faced with ever-increasing repositories of visual data, and they’re using text-based tools to interpret this content. It’s like trying to understand the complexities of deep space with a magnifying glass.
There is a huge buzz in the industry about how artificial intelligence (AI)—or more specifically computer vision—can change this.
From accelerating programmatic video capabilities, to uncovering semantics in Facebook uploads and analysing human emotions, machines can detect visual triggers invisible to the human eye. And, the reality is, today’s tools barely scratch the surface. So how is computer vision offering new insights and capabilities to hack real brand performance?
1. Hyper-Personalised Ad Selection Through Video Fingerprinting
Computer vision lets marketers add meaning and categorisation to large amounts of video footage. Through facial, object, audio, and speech recognition, machine tools automatically tag every frame with a huge amount of metadata. This could be anything, from a model’s hair color, to their gender, facial expressions, the time of day, or any object in a single shot.
This process of transcribing video, adding semantics and defined categories, can be used to catalogue brand video, enabling hyper-personalised ad delivery in real time. A combination of this AI and programmatic creative changes allowed Unilever’s Axe to deliver 100,000 different versions of its “Romeo Reboot” campaign, personalised to user tastes. Axe partnered with research firm Box1824 to target different versions of the video to audience profiles, using known insights such as known musical tastes or purchase history to select the most appropriate content.
In this way, brands use computer vision to create customised ad campaigns that will resonate with individuals. Machine learning can also help to identify larger trends and optimise campaign delivery over time. Marketers can apply this technology to ad placements, evaluating the context, from static pages, text, and images to full TV shows to select appropriate ads in real time. So when Cristiano Ronaldo scores a goal in the game, at half-time you can make sure to place the Nike ad with Ronaldo in it.
2. Extract Audience Insights From UGC
Cataloguing visual brand content to power personalised ad delivery is only the beginning. Users don’t just interact with videos, increasingly they are capturing and uploading their own user-generated content (UGC) via social media.
On platforms such as Facebook, WhatsApp, and Snapchat, users upload and share over 1.8 billion photos a day. And as the digital world turns to video, on YouTube alone users are uploading an average of 300 hours of video every minute. Adding semantics to a growing mass of UGC means brands can gather further audience insights, helping them to derive personal characteristics, tastes, and experiences.
Visual intelligence platform GumGum uses computer vision to scan visual social media content, providing marketing teams with actionable audience information. Using this information they helped L’Oreal’s Feria to target customised ads to women with different hair colours. With techniques like this, brands are able to search social media for visual content that relates to their product or offering. They can respond with campaigns catered to intelligent audience insights, or even augment UGC with branded additions.
3. Synchronised Advertising And Improved TV Attribution
Scheduling digital ad delivery in sync with TV commercials can now be done through computer automation. Computer vision keeps an eye out for the commercial and triggers a complementary ad on mobile or desktop. According to Accenture, 87% of consumers use a second screen. AI provides a way for marketers to take advantage of these habits, linking linear viewership to the digital space, offering tailored calls to action in sync with TV content or paid commercials.
This also means marketers can schedule ads and launch conversations during their competitors’ commercials, or even just in tune with relevant TV content. This means an individual watching a travel documentary about Southeast Asia on TV, for example, could be served a tailored video campaign on their mobile device offering a discount flight to the beach paradise of Koh Phi Phi. In the future, employing computer vision to gather user insights means it could even take insights from Instagram posts of users’ last vacation.
And real-time data that ties this all together gives marketers analytics on attribution, helping them to measure downloads and page visits in correlation with TV ad spots. As attribution windows continue to shrink, computer vision will become essential to accurate analytics across multiple platforms. Furthermore, brand identification in scenes means businesses can measure earned media and also ensure copyright protection.
4. Sentiment Analysis Via Emotive Computing
Today, marketers are also able to take visual data from the other side of the screen, with the use of cameras to measure emotional responses or identify a consumer’s surroundings, to deliver more relevant ads. Agencies such as Omnicom and WPP’s MediaCom are making strides with “feature extraction” and sentiment analysis, measuring audience reactions when exposed to different ads.
M&C Saatchi in London used this form of visual analytics to create the world’s first outdoor AI campaign. Using a camera to capture a viewer’s facial expressions, M&C ran a series of AI posters at bus stops, switching the layout, copy, font, and images through a “genetic algorithm” using a person’s visible emotions, for instance if they looked happy or sad, to build a tailored creative.
Advanced forms of emotive computing, known as affective computing, are also able to detect hidden human responses. For example, software startup NuraLogix uses a process called Transdermal Optical Imaging, which uses cameras to measure blood flow information and determine emotional responses.
Of course, using personal data in this way is controversial, and there are various barriers to brand adoption. Collecting biometric data such as “faceprints” violates privacy acts, both Google and Facebook currently face lawsuits regarding biometric data. While facial recognition is possible on Android, Google does not currently allow this on the MyGlass app store.
According to U.S. media company PSFK, Apple also bars facial recognition via the camera on its phones, relegating users who want to experience this technology to the desktop. However, through its purchase of AI startup Emotient, the tech giant has furthered its own advances in emotional detection. What Apple plans to do with these new-found capabilities still remains to be seen.
Nevertheless, computer vision continues to advance rapidly. Artificially intelligent sight has overtaken human vision and perception. The effect of this is that, as video becomes the digital norm, vast quantities of branded video, UGC, and even consumer emotional responses can be translated, communicating meaning in real time, and making this all searchable for the future.
When paired with big data and response measurements, brands infer valuable audience insights, helping them to co-ordinate marketing activities to create engaging campaigns that work in synergy. As the visual online world gains momentum, AI technology will be our eyes. Computer vision will guide marketers through the noise of all this video content, helping to uncover new intelligence that will transform brand performance insights, light years from the guesswork of old TV advertising.