Video is going Digital. What does it mean for advertisers?


In 2016, for the first time, according to Interpublic Group’s Magna report, digital ad sales surpassed traditional television ads in the US with a budget of $73B vs $70B. Furthermore, the expected social video ad spend in 2017 is expected to double. After decades of the television ruling as the dominant medium for video ads, digital ads are now taking the lead.

This isn’t surprising, as we all see the consumers are getting more and more of their content from digital video – a trend led by Facebook Live, Snapchat, YouTube, Netflix, and others.

As Mark Zuckerberg stated before – “We see a world that is video-first with video at the heart of all our apps and services.” Video has always been the most effective, most engaging advertising method and the industry and publishing platforms are well prepared for the rise of digital video by introducing innovative ways to deliver, analyze and create video ads.

Digital media has always been differentiated by its ability to adjust rapidly to the audience, and be hyper-local and hyper-personalized. In digital text and image ads, ad creative often shifts based on the data collected about the user (location, browsing history, demographics etc.)

In video, however, producing and personalizing ads can be a long and an expensive process, but without it, we won’t be able to gain the full benefit of digital, applying traditional optimization methods like A/B testing.

Just imagine a short 30 second video. It can have so many pieces of creative – imagery, texts, fonts, colors, the scenes, and many more! One short video can have more than a million video variants, each just a little bit different from the other, but with a tremendously different effect on KPIs and conversion rate.

Preforming traditional A/B video optimization isn’t possible for more than a million video variants – it will be too expensive to produce and will required huge media buys to measure its efficacy. However, by leveraging Artificial Intelligence and Big Data analysis this is made possible.

Machine Learning and Artificial Intelligence plays a critical role in optimizing video ads content. By measuring consumers’ engagement with the content, the machine can render and deliver incremental tweaks to the video, and with a closed feedback look – measure the impact of each change to determine the next change, it can quickly evolving a given video to its best performing variant.

Doing this at Wylei, our team was able to show uplift higher than 35% for our client’s Facebook video ads. By using reinforced machine learning and the K-multi armed bandit approach, we iteratively render in real-time new video variants and optimize the active variants in closed explore-exploit test-control environment, reaching the most optimal variant. This is an exciting era in which we can achieve personalization at scale leveraging artificial intelligence.