By Jeremy Fain, CEO and Co-Founder of Cognitiv
As new streaming services proliferate and more households opt to cut the cord, Connected TV is increasingly the platform of choice for brands looking to reach out to new audiences and reconnect with existing ones. However, as with any other platform, brands have to make sure they are reaching the right households, and that they are able to track that exposure and tie it to a conversion. With deep learning, marketers can be assured that their ad dollars are flowing to the individuals most likely to buy their brands while also gaining valuable insights into consumer behavior.
With traditional TV advertising, brands could never be completely sure that they were reaching the audiences that they hoped to. Moreover, television ads are expensive investments, both in terms of the cost of purchasing a spot as well as the production required to create a sleek, engaging ad. This year, a 30-second spot at the Super Bowl cost advertisers a cool $5.5 million, despite the fact that big brands like Coca Cola, Budweiser and Hyundai opted to sit out this year’s football extravaganza and overall viewership was down by 15% compared to last year.
Advertising on CTV platforms is, for the most part, similar to advertising on any other digital channel. Unlike linear television, which relies on statistical models to make estimates on viewing numbers and audiences, brands have – in theory – a much more precise view into the performance of their CTV ads and better insight into the audiences they reach. Consequently, brands can take what they know about customers on other channels and use that information to target households on streaming services, and they can also use information taken from streaming services (about watching habits, popular shows, etc.) to build out more comprehensive audience profiles.
However, in order to maximize the impact of their CTV advertising, marketers must first figure out which audiences would be best to allocate their resources to, and how to find the right balance between accuracy and scale. Deep learning is especially valuable here, as its ability to plow through reams of data and find patterns is unparalleled, and goes far beyond a human’s capabilities. Marketers can use deep learning to identify audience clusters composed of households who exhibit similar traits and behaviors to their existing customers, thus marking them out as good candidates for conversion. Importantly, these clusters can be any size, comprising anywhere from a handful to millions of prospects, meaning that marketers no longer have to make broad generalizations about who their audience is. Instead, they can use these insights to tailor their advertising to the needs of each cluster, thus giving greater depth and breadth to their reach.
Deep learning is a data-thirsty enterprise, so the more information marketers can feed it, the better the results. Consequently, the data that they get from CTV providers about the households they are advertising to is extremely valuable because it enables a broader understanding of consumer behavior. By parsing data on households’ viewing habits, marketers can also gain more insight into the individuals within the household – for instance, by noting whether or not there are children, and what their relative ages are – which will in turn influence the types of products or ads that are shown. Similarly, what people watch is often a window into their personal preferences and interests, which will help deep learning algorithms narrow down precisely which households to target and which to avoid. For example, if you can see that a particular household enjoys cooking shows, and your research on existing customers shows that they tend to enjoy cooking or food-related subjects, then that is a good indication that you should be targeting that household.
Because CTV ads are only considered viewable if the video actually plays, regardless of how many times the ad has been loaded, marketers can get a better sense of how many people actually saw the ad, and how long they watched it for. A deep learning algorithm can take this information and determine which types of ads to show people depending on how they interacted with previous ads as well as their content and subject preferences, thus allowing marketers to create a variety of dynamic experiences designed to best fit the requirements of the individual or household.
Marketers spent $8.11 billion on US CTV advertising in 2020, and that number is expected to increase to $11.36 billion this year. eMarketer estimates that, by 2024, $18.29 billion will be spent on CTV ads, more than double what was spent last year. With the pandemic, more people are streaming TV than ever, which means that there is more data available on people’s viewing habits and preferences. Consequently, now is an ideal time for advertisers to push more money into CTV, especially considering that many areas of the country are still under some form of lockdown or subject to restrictions.
Every new channel represents a great opportunity for brands to get in front of new audiences, but that opportunity is only worthwhile if brands develop a targeted approach to advertising and a plan to maximize the value of the information they receive. Deep learning helps marketers achieve both of those things with levels of sophistication and accuracy far beyond what marketers could achieve on their own. It is clear that CTV as a channel will hold increasing importance to advertisers in the years to come – but it is up to marketers to decide how best to leverage it.