Did you miss the session at the Data Summit? See on-demand here.
This article was contributed by Jeremy Fann, CEO and co-founder of Cognitive,
It’s easy to see the appeal of a self-service programmatic ad purchase. Instead of limited transparency on pricing and placement, advertisers can direct exactly where their campaign costs go and how much they pay for each impression. Yet, as many merchants have unfortunately discovered, this freedom comes with serious costs. Not only does it take a lot of time and energy to effectively optimize efficiency, it is also extremely difficult to produce consistent results on the required scale. With in-depth study at their disposal, advertisers can avoid this endless slog of tedious, unsuccessful work and instead rely on AI algorithms distributed through integrations such as Dynamic Private Marketplace (otherwise known as DLID or PMP), which automatically Will constantly optimize to the maximum. Operation
At the moment, there seems to be a fairly even division between the number of brands and agencies that rely on managed services and the number of organizations that prefer self-service. A recent survey by Advertiser Perceptions found that 56% use some form of managed service, while 46% say they use self-service. At the same time, 52% of buyers reported increasing their self-service costs this year, while only 17% intended to increase their spending on managed services. The same survey found that The Trade Desk, Amazon Advertising and Yahoo! With each passing year the majority have become self-service platforms, which speaks to the widespread desire for more transparency in programming, especially in terms of fees.
However, despite all the optimizations around self-service, it does not address many of the challenges in display advertising. For example, self-service requires a large team of traders to effectively manage day-to-day operations. With new information coming in about market conditions, consumer preferences, trends, etc., merchants need to be able to quickly synthesize that information in order to run campaigns efficiently and accurately. However, humans are not robots – we need time to examine information and analyze related patterns before we can devise an effective strategy. Given the many marketing teams that operate under tight deadlines, there is no real way for marketers to consistently generate fully-optimized campaigns, leading to wasted costs in the long run. It is also rare for a trading team to have enough traders to effectively cover all of their clients’ campaigns. They usually spend most of their time on two or three of their most important clients, while the rest get less time and effort.
This system of continuous, non-scalable trial-and-error also makes it incredibly difficult to operate on a scale. Plenty of tricks start strong but quickly fade, so marketers are scrambling to come up with new ideas trying to manually optimize their campaigns. This only serves to make self-service programmatic more boring and inefficient – and more difficult for marketers to achieve long-term success.
The number one reason given by most advertisers transitioning to self-service, according to Advertiser Perceptions, is the “desire for visibility in programmatic fees.” More than half (56%) of advertisers cite fee optimization as the primary reason for transitioning – which is understandable given the stress caused by the epidemic on the marketing budget. If advertisers are unable to effectively optimize their costs, switching to manual self-service may not result in the expected cost savings.
Self-service is here to stay, but lack of time and difficulty finding scalable tricks is a huge limiting factor for success. Self-service advertisers should find solutions to these problems. Solutions that will do the tedious task of optimization for them will free them up instead of being able to cover all of their campaigns equally and focus on strategy and long-term concerns. In particular, various forms of machine learning, such as deep learning, are used by brands such as DoorDash to ensure that they can optimize their advertising costs on a scale.
Deep learning is a valuable tool due to its self-learning, constantly evolving predictive capabilities. For example, if you want to train a deep learning algorithm on customer data, they will be able to identify key features of your target customer and use that information to predict how new prospects will respond to your ad. This enables the algorithm to avoid advertising for people who are unlikely to be converted, and respect those who are. Best of all, he does this automatically and in real-time, and he will adjust his predictions as he learns more about your audience and how they respond.
These algorithms are sophisticated and powerful enough to individually evaluate each media buying opportunity – meaning that instead of setting arbitrary rules about who to target, the algorithm itself decides who is worth investing in and who to avoid. As a result, it enables a separate and dedicated analysis of campaigns even when the campaign is live so that real-time market conditions and consumer behaviors are constantly considered.
Manual self-service advertising is difficult, if not impossible, to master in the long run and on a scale. Ignoring the complexities of human behavior, it is necessary to take into account many different factors that may change in a moment’s notice. With deep learning algorithms, self-service advertisers can be relieved of the never-ending pressure to improve performance, and instead rely on technologies such as AI-powered dynamic PMPs that will continuously optimize campaigns and ultimately manage all of their tasks effectively. Will give time to do. Campaign
Jeremy Fan is the CEO and co-founder of Cognitive,
Welcome to the VentureBeat community!
DataDecisionMakers is where experts, including tech people working on data, can share data-related insights and innovations.
If you would like to read about the latest ideas and latest information, best practices and the future of data and data tech, join us at DataDecisionMakers.
You might even consider contributing to your own article!
Read more from DataDecisionMakers