The Role of AI and Machine Learning in Programmatic Advertising
 |
Pathlabs Marketing |
 |
October 26th, 2023 |
In an era burgeoning with technological advancements, buzzwords like 'artificial intelligence,' 'machine learning,' and 'programmatic advertising' have almost become mainstream. Yet, beneath the jargon, there's a haze of misconceptions regarding how these three fields intersect and what we need to understand about each of them. Let's break it down for you.
Understanding AI in Programmatic Advertising
The Basics of AI and Machine Learning
Artificial Intelligence
Artificial intelligence is a field of computer science that aims to create machines and systems capable of performing tasks that simulate human intelligence.
AI encompasses various subfields where scientists are trying to develop machines and systems modeling the different features of human intelligence. These fields include robotics, language learning models, voice and vision recognition systems, machine learning algorithms, content generators, etc.
The ultimate goal of AI is to get these machines and systems to be so complex and in-depth that they can work autonomously, allowing humans to offload many of their remedial tasks to the AI.
However, despite this big goal, we are still at a nascent stage of AI. The main advancements we have seen so far are driverless cars, ChatGPT, chatbots, translators, and data analysis tools. These individually are very complex and exciting AI innovations, but we consider them Narrow AI, as they still require human oversight and aren’t perfect.
Machine Learning
Machine learning is a subfield of artificial intelligence focused on developing machines and systems that progressively learn over time based on training data and refining algorithms.
These machines and systems ingest various data inputs and use machine learning algorithms to analyze the data and develop models that produce outputs. These outputs can be in the form of making predictions and decisions on outcomes, identifying patterns in data, or generating results or recommendations. Then, the system autonomously trains and improves over time based on the accuracy of the output and the more data it takes in, without specific programming.
Although machine learning is a subfield of AI, other AI fields like language learning models and generative AI use it to help the models continuously learn and improve.
The Basics of Programmatic Advertising
Programmatic advertising refers to the buying and selling of digital ad space using algorithms, technology, and automation.
When we say programmatic advertising, we generally refer to the intricate landscape of advertisers buying digital ad space and placing ads while publishers sell space.
Both parties rely on different technologies, platforms, and technical intermediaries to go about this process and facilitate these transactions, making it ‘programmatic.’
On the buy side, a primary way for advertisers to programmatically procure ad space and place ads is by using a demand-side platform (DSP). They build out their campaign in the DSP, detailing objectives, targeting parameters, budgets, and more. Upon turning the campaign live, the DSP autonomously interacts with ad exchanges, bidding in real-time to place ads in front of users it deems relevant.
Learn more about the intricacies of programmatic advertising here.
How AI, Machine Learning, and Programmatic Advertising Intersect
The buzz around AI and ML dominating the world of programmatic advertising is often exaggerated. Many think AI is the main driving force behind the processes involved in programmatic advertising, but this isn’t the case.
The programmatic ecosystem does rely on technology, systems, and algorithms, which are very complex; however, these aren’t explicitly forms of AI or ML.
Instead, when we look at AI and machine learning in the context of programmatic advertising, we are currently seeing the industry apply it in the following three ways:
Generative AI tools
Machine Learning in Programmatic Software
Proprietary AI and ML in Large Tech Companies
Let’s discuss each of these applications a little more in detail below.
The Role of AI and Machine Learning in Programmatic Advertising
Generative AI
Generative AI, ‘genAI’ for short, refers to tools like ChatGPT, Dall-E, and other text, image, and video generators. This AI field primarily leans on machine learning, language learning models, and computer vision. Users can provide a text prompt, then the genAI system’s output will be text, images, or videos.
Not many programmatic software have genAI features, but advertising teams have been independently using genAI to develop ad creative and copy such as ad headlines, descriptions, and SEO terms – they even use it to write entire blogs. Apart from writing copy, genAI can develop images and videos for campaigns or simply assist in brainstorming and fact-checking.
Machine Learning in Programmatic Software
Certain software used in programmatic advertising, particularly demand-side platforms, have begun incorporating AI features. These features mainly leverage machine learning algorithms and models that help improve data analysis, campaign oversight, and performance.
A prime example is The Trade Desk DSP’s Koa feature. Koa uses AI and machine learning algorithms to look at the different decisions the advertiser makes when building out their campaigns in the platform; then it performs the following tasks:
Generates recommendations on channels and audiences based on goals and KPIs the advertiser sets
Helps analyze campaign data to identify patterns and potential insights
Modifies bid rates in real-time to meet advertisers’ pacing goals
AI isn’t running the entire campaign for advertisers, but the additional AI features help guide it and provide more insight.
Proprietary AI and ML in Large Tech Companies
Large tech companies, especially search engines and social media behemoths like Google and Meta, have integrated AI and ML features into their advertising technologies.
We consider these spaces part of the programmatic ecosystem, as they help facilitate the buying and selling of digital ad space using technology, algorithms, and automation. However, they are unique because they primarily sell their internal ad space inventory to advertisers.
Additionally, they operate somewhat opaquely. We know they incorporate AI and ML into their operations, like how Facebook uses machine learning to decide who to place ads in front of or how Google Ads has an AI smart bidding tool.
But, we don't fully understand the inner workings of the machine learning or the precise data they utilize to develop these models. Additionally, they maintain a high degree of confidentiality about auction outcomes and ad placements. Because of this, we consider them ‘black boxes’.
Advertisers can certainly continue to use these platforms for their ad campaigns, as they have powerful reach; they just will not have as much transparency into the AI and ML functions these technologies offer and promote.
The Application of AI in Different Advertising Networks
Search Engine Advertising
Search engines develop machine learning models to analyze search queries and generate search engine results accordingly. They will also incorporate machine learning models to decide which search engine result pages to place paid search ads in.
Social Media Advertising
Social media platforms incorporate AI in the form of machine learning, natural language processing, and computer vision. They use this AI to analyze the content uploaded into the platform, classifying it based on topic, what the visual content talks about, etc.
They will also use machine learning to analyze the data from user activity on the app, serving content and advertisements to classified groups of users. Based on the resulting engagement, the model will continue to improve.
Video and Content Advertising
As mentioned, generative AI is currently the main application of AI in the video and content creation side of programmatic advertising. Teams use ChatGPT to write copy for blogs, landing pages and paid search campaigns while also using tools like Dall-E and Runway to generate video creative.
AI-Related Challenges in Programmatic Advertising
Data Privacy
Data is a tricky topic in AI and the bigger picture of advertising. For any field of AI to work, especially machine learning, there has to be a lot of data available. This data primarily comes from users, operational data, and other sources.
Collecting this data is tricky because teams must follow many regulations and rules to attain it. It also prompts questions on the advertising side regarding whether it is ethical to collect specific data from a given campaign, initiative, or user and then use it to train and develop AI models.
Transparency
As mentioned, many companies, especially tech giants, are considered black boxes, as they do not always disclose how their algorithms, models, and different AI offerings work and arrive at conclusions.
“To further complicate things, the intricacies of a machine learning model and the conclusions they make are often not able to be explained; given their neural networks and many learning layers.”
This can lead to mistrust from advertisers who want to leverage these newer tech innovations but are in the dark about the validity and reliability of these offerings.
Algorithm Bias
Unfortunately, many cases of bias and prejudice within machine learning outputs have occurred. The algorithms may disqualify certain users based on data, or, as we see with generative AI, the system may generate inappropriate content or perpetuate stereotypes. Along these lines, language generators can produce incorrect facts, and users can easily bias the outputs.
Quality
Right now, we can really only use AI to enhance how we do our jobs and boost efficiency. Using ChatGPT, for example, we can leverage this for idea generation and developing content drafts. However, it is not at the point where we can generate a blog in one go and publish it without reviewing. Alternatively, we can take the suggestions that The Trade Desk’s Koa offers regarding audiences to target, but it is crucial to still have human quality assurance.
The Future of AI in Programmatic Advertising
Right now, there are three AI trends to anticipate in the future of programmatic advertising:
More Widespread Adoption
AI is exploding mainly because it helps make our jobs easier. Instead of having to perform as much data analysis, spend time writing content, or make certain advertising predictions, we can rely on AI technologies to do so or at least help. This will continue to be the driving reason for more AI technologies to develop as more people want solutions to make their lives easier.
Discrepancies in Knowledge of AI
We will continue to have people who understand and love AI, while others have a weak grasp and fear it. With more adoption, more users will have to leave their comfort zones and understand how to leverage these tools.
They must also understand that they don’t need to be expert data scientists to use AI and benefit. Instead, it is more about being aware of AI options and using them when applicable. For example, you don’t need to know the inner workings of ChatGPT’s back-end algorithms, but you should learn how to leverage this tool and when to apply it to be more efficient and effective.
AI for Creative Development
Generative AI is taking the creative side of advertising by storm. However, many advertising leaders still want to have a significant role in developing advertising creative (graphics, banners, ad copy, etc.), ignoring the capabilities of genAI.
We anticipate there will be a shift where these leaders have more buy-in and are more comfortable leaning on AI for creative development, also allowing AI and machine learning algorithms to run multiple tests on different iterations of the creative to see which features perform the best.
We may even have a model where advertising content is dynamically created from scratch in real time based on the specific user the ad is about to serve to.
Conclusion…
Artificial intelligence (AI) is a broad area of computer science aimed at building machines that mimic human intelligence, encompassing everything from robotics to language models. Machine learning (ML), a subset of AI, focuses on systems that adapt and grow through data analysis without being explicitly programmed. On the other hand, programmatic advertising involves automated buying and selling of digital ad space using technology, algorithms, and automation.
Currently, the overlap between AI and programmatic advertising is primarily in generative AI tool usage, ML features in programmatic software, and the AI frameworks of major tech companies, especially on search and social media platforms.
As the field progresses, we can expect more technological advancements, increased attention to data privacy, and broader acceptance and utilization of AI solutions.
