Artificial Intelligence or AI has been a fashionable buzzword over the last decade across different industries and the ad tech industry is no exception. Take a look around the ad tech landscape and it will be hard to find a platform that doesn’t claim to have AI. Because of this, it is difficult to discern which platforms truly have it and which ones don’t.

So how can one differentiate between the powerful and not-so-powerful AI? Here are three factors to consider when evaluating ad tech AI:

1. What is the source of the data?: AI systems are only as good as the data that they operate on. AI that operates on stale third-party data is making decisions based on outdated, often loosely classified inputs. This leads to substandard decision-making and, ultimately, to ineffective advertising. To get it right, systems should not only operate on first-party real-time data, but also have the infrastructure to handle this real-time data–i.e., infrastructure that can handle massive scale (big data) at a very low latency (minimal delay).
2. What type of AI? Not all AI techniques are the same. The term AI is incredibly broad, encompassing everything from robotic automation and image recognition to natural language processing, machine learning, and more. Some in the ad tech industry claim that a set of automated scripts is AI. However, all these scripts can do is adjust a handful of input factors to optimize campaigns, whereas advanced techniques such as machine learning can first learn and then adjust thousands of variables to drive desired campaign outcomes, at scale. As a simple analogy, the former is akin to implementing cruise control in a car, while the latter is more like building a self-driving car. Being able to influence human decision-making through perfectly timed, uniquely relevant advertising is a complex, nuanced problem that cannot be reduced to a handful of input factors. This means that to truly deliver superior business outcomes, AI engines in ad tech need to deploy machine learning methods (e.g., neural networks).
3. How closely does the AI interact with the raw data? AI is much more effective when operating on the raw data as directly as possible. Manually abstracting the input down to a few arbitrary variables before the AI processes it can dramatically reduce its effectiveness. Instead, operating directly on the raw data allows the algorithms to self-learn features that matter (a technique called training) and then optimize those features to achieve the desired goal. As an example, an AI image recognition system built to classify different types of cats would get much better results if the algorithms used the raw input images with all the details instead of manually abstracting the images down to just a few factors such as color, size, and features. Similarly, in ad tech, AI that adjusts a handful of arbitrary bid factors like age, gender, and location to evaluate the value of a bid are not as effective as AI that directly operates on the raw data and lets the machine learning algorithms learn and adjust the factors that matter. 

Quantcast's DSP and its powerful capabilities

When we were building the AI and machine learning engine that powers the Quantcast Advertising Platform, the above considerations were definitely top of mind. While the job of the demand-side platform (DSP)’s user experience is to keep simple things simple, our AI/ML engine was tasked with making complex things possible (and yes, we drew inspiration from Alan Kay’s famous quote). 

Below are the three key capabilities that make it a powerful AI engine: 

Advanced audience analytics: The platform operates on real-time data, which means it handles massive scale (up to 20-40 petabytes of data processed daily) and can operate at a low latency, or minimal delay. This is made possible with a data analytics system built from the ground up–one that turns huge amounts of data into an insights playground. It queries a database of over a trillion online signals in under 100 milliseconds to provide an interactive and instantaneous experience. It is this kind of surrounding infrastructure that helps realize the true benefits of AI. 

Real-time predictive modeling: The platform is able to react to the most recent events across the internet and capture ever-changing consumer behavior, understand consumer interest, and infer consumer intent. To achieve that kind of intelligence and sophistication, it uses advanced machine learning algorithms to build custom predictive models for each campaign. In addition to these campaign models, it also builds media models for viewability and brand safety as well as general models, such as a topic model of the open internet. All this modeling is done using advanced machine learning techniques such as neural networks and, in some cases, deep learning (topic modeling). 

Autonomous campaign execution: The Quantcast Advertising Platform focuses on outcomes. Practitioners are able to specify goals and KPIs and have the confidence that the platform delivers those outcomes without having to manipulate a handful of input factors. This is achieved with a closed loop system that analyzes real-time audience behaviors as well as live campaign performance, factors in campaign objectives, and autonomously tunes 10,000 variables every minute to optimally bid on each impression. These are not human-defined variables, but those that the platform's AI/ML engine has itself learnt and found to be highly relevant to campaign performance. 

In summary, this powerful engine can: 

• Process real-time data at scale with custom built data analytics and predictive models
• Turn data into insights and intent using advanced machine learning techniques 
• Self-learn the factors that matter for effective advertising campaigns and autonomously control them in a closed-loop fashion.

It empowers you with: 

• Fast access to detailed audience and campaign insights to fuel experimentation and innovation
• Perfectly timed and uniquely relevant advertising based on consumer behavior, interest, and intent 
• Greater ROI, at scale, leading to superior business outcomes 

AI powers superior outcomes

The results speak for themselves. It takes brands and agencies less than 3 minutes on average to set up an advertising campaign with the Quantcast Advertising Platform; with many other DSPs, set up can take anywhere from 30 minutes to several hours. Campaigns executed with the Quantcast Advertising Platform have 120% better performance than goal on average. This ease of use, efficiency, and effectiveness enables practitioners to focus their creative energies on driving advertising and marketing innovation. The confluence of human ingenuity with technology is what is delivering the best outcome for you, our customers.

"Quantcast has proven its ability to ingest and process web signals and build sophisticated targeting for me. It leverages machine learning on top of that to drive progressive improvement in campaigns.” - Mike Ru, Sr. Marketing Manager, Microsoft

“The power of Quantcast’s algorithm, machine learning, and user scoring has just driven best-in-class performance at scale for us.” - Alex Grover, Digital Director, MediaCom

“Where Quantcast has been really effective for us is leveraging the technology of Quantcast and the AI to identify audiences who are qualified and likely to respond.” - Jason Crawford, VP, Head of Display Media, iProspect, Dentsu

Our vision for an AI-powered platform

We built a team and a DSP that could realize our vision of using AI and machine learning to achieve results for you. Quantcast started this effort years ago with a large investment in R&D so that we could not only apply state-of-the-art technology, but also innovate and push the envelope of what is possible. We are proud of our large and growing team of engineers, innovators, and dreamers that made this possible. With over 100+ patents and a thriving R&D team, we are taking on the challenges of the future by driving continuous innovation and transformation to help you achieve your advertising objectives. 

Learn more about the Quantcast Advertising Platform here.