The Power of Machine Learning to Enhance DSP Campaign Performance

Every day, Amobee’s Demand Side Platform (DSP) receives billions of ad requests. Each of these requests is a unique opportunity for brands to reach target audiences on their favorite website, app, video, or connected TV. Powered by AI, the Amobee DSP ingests all of the data from these requests to generate smart insights for the right price, the right property, and the right time for brands to reach their audience efficiently at scale. 

Delivering accurate predictions is more important than ever, given the rising costs of Click Per Acquisition (CPA) in the advertising industry and tightening ad budgets amidst broader economic uncertainty. Recently, Amobee’s Data Science and Engineering Teams leveraged machine-learning to enhance the algorithm that predicts Action Rate at bid time to make it more accurate. Let’s take a look at how Amobee’s DSP utilizes machine-learning, and how the recent enhancements to the algorithm improved CPA for major retail brands.

AI Tailored to Campaign Goals

When an advertiser creates a campaign in the DSP, Amobee AI takes over,  optimizing each aspect of the campaign to deliver campaign goals most efficiently. When the opportunity to show the ad appears, Amobee AI answers two fundamental questions: 1) to bid or not to bid for the impression opportunity, 2) how much to bid? The Amobee DSP utilizes machine learning and control theory-based predictive algorithms that combine data about the user, website, web page, campaign, and ad creative attributes to make predictions of click and conversion rates for each ad display impression. These predictions are combined with a campaign’s viewability, completion, CPC, or CPA goals and campaign delivery settings to generate an optimal bid response for each request. 

In adtech, it’s common for DSPs to use all-purpose machine learning models, but Amobee’s data scientists and engineers have found that there is no one-size-fits-all approach to AI; a variety of advanced machine-learning techniques drive better campaign performance. 

Model Chaining – When an ad request is received, depending on the nature of the ad opportunity, Amobee AI “chains” several models together, each serving a specific purpose, in a sequence that uses the output of one model as the input for another one, which optimizes the response to the bid request.

Model Splitting – Amobee AI uses models tailored to an advertiser’s campaign goals and inventory, whatever those goals may be. These specific models outperform all-purpose models in head-to-head comparison, optimizing to an advertiser’s campaign and budget goals.

Transfer Learning – Amobee AI leverages insights gained from running one model to create new, more optimized models that help deliver on campaign goals. 

Rules, Heuristics, Math, and Models – AI capabilities are leveraged throughout the Amobee stack and across every phase of the marketing cycle for advertisers. Both proprietary and non-proprietary technologies are leveraged in combination with machine learning, simulation, statistical models, heuristics, and rule-based techniques to optimize campaign outcomes. 

Amobee’s data scientists constantly monitor, evaluate, train and retrain the DSP’s AI models to improve accuracy, reduce media waste, and meet each campaign’s goals. 

Enhanced DSP Algorithm Improves Platform-Wide CPAs

Recently, Amobee’s data scientists and engineers enhanced the DSP algorithm that predicts action rate at bid time to improve accuracy. Platform-wide CPAs improved by 33% while two major DTC retailers saw 36% and 39% improvements respectively, with minimal impact to campaign scale/pacing, outperforming competitive platforms including search and social.

[TABLE FROM SALES SHEET]

The Amobee team is now in the process of enhancing machine learning for additional CPA, CPC, Viewability Models, as well as making optimizations to the win price prediction models and bid scoring flows. Amidst rising industry costs and ongoing economic uncertainty, optimizations like this are more valuable than ever to campaign performance. 
Curious to learn more about Amobee’s solutions for advertisers? You can read more success stories, watch a demo, or contact us.

 

About Amobee

Founded in 2005, Amobee is an advertising platform that understands how people consume content. Our goal is to optimize outcomes for advertisers and media companies, while providing a better consumer experience. Through our platform, we help customers further their audience development, optimize their cross channel performance across all TV, connected TV, and digital media, and drive new customer growth through detailed analytics and reporting. Amobee is a wholly owned subsidiary of Tremor International, a collection of brands built to unite creativity, data and technology across the open internet.

If you’re curious to learn more, watch the on-demand demo or take a deep dive into our Research & Insights section where you can find recent webinars on-demand, media plan insights & activation templates, and more data-driven content. If you’re ready to take the next step into a sustainable, consumer-first advertising future, contact us today.