Predicting how users will engage with ads can be challenging. Traditionally, advertising companies use cookies and embedded monitoring codes that produce a record of how users engage with ad content. But when it comes to predicting user engagement ahead of time, a novel approach is needed.
Zhong Ding of Xinjiang University and Xing Feng Fan of Sichuan International Studies University demonstrated a new, unique approach to predicting user engagement in a paper written for the 2024 International Conference on Interactive Intelligent Systems and Techniques (IIST). Ding and Fan outline how to use a deep learning model to predict user click-through rates on ads. Their solution enables marketers to maximize their ad ROI and reduce the amount of time it takes to train predictive deep learning models.
How AI Interprets User Interaction Data to Gauge Engagement
Ding and Fan’s AI interprets user interaction data to assess engagement by using recurrent neural networks (RNNs). RNNs do this by employing:
- Sequential data analysis. This involves the RNNs analyzing historical user behavior, such as their clicks and views. This enables researchers to identify trends regarding how users engage with ad content.
- Layered processing. Deep learning models process data using multiple layers, gradually mapping the data up to higher levels of abstraction. In this way, RNNs surface insights that make it clear how users are engaging with ad content.
- Prediction of click-through rates. Ding and Fan use RNNs that predict click-through rates, which makes it easier for marketers to know how users will engage with their ads.
- Using dropout technology. Ding and Fan use dropout technology, which omits nodes at random during the training process. In this way, it prevents the model from memorizing training data (also known as “overfitting”) because it forces many neurons to learn genuinely useful representations.
One reason RNNs make it easier to accurately gauge user engagement is that they use recurrent connections. This makes it possible for RNNs to leverage both past and present data at the same time. By combining historical and current information, RNNs can do a better job of predicting how users will engage with ad content.
What These Findings Mean for Ad Creatives, Placement, and Timing
Ding and Fan’s discoveries have a considerable impact on the work of ad creatives, as well as strategies around the placement and timing of ads.
Ad Creatives
For ad creatives, the study found that emotionally charged ads can have a significant impact on shaping positive attitudes in audiences over time. Also, AI can make it easier for creatives to identify the visual and promotional features that resonate best with their target audiences. Using this data, creatives can deliver ads that are more engaging and effective.
Ad Placement
Ding and Fan also outline how to optimize ad placement, specifically using demand- and supply-side platforms (DSPs and SSPs), as well as visual chain technology. DSPs and SSPs ensure your ads are reaching the right target audiences. Visual chain technology enables users to access product information as well as click on purchase links without having to navigate to another screen. This can boost click-through and conversion rates.
Timing
Ding and Fan’s discoveries underscore the importance of predicting the optimal times for displaying ads. Getting the timing of ads right is particularly important when advertising in on-demand environments, such as via YouTube and other online video platforms. Using deep learning, marketers can identify the best times to show ads to specific target audiences.
Ethical Considerations
As is the case with many machine learning breakthroughs, data privacy and transparency arise as core ethical considerations for anyone looking to use Ding and Fan’s work to improve the effectiveness of ads. It’s important to ensure any system built adheres to standards, such as GDPR and CCPA, when it comes to collecting, sharing, and storing user data.
For example, to comply with CCPA, advertisers or brands may need to explicitly convey to users how they’ll be using their data. They also may need to provide a way for users to opt out of having their data used, such as for training deep learning models to increase the accuracy of engagement predictions.
Privacy concerns may also arise if advertisers fail to let users know their data will be shared with DSPs and SSPs. There should be consent mechanisms in place to empower users to decide which data they want to keep private from these systems.
Transparency is the key. Advertisers and brands should use clear, simple disclosures that outline exactly what happens with user data, how long it’s stored for, and who they may share it with.
Future Trends
Ding and Fan demonstrate how their model succeeded using Avazu and Criteo data sets, both of which are widely used datasets for predicting click-through rates (CTRs) for online ads. Given this success, the future could be bright for marketers who use deep learning in a way similar to what Ding and Fan demonstrated, applying AI in consumer behavior analysis.
To learn how you can optimize ad timing, target specific markets, and inspire higher click-through rates using deep learning, check out the full paper here.
Download “Discussion on Advertising Communication” Article
Disclaimer: The author is completely responsible for the content of this article. The opinions expressed are their own and do not represent IEEE’s position nor that of the Computer Society nor its Leadership.
