Perspectives Product

Adaptive Timeout™ – Incorporating Adaptive Machine Learning into Header Bidding

Index Exchange is excited to offer a new feature to our Publishers – Adaptive Timeout™. Today, all wrappers available to Publishers use a fixed-value timeout, and most Publishers use a single fixed timeout for all of their header bidding supply, regardless of factors such as device type or the user’s network conditions. This presents a significant opportunity for optimization. With Adaptive Timeout, our Publishers now have an adaptive machine learning feature that accounts for these variables. This feature helps increase the number of bid responses for all Bidding Adapters that Publishers can evaluate and increases Publishers’ revenue while improving user experience. Static timeouts led to a trade off, whereby publishers had to sacrifice user experience to wait longer for header bidding, or sacrifice revenue by setting aggressive timeouts.

Header Bidding With Adaptive Machine Learning – Why it Matters

Across all Publishers using the Index Exchange wrapper, the average timeout for all Bidding Adapters is around fifteen percent, which led to publishers missing opportunities while waiting for bid responses in latency-constrained environments. Because not all potential bids are being evaluated due to timeouts, a Publisher can lose out on many bids, including the highest bids from their demand partners. We realized that by creating an adaptive machine learning algorithm that accounted for device type, geographical location and network conditions, we could optimize the Publisher’s Wrapper timeouts.

How It Works

Header Bidding With Adaptive Machine Learning - How it works

The Index Exchange Adaptive Timeout is an adaptive machine learning algorithm based feature which optimizes a Publisher’s wrapper timeout for each individual user on each individual page view. This approach maximizes a Publisher’s revenue without any development work, while enhancing the user experience. Adaptive Timeout allows the Publisher access to an adaptive machine learning algorithm to adjust the wrapper timeout value as the network conditions for the user become faster or slower. It also takes into consideration other factors such as the device type and the geographical location of the user. With all of these inputs, Adaptive Timeout can ensure that the optimal time is given to header bidding and the optimal browsing experience is given to the end user.

Header Bidding With Adaptive Machine Learning - how it works

“Our long-standing partnership with Index Exchange is a testament to the company’s continued commitment to innovation and maximizing publisher revenue,” said Jeremy Hlavacek, Head of Revenue, Watson Advertising. “This latest feature will ensure more bid responses are making it into our ad server, while improving user experience. It’s a huge opportunity for revenue growth.”

To learn more about Adaptive Timeouts, head to our Knowledge Base.



Leave a Reply

Your email address will not be published.