Index Exchange Is Bringing Machine Learning to Header Bidding

Index Exchange has unveiled a new product called Adaptive Timeout that it claims will balance publishers’ monetization and consumer experiences by using machine learning to maximize ad opportunities without significantly affecting webpage load times. The new product makes publisher page speeds smarter by adapting them to each user, particularly mobile device users, many of whom will have a varying experience based on the quality of their network signal–either Wi-Fi or cellular. Adaptive Timeout helps publishers collect the optimum amount of bid requests for a given impression in an ad auction using a decisioning-logic called header bidding while minimizing the potential negative impact it may have on the consumer experience—i.e., page load time—according to Index Exchange. Critics often cite the negative impact header bidding, which maximizes the number of bid requests as a means of removing the widely perceived Google bias in most online ad auctions, can have on the consumer experience. This is because the wider solicitation of bid requests from media buyers delays page load times. On mobile devices, which have smaller processing power and more variable network signal, this is an even more critical issue. Adaptively modifying the timeout (when it stops collecting bid requests) can ensure the maximum number of bids that make it to the publisher, thereby increasing revenue while minimizing the impact on the user experience in each unique condition, according to Index Exchange. Previously, publishers had to rely on imprecise, fixed-value timeouts for all users, which led to publishers missing opportunities while waiting for bid responses in latency-constrained environments, according to Gabriel DeWitt, vp of product at Index Exchange. “Incorporating machine learning into header bidding is a natural evolution of the product line, and Adaptive Timeout is the first of many innovations we’ll be enhancing with this technology,” DeWitt said. The independent ad exchange has been a vocal advocate of publishers using header bidding to bolster their revenue opportunities, although last year, it was embroiled in a controversy involving a tactic called bid caching, which it later suspended. Bid caching involved holding onto the details of a demand-side platform’s unsuccessful bid request for an individual ad impression and then using the details of that same request to serve an ad on behalf of the DSP on a later piece of content viewed by the same consumer. DeWitt was adamant that Adaptive Timeout in no way involves the caching, modification or bid retention. “Instead, it intelligently adjusts the IX Wrapper timeout value based on the device type, network conditions, and the user’s browsing behavior,” he said. Adaptive Timeout will be followed by a number of other header bidding offerings planned over the second half of 2019.