Optimizing Audio Advertising Campaign Delivery with a Limited Budget

As audio advertising gains wider adoption in the industry, more marketers and advertisers begin to realize the opportunity and importance of audio-ad campaign optimization, i.e. leveraging different tools to increase the return on investment (ROI) of their campaigns. Recent research has shown that measuring advertising effectiveness and showing efficiency of targeting are highly non-trivial. In addition, most of the related literature are on display/video or sponsored search ad campaigns. In this work we present a flexible targeting framework for audio advertising and compare different models' performances on real world datasets. We also run a follow-up A/B test with our best performing model and find it to drive more users to use the new promoted feature when compared to our baseline. The statistically significant positive lift observed for our baseline and our proposed model, as compared to control, demonstrates the ad campaign effectiveness. In addition, our model outperformed the baseline by a large statistically significant lift on all of our metrics of interest, ranging from 13 % to 52 %, which demonstrates the targeting efficiency of our model.