Take control of your monetization strategy with machine learning: Introducing MoPub Dynamic Waterfall

January 15, 2018

Tags: 2018, Publishers

Calvin Yeung


See all blog articles

Note as of March 2018: While we have observed positive impact to the yield of certain apps thanks to Dynamic Waterfall, we believe the performance could be even more impactful with additional investment. With that, we have decided to temporarily turn this feature off, which will allow us to iterate more quickly and offer new solutions in the future. Please feel free to reach out to your MoPub account manager or MoPub support for more information.

Autonomous driving is all the rage now. Experts and industry leaders predict that no human will ever have to drive themselves within the next 10-15 years. However, before we completely relinquish controls to a bunch of sensors, cameras, and lines of code, it’s worth noting that we’ve made many iterative progressions since the Ford Model-T -- from manual to automatic transmission, cruise control to adaptive cruise control to various smart safety features such as lane-keep assist, sign recognition, and more -- to make our driving experience safer and more enjoyable, while still being in control.  

In a similar vein, MoPub believes that a transparent and flexible platform that combines dynamic optimization with human control is the ideal solution for publishers in optimizing their monetization strategy.

On the heels of some recently introduced machine learning-powered enhancements, we are excited to announce Dynamic Waterfall, a dynamic optimization feature aimed at increasing yield and improving user experience by reducing latency in ad response time. 

Based on data and conversations with our publishers, we understand that publishers are limited to choosing between optimizing for either CPM or fill rate, and latency data isn’t always readily available when setting the waterfall.  This often leads to unintended negative consequences for the publishers, such as:

  • Demand sources that rank high in priority due to their historical higher CPM could have low fill rate, thus increasing latency of filling the request.

  • Cumulative latency from repeated unfillable attempts could ultimately lead to time-outs and missed revenue for the publisher.

  • Both of the above combined would ultimately lead to suboptimal revenue and poor user experience for the users.

We want to be able to increase as many buyers bidding/buying our inventory as possible without adding more latency because we know adding latency would be detrimental to our success.

—Nolan Fox, Director of Ad Operations, The Meet Group

To address this issue, MoPub has launched machine learning powered dynamic optimization that looks at each of the network’s historical fill rate and latency, in addition to the eCPMs, for a particular ad unit and automatically re-ranks them within the ad waterfall priority (only within priority 12 for now) to maximize for impression and revenue.  Our beta partners have seen positive increases across fill rate, CPMs, and publisher revenue for their participating ad units.  

With Dynamic Waterfall and continued investments in machine learning that complement the built-in flexible controls on the platform, MoPub is committed to providing a full mobile monetization solution that is designed to assist publishers control their own destiny, rather than just hope for the best.

To learn more about how it works, check out this 1-sheeter.

Dynamic Waterfall is turned on by default for all publishers and there’s no action needed from you. If you have any questions, please see our support docs or contact your account team or support@mopub.com.


Contact us

Contact us

Ready to maximize your mobile revenue?

Girl with phone