At Bannerconnect, we’re lucky to be one of the first parties to get our hands on AppNexus’s game-changing APB. For those not yet acquainted, APB (AppNexus Programmable Bidder) allows any advertiser or agency to upload their own version of the algorithm to the AppNexus platform. We’ve been working in partnership with AppNexus to test APB on real campaigns to help fine-tune the technology. The ability to program your own algorithm comes with huge promise, and APB delivers; we found significant improvements in campaign performance when using APB compared to typical campaign set ups. It also ended up saving a decent amount of time on campaign management. That’s something to get excited about.
A trusted partnership
Bannerconnect was chosen by AppNexus to run alpha tests on APB to see it in action and quantify the advantages. We’ve always had a good relationship and have researched AppNexus algorithmic bidding extensively, so it was a good fit. We were able to provide insightful feedback to AppNexus while exploring the impact of APB on campaign performance, and how much time it could save on campaign operations.
What we did
We tested APB on a real campaign for a leading automotive brand*. The set up allowed for nine distinct segments, each containing visitors that viewed a specific kind of car. Historically, the segments performed similarly; the minor differences meant it wasn’t worth setting up separate bid strategies per segment. The test was performed in three different parts:
First, a benchmark campaign. As the name suggests, this was our control. The benchmark campaign treated all nine segments equally without additional recency targeting.
Second was standard recency targeting; a typical campaign treating all nine segments equally, split into four different recency settings.
Finally, APB targeting; a single APB campaign. This campaign grouped the nine segments into three groups with specific bid strategies per group. Next to that, each group was split into nine different recency settings. APB saved us setting up and managing 27 different campaigns.
What we found
Performance sees a drastic improvement when using APB compared to both the benchmark and the standard recency targeting campaign. APB achieved a 76% lower eCPA than the benchmark campaign and a 35% lower eCPA than the standard recency targeting.
APB also saves campaign operations time: The APB campaign required 60% less time on campaign management than the benchmark campaign. While the benchmark campaign needed less time to set up, it needed much more optimisation to reach an acceptable performance. Compared to the recency targeting campaign, APB required 20% less time on campaign management. The main time saving was setting up new recencies. The standard recency retargeting campaign required extra campaigns to set up, where APB only needed a small extension to the bonsai tree.
So there you have it. Bannerconnect has found that APB will give advertisers much more specific targeting and better performance on a campaign, while saving time on campaign set up and optimisation. By being able to make the algorithm do whatever you want, advertisers have the power to differentiate and, as AppNexus CEO, Brian O’Kelley, puts it “Build your own secret sauce”. We can’t wait to implement APB with more of our clients’ campaigns and help their advertising go further. APB is going to revolutionise campaign management, and we’re proud to back it. How will you build yours?
*Background on Retargeting
We decided to focus our trial with APB on a retargeting campaign, as inventory available in alpha phase was limited. It also happens to be one of our areas of expertise. We found that to increase the performance of a retargeting campaign, it’s best to create a separate bid strategy for recent website visitors, known as recency targeting. Next to this, not all segments perform equally; a product page generally performs better than a home page and is therefore more valuable. Implementing these two strategies together typically requires a lot of separate campaigns to set up and manage.
https://www.bannerconnect.net/wp-content/uploads/2018/10/featured-image-livable-code.png400400John Hotterbeekxhttps://bannerconnect.net/wp-content/uploads/2015/09/logo_bannerconnect_72.pngJohn Hotterbeekx2018-10-25 10:00:392018-10-25 13:56:48Livable code, Embrace the Practical Mess Part I
https://www.bannerconnect.net/wp-content/uploads/2016/08/Alice-in-Dataland-02.jpg400400Alice Bezetthttps://bannerconnect.net/wp-content/uploads/2015/09/logo_bannerconnect_72.pngAlice Bezett2018-04-03 10:00:372018-05-16 17:11:29Regression Models for Trait Prediction Part II
https://www.bannerconnect.net/wp-content/uploads/2017/12/you-can-learn-how-to-analyse-featured-image.jpg515515Kristof Soomershttps://bannerconnect.net/wp-content/uploads/2015/09/logo_bannerconnect_72.pngKristof Soomers2017-12-05 10:10:582018-05-17 11:04:29You Can Learn How to Analyse