Advertisers care about users taking actions, whether that action is a landing page visit, a product hover, or a checkout that requires credit card information. In the past, we focused our efforts on solving click and viewability based goals (CPC, CTR, post-click CPA, and viewable CPM) for programmatic buying. After fine tuning these optimization algorithms, we began supporting conversion goals with the release of post-click CPA optimization. While this optimization led to positive performance, it only took into consideration post-click conversions and was built on the backbone of optimization algorithms that value inventory based on clicks. As a result of our system not supporting optimization to post-click and post-view conversions, our clients had to use manual optimization in order to achieve the true goal associated with a media budget. Now AppNexus clients will be able to optimize to a CPA value for retargeting line items.
CPA goals are usually broken down into two strategies: prospecting and retargeting. Generally speaking, prospecting strategies drive users into the conversion funnel and retargeting strategies drive users down the conversion funnel. As we dug into the problem, AppNexus’ Data Science team concluded that we would need a unique solution for each. A prospecting line item doesn’t use first party user data, meaning traders are aiming to find potential converters from the wide span of available users on the internet. The challenge with prospecting line items is discovering inventory and audiences that perform. A retargeting line item targets first party user data. In most cases, this means users who have visited the advertiser’s website or users who are important to the advertiser for another reason, such as having purchased a product in store or having a subscription that is up for renewal. The challenge with retargeting line items is reaching users and bidding the correct price. We are releasing a solution for CPA optimization for retargeting line items first. Below you will see how the new feature removes the need for a trader to manually optimize and how you can use it in APP today.
Apply Machine Learning to Audience Valuation
In the case of retargeting line items, our optimization algorithms help you reach the most valuable users and pay the optimal price. We do this by focusing on the variables that have the most impact on CPA performance: the segments targeted by a line item and how recently users are added to these segments.
Retargeting segments will often reflect the conversion funnel. Therefore, a trader will value each segment and when users are added to a segment differently. Take, for example, a typical e-commerce interaction. In most cases, a user that has added an item to a cart is more valuable than a user that viewed an item. Similarly, someone who added an item to a cart today is probably more prone to convert than someone who added an item to a cart seven days ago. Below is a representation of how a trader may have manually bid on users based on the segment and recency.
The algorithms built for CPA retargeting optimization automate these manual breakouts. Instead of traders estimating what to bid, our machine learning algorithms build a unique data model for each line item that determines which segment contains the most valuable users based on historical conversion data. We then modify a bid based on how frequently and recently a user was exposed to impressions from the advertiser and how recently the user was added to a segment.
In the new workflow, a trader must indicate that a line item is a retargeting line item. After doing so, a user can input a CPA goal and choose a single conversion pixel for our optimization algorithms to use. If the line item is not a retargeting line item, a trader will only have the option to optimize to a post-click CPA.
Insights and Analytics
As is the case with all our optimization products, we empower buyers by sharing insights and data to monitor a line item and make decisions. A buyer can view how the line item is performing against the CPA goal and delivering as well as view other key performance indicators.
We will be releasing this feature to all AppNexus Programmable Platform clients this month. Watch our Release Notes page for the specific date.
Concurrently, our Data Science team is working on building an optimization solution to support a CPA goal for prospecting line items. Expect this second CPA solution this year and, as always, drop us a note if you have questions or ideas for optimization – we love hearing from you!