Optimize Ad Buying with AI-powered Recommendations

You’re probably hearing a lot about AI in advertising lately and may not be entirely sure what that means. We get that most users don’t want to hand their keys over to a black box and want to make informed decisions instead. So, we’re going to break down how AI should be used in an ad platform to make it more tangible, starting with campaign optimization recommendations.

A big part of Yahoo Blueprint and the DSP Dashboard is our AI-based recommendations, which recommend campaign adjustments to maximize delivery and expand the success of the best performing lines. 

First, AI is not magic! It actually means that we have computational methods of analyzing large amounts of data, processing them and generating information that’s usable. So, our recommendations are not popping out of thin air, but instead we parse all of our running lines through machine learning models, such as neural networks, to determine how tweaking various parameters might optimize delivery. 

This is only possible because we analyze over a million data signals from the daily processing of more than two trillion bid requests to calibrate those models. These data signals include:

• Past line delivery
• Forecasted supply (the same as available through Omniscope!)
• Optimization modeled performance
  

These data signals analyze past delivery metrics and compare them with the predicted delivery based on our forecast. Then, we can extrapolate how different optimization strategies would perform in order to provide sound data-enhanced recommendations.

The advantage of utilizing these different data points is that our recommendations aren’t just simply telling you to do something. By leveraging our forecast, for example, we’re able to provide clear rationale for our guidance and show you the estimated impact that such a change will have!

In the end we’re providing a few different recommendations:

• Incremental Budget
  For CPA and CPC lines that are reaching their budgets and staying under their goals, we estimate how much budget could be added to the line to push performance with minimal impact to their CPA or CPC goals.

• Max CPM
  For the same CPA / CPC lines, this recommendation shows how changes to the maximum CPM can help fully spend the budget or increase their KPI performance.

• Viewability
  If viewability targeting is leading to under delivery of impressions, this recommendation will show the effects of adjusting the viewability target.

And we’re not done. This space is iterating quickly across our industry and platforms in particular. Yahoo DSP already has more recommendations and insights being cooked up on the Dashboard, leading to more options and insights for buyers. Make sure to be on the lookout for more!