AppNexus Programmable Bidder: How Data Science Gives Buyers Greater Control Over Bidding Strategy

Comments Off on AppNexus Programmable Bidder: How Data Science Gives Buyers Greater Control Over Bidding Strategy

Programmatic advertising has enabled marketers to target their audiences more efficiently and buy inventory they wouldn’t have access to otherwise. But digital advertisers haven’t had great options for implementing a bidding strategy that perfectly aligns with each campaign’s unique KPIs.

That’s the problem we set out to solve when we built the AppNexus Programmable Bidder (APB) two years ago. APB allows marketers to upload their own algorithms onto the AppNexus platform, change them at will, and see their bidding strategy change accordingly in real time.

In this post, we’re going to look at how data science makes APB such a valuable tool for marketers and give you a sneak peek at some of its upcoming improvements.

The problem: Bid strategy optimization is either too simplistic or too costly

 

As we mentioned, options have historically been limited for marketers who want to build out robust bidding optimization strategies. They can use an out-of-the-box DSP, which makes it quick and easy to set up a bidding strategy using one-size-fits-all algorithms. Many marketers find this sufficient, but for big brands and their agencies, it means losing the ability to apply proprietary data on customer behavior or likelihood to buy.

The other option is to build a bidder from scratch. While this would give marketers near-total control over bidding strategy, the cost is simply too high. Building a sophisticated bidder takes more time, money, and engineering resources than most advertisers can afford. So, most buyers had to bite the bullet and use a third-party solution, knowing it couldn’t get them the best possible results.

 

How does APB solve this issue? A healthy dose of data science

 

Bidding strategies are determined by several different algorithmic components. One of them, for example, is the expected value (EV) assigned to each impression based on the KPI the buyer has chosen for the campaign in question. So, if a buyer wants to optimize for clicks, the EV would be based on factors such as value of a click and likelihood of a click.

APB gives buyers the ability to granularly control the bidding process by uploading their own predictive algorithms that model these components (based on information in the bid request like ad size, domain, and user segment) onto the AppNexus platform using a simple API. Other components buyers can build their own models for include probability of conversion after a click, creative selection, and many more.

When we first released APB, we also released our own proprietary language for marketers to build their bidding algorithms and upload them onto APB: Bonsai. Bonsai is a decision-tree based language that lets buyers determine bid valuations with simple “if this, then that” logic. Every tree branch leads to a leaf value, which is the final figure to be used as a bid price or bid multiplier.

Our clients had a lot of success with Bonsai, but many wanted even more granular control. That’s why in the newest iteration of APB, we’re giving buyers the ability to upload logistic regression models, which can predict “yes or no” events like a click with greater accuracy across a wider array of scenarios.

We’re also giving buyers the ability to tap into the power of our optimization models at the same time they’re using their own. Appnexus has its own default models for how each of the bid valuation components we mentioned will change in different scenarios. With the new APB, buyers can now easily utilize those defaults while inputting their own models for other components. That way, buyers can capitalize on their strengths in areas where they have extremely useful data and let us handle the rest, leveraging data from the billions of transactions that take place on our platform every day.

 

How you can take advantage

 

APB is still in alpha testing with a limited group of buyer clients. But you can start preparing yourself now by investing in data science – consider making it part of your next hiring or training initiatives. It’s an investment you won’t regret, especially as machine learning continues to improve the way advertisers do business.

 

Want to learn more about our APB offering? Contact us to speak with one of our experts!

Filed under AppNexus Updates, Technology.

Comments Off on AppNexus Programmable Bidder: How Data Science Gives Buyers Greater Control Over Bidding Strategy

Comments are closed.