Media Planning & Buying
How to overcome the uncertainty of programmatic buying
July 08, 2014

Programmatic buying is like fortune telling: It's based on predictions that might not be accurate. Here's how to use the unknown to your advantage.

Programmatic advertising was a natural evolution of the digital landscape. On the one hand, buyers need to target at scale, while on the other, sellers want to monetize remnant inventory. Technology is the glue that makes it possible. We saw it happening on Wall Street years ago. Now, we are living through its growing pains and learning curve in digital advertising.
 
Buying ads programmatically is in a sense a lot like buying wholesale inventory -- there is no guarantee for each item to be perfect. You can inspect a sample and count on it being a good representation of the quality of inventory. You realize that you can't inspect each and every item and allow for some items to be imperfect. Inspecting every item is simply too costly and impractical and will not allow scale. As long as the average outcome (in this case, percent of defective items) is predictable, you can determine the price of the wholesale purchase that is sensible to you.
 
When you buy programmatically, you have to give up control of every single impression in order to scale. Your targeting will not be 100 percent accurate with every impression. However, if someone has already measured the imperfection rate of thousands and millions of identical impressions, you can still target a certain level of performance and count on it reaching the goals of your campaign without limiting scale. With the help of technology, you have the ability to target and inspect inventory -- similar to when you buy wholesale inventory -- and predict the overall campaign performance.
 
Ad viewability is one of the things that advertisers are focused on these days, and for a good reason. Ads that are not in view don't have the opportunity to influence the users' decision to buy a product or make them remember a brand. Technology allows buyers to bid on a specific likelihood of in-view impressions. But can anyone guarantee viewability of any ad before it is served?
 
Viewability depends on the user's interaction with content and can only be measured after the ad is served. If the specific user never scrolls down, none of the below the fold ads will be in-view. But if the user does scroll down and spends time below the fold, the outcome will be very different. Programmatic buying by definition is about bidding on inventory before the ad is actually delivered and rendered, and therefore there is no way to know if an ad will be in view for sure. Regardless of how you look at it, it is all about knowing the probability and ability to leverage this knowledge.
 
Don't be fooled by guaranteed viewable impressions. It's an illusion -- a dangerous one. Your desire to accurately predict the future is tempting, but the only way to do it is by building a time machine.
 
Scale allows us to use probability to our advantage. With a large sample of impressions on a placement, we are able to predict viewability and help to price the bid accordingly by discounting for the rate of unviewable impressions. Just like buying wholesale, your overall ad buys allow you to achieve the results that you need at scale.

Kiril Tsemekhman is senior vice president and chief data scientist at Integral Ad Science.

On Twitter? Follow iMedia Connection at @iMediaTweet.

WHITE PAPER LIBRARY

View More Research »