In the increasingly competitive world of display advertising, the technology used by first-generation ad networks is no longer enough to optimize campaigns. More sophisticated buyers now participate in auction markets using new technologies which enable radical price-performance advantages. These new buyers take advantage of inefficiencies in the competitive landscape by continuously adjusting their bids based on different predicted values of each individual impression. How can you keep up with all the new tricks and combine them to achieve the best results? One way to achieve superior online advertising results is through artificial intelligence (AI).
Time for a change
For years, online display advertising companies have been able to take advantage of low hanging fruit. Ad networks have easily delivered impressive results through relatively simple and one-off techniques like retargeting or content targeting. By blending these optimization techniques with other "filler" tactics such as broad geographic targeting, ad networks can meet customer performance objectives and maintain high margins.
When reviewing the actual price bid by our competitors in the real-time exchanges, tricks such as manual bidding and buying become clear. Very frequently, we see the same rounded prices coming up: $1, $1.50, $2, $2.50 CPM. These prices reflect a strikingly unsophisticated buyer moving through the marketplace in a ham-handed way and making optimization adjustments on a gross scale.
How AI is changing real-time bidding
The field of AI has a storied history. Since 1960, researchers have made steady progress at creating software that can perceive patterns in data better than humans can. The idea is simple enough: A person scheduling ads can look into the interactions of one or two variables at a time whereas smart software can simultaneously consider thousands of variables. A machine learning algorithm learns from all the attributes of each impression, treating each impression as data that contributes evidence toward thousands of simultaneous hypotheses. Instead of focusing on either behavioral or contextual targeting alone, machine learning predicts the value of an impression using all of these techniques together.
Equally important with machine learning are smart control systems. Sophisticated ad buying software excels by adjusting bids automatically in response to the changing competitive landscape. For example, rather than achieve even pacing throughout the day by buying an equal number of impressions each hour, smarter modern ad buying software adjusts pacing based on the availability of traffic, as well as its predicted conversion rate. This sophisticated software will vary bids based on the predicted probability of success (a click or conversion) instead of using the same bid for any impression that matches its hardcoded criteria. In the new world of real time bidding for exchange inventory, opportunities for smarter ad buying become even more viable because the software enables both cherry-picking the best impressions and the ability to stop campaigns on a dime.
Machine learning models also enable ad buying decisions to be extremely personalized. For example, the optimal frequency cap per day, or for a whole campaign, may not be a fixed value for every campaign or for every user. By including frequency along with other attributes in the model, an intelligent network can determine the exact value of showing an incremental impression to a given user, on a given page, at a given point in time.
What nourishment does this software need? Artificial intelligence software is hungry for data. Billions of impressions, each of which can be coupled with prior user history that may be relevant to ad serving decisions, flow through the open mouth of large ad network databases each month. Taking advantage of this much data necessitates an experienced software development team to design and operate the large scale data processing clusters which manage it all. The impact of it all, though, is significant for users. Ads are increasingly relevant and can be uncannily accurate in presenting an appealing and timely call to action.
Reaching consumers with relevant and targeted ads is no longer a job for companies with a single trick. Ad networks are now emerging with a full bag of tricks to ready to capitalize on all that AI, smart controls, and large scale systems bring to table. The future of online advertising means using "artificial intelligence" to drive real results.