Features

Big Data’s Killer App: Mobile RTB

Written on
Jun 18, 2013 
Author
Dr. Saed Sayad  |

ADOTAS – As you may have noticed, mobile real-time bidding (RTB) has gained popularity recently.  Mobile RTB ad buys were not particularly common as recently as last year, but in just the past 12 months we have seen the market explode. In Q1 2012, mobile RTB ads accounted for only 8% of ads served on mobile devices, but by Q4 2012, the RTB market accounted for 64% of total mobile ad requests.

The shift to mobile RTB is well deserved, due to its capability to harness the limitless streams of data that let advertisers intelligently target and serve their ads to consumers. There are many outstanding benefits to mobile RTB for both advertisers and publishers, but in a nutshell:

  • Mobile RTB allows buyers and agencies to accurately target their audience, control their campaigns and give complete transparency, yielding increased return on advertising spend.
  • Mobile RTB gives ad sellers and publishers greater yield optimization, a higher value of inventory, and revenue for display ads sold outside of their direct relationships.

Moreover, big brands need comfort that they can handle their big data needs across all audiences, but ROI remains the gaping hole in mobile advertising. In fact, RTB is the killer app for big data, yet to date the missing link has been the ability to use predictive modeling and targeting to harness the oceans of available information. Effectively using RTB depends on how intelligently and efficiently you use big data – this requires a technology that is capable of processing massive amounts of data with great speed, accuracy and — if possible — in real time.

First Things First: What is Mobile RTB?

Mobile RTB can be broken out into 4 basic components:

1) Bid Requesting Inventory: Advertising exchanges enable their inventory to be bid on by networks, and they provide information with the bid request. This can include the publisher’s name or URL, the ad’s position on the page, the type of ad unit acceptable, latitude and longitude, to name just a few

2) Bid Responding System: The system’s sole job is to ‘answer’ back to the bid request from the ad exchange with a bid price and the creative to be placed.  The bid response first digests the information coming in with the request, then obtains and digests any 3rd-party data available.  This must be done within 100-200 milliseconds before the publisher chooses the highest bidder and places the ad.

3) Data Enrichment: he impression request is enriched by using 3rd-party data partners and through building your own database of 1st-party information. Both are important.

4) Decision Engine: This is what makes or breaks mobile RTB platforms, and what the rest of this article will focus on.  Decision engines are based on predictive models – essentially, they aim to predict future actions based on historical data.

Putting Mobile RTB to Work

Unlike first generation mobile ad buying platforms, the second-generation mobile RTB ad networks enable advertisers to buy inventory through ad exchange platforms on an impression-by-impression basis. These impressions are bought via an RTB platform, and chosen based on predictive models that tell the platform whether or not the impression fits the advertiser’s desired criteria.

A critical variable in the effectiveness of a predictive model within the context of an RTB platform is both the speed and accuracy with which the model can process massive amounts of data. These elements are the key differentiators of a truly real time predictive modeling platform; as such a platform is confronted by constantly expanding data.

Real Time Predictive Targeting: Pushing the RTB Barometer

What do we mean by “real time predictive modeling”, and why the need for progress?  Recently, we have seen predictive modeling in RTB moving well beyond static activities, such as segmenting and simple retargeting campaigns, toward predictive targeting and real-time decisions. The term “real time” is used to describe how well a predictive modeling system can accommodate an ever-increasing data load instantaneously.

Traditional predictive modeling – a process that can take days, weeks, or even months, in the era of big data is simply not working. The need for advancements in predictive modeling with the capability to learn and predict in real time is getting more obvious as companies look to define the ROI on their campaigns.  Such real time problems are usually closely coupled with the fact that conventional predictive modeling engines operate in a batch mode, which requires the model to have all relevant data at once, and it doesn’t react to new data in a dynamic manner.

Now We’re Getting Somewhere

In our experience, real time predictive modeling showed an unparalleled ability to cherry pick desired brand target audiences and impressions. The participating advertisers have enjoyed improved engagement levels, such as significant uplift in click through rates (ranging from 100% to 1000% increases in industry averages) and awareness, and considerable reduction in cost per acquisition rates (decreases in CPA rates up to 70% ).

As advertisers benefit from these results, it should increase industry standards to push the barriers on networks offering traditional predictive modeling.  Mobile RTB exists to provide advertisers with an efficient way to leverage this technology and reach their desired audiences, one impression at a time.





Dr. Saed Sayad, Chief Data Scientist of AdTheorent

Saed has more than 20 years of experience in data mining, statistics and artificial intelligence and designed, developed and deployed many business and scientific applications of predictive modeling. Saed is a pioneer researcher in real time data mining and big data analysis, an adjunct Professor at the University of Toronto, and has been presenting a popular graduate data mining course since 2001.

Reader Comments.

No comments yet

Leave a Comment

Add a comment

Tags: , , , and
Article Sponsor

More Features