Why online advertising needs data scientists and Artificial Intelligence
03 Oct, 2012
Better understanding of Big Data will help advertisers define their audiences and find new users.
Ad agencies increasingly face the challenge that while the online advertising industry is now very well equipped to target current audiences, it’s poorly equipped to identify new audiences.
Online advertisers should look to new technologies to target potential customers and prospect for new users.
Artificial intelligence (AI), computer modelling and advanced mathematics will automate many online advertising processes and decision-making by mirroring and enhancing human expertise and experience, enabling agencies to both find and target new audiences.
The adoption of trading desks, demand-side platforms (DSPs) and real-time bidding (RTB) has fundamentally changed the online advertising industry in recent months, giving more agencies the opportunity to run online ad campaigns without intermediaries like ad networks.
Now that agencies have experimented with these new tools they are able to consider the benefits and pitfalls of going it alone, and some interesting outcomes are emerging.
The general consensus seems to be that no single solution improves performance across the board and that there is nothing unique that allows agencies to separate their delivery methods from one another. But some technologies are effective on some campaigns, sometimes.
Retargeting is straightforward, and using first party data in a DSP or trading desk can be effective for re-messaging users who have already visited a site – but it’s necessarily limited to the finite amount of current site users.
The difficult bit is to continually build brand awareness and prospect for new customers, to reach the many thousands of potential customers who have never visited a client’s site. Third party data purchased from data suppliers can help, but advertisers often end up using the same data as their competitors.
This pushes the costs of data up, and means that agencies cannot reach unique audiences as they are only accessing aggregated inventory available on exchanges.
A data management platform using AI and computer modelling techniques will help advertisers define their own audiences and find new users.
The most cost effective way to improve performance is through using unique data and unique strategies that set advertisers apart from their competitors. This is why, since 2009, Crimtan has steadily invested in developing its own expertise in data profiling and in building its own data management platform and ad serving technology.
It’s also why the company reached out to the University of Brighton’s School of Engineering, Computing & Mathematics; to help understand how AI, computer modelling and advanced mathematics could improve advertising effectiveness and deliver advertisers better targeting, optimisation and performance for their online advertising.
Advanced mathematics has helped the internet evolve into what it is today, but there’s still a great deal of room for improvement. Key to this is the ability to collect, store and manage vast amounts of data – including behavioural data – so that the maximum value can be extracted and used within a sophisticated, large-scale, real-time web delivery environment.
But managing ‘Big Data’ needs a combination of the right technology and the right people – experienced data scientists who understand online advertising and can mine the data effectively to develop algorithms that address multiple market sectors.
At a basic level, these AI advertising technologies will define logical rules that underlie both predictable and surprising behaviour of internet users. This will result in radically more intelligent and vastly accelerated decisions, such as which ad should be served to a particular user, in which format and when – vastly improving campaign effectiveness for innovative advertisers, while enriching the online experience for internet users.
A key part of Crimtan’s work with the University of Brighton will be investigating how negative reasoning techniques can inform AI technologies and optimise real-time targeting systems.
At present, online ad campaign decision-making is primarily governed by user actions, such as clicks, online purchases, engagements and the like. To date, the value of inactions has not been effectively incorporated into online ad systems. Using data scientists’ negative reasoning techniques will help AI technologies to define decision-making rules based on inactions, non-occurrences and non-events.
Put simply, this relates to data regarding users not clicking on an ad, not purchasing after entering an online store, or not hovering over an expandable rich media advert. The resultant new systems will limit wasted ad impressions, giving improved return on investment.
Data driven advertising has a lot to offer, and agencies that are willing to work closer with partner networks to seek out the latest innovations and develop new data strategies will be the ones that gain a competitive edge.
Real-time technologies using AI, advanced mathematics and predictive modelling are next-generation tools that will bring advertisers more efficient brand building and prospecting, along with greater customer insights and more long-term strategic value.
By Tony Evans, Corporate Development Director, Crimtan.
This article first appeared on www.brandrepublic.com on 24 September 2012