If you have attended any industry events or perused media news sites in the past few months you’ll be well aware that we are no longer in the decade long ‘Year of the Mobile’ and have well and truly entered the ‘Year of Artificial Intelligence (AI)’.
Alongside Artificial Intelligence, Machine Learning also features as a buzz term of the moment. Whilst these terms are perceived to be synonymous, they are in fact quite different things and understanding the distinction is important to understanding both how they can be used and the impact they will have.
To put it simply, AI is a computer science which strives to create machines that simulate intelligent human behaviour – making computers actually thinklike a human (albeit a very smart one). An entirely autonomous and reliable self-driving car is a prime example of AI.
Machine Learning is a subset of AI, forming part of the wider AI toolkit. Machine Learning is defined as a process by which a machine can consume data and learn from it without any human intervention. Such machines use probability as well as past performance and real time learnings to recognise patterns in data and make predictions. The feedback loop of the data telling the machines whether their predictions are true or false will then impact their future decisions; thus making it a process of learning. Last year my bank knew that I had lost my credit card before I did as their machine learning tools had detected and notified them of unusual purchase patterns on my account.
Although distinct types of technology, AI and Machine Learning are interdependent – Machine Learning is born from AI, but AI needs Machine Learning to make machines truly intelligent.
From Google to Microsoft, automated bidding to customer service chat bots, Machine Learning is pervasive and whether you are a consumer or a marketer, and whether you know it or not, you are being affected by it in some way.
The power and impact of AI and Machine Learning on day to day life is undeniable and the media industry is no exception.
October 2016 marked a pivotal point for AI within the media industry. Cosabella, the US based lingerie brand made the decision to cease the “time-consuming and difficult” task of trying to communicate their brand to a digital agency of “very lovely people”. Instead, they passed the task of reaching and converting consumers to a single employee, Albert.
Albert is a highly efficient executive, planner and manager who can optimise any number of campaigns at one time (like clockwork), he is in the office 24/7 and quite literally, lives for data. Now recruiters, don’t get too excited, Albert isn’t a prospect for your books, he is an AI platform created by Adgorithms. Albert isn’t cheap, with a salary equating to 18% of Cosabella’s media spend but according to Cosabella, he’s more than worth it. With the help of Albert, Cosabella has seen a cool 50% increase on search and social Return on Advertising Spend alongside a 12% reduction in ad spend. The official line from the marketing director at Cosabella, Courtney Connell was “After seeing Albert handle our paid search and social media marketing, I would never have a human do this again”.
There are two clear winners in this scenario: Cosabella and Adgorithms. Both have reaped the rewards of being innovators – one by adopting the technology and one by creating it. The key lesson here for both brands and agencies is to embrace AI and incorporate it into your business to derive the maximum value from your marketing investment.
Fortunately for Dentsu Aegis Network, instead of competing with Albert, we have created an Albert of our own. Built by our Data Science team in India, it consists of a number of proprietary machine learning tools, one of those being the Intelligent Automation tool. The suite of available tools range from initial audience and budget planning, to producing new creatives based on live performance, and real time targeting, bidding and scheduling in line with your chosen objective.
iProspect Australia ran the first APAC test of these tools in April and here’s how it went…
Human was pitted against Machine with equal budgets, flight dates, creative and the primary objective of driving Return on Investment (ROI) for the client. The Human campaign was planned and implemented as normal – the planner employed their experience and historical performance. The Machine campaign used automated tools to create and segment three affinity audiences, and using the campaign objective and historical performance, it allocated budgets and built a bidding strategy.
Once live, the Human spent the next ten days monitoring performance, extracting and analysing data and making optimisations. This process was repeated a number of times over each working day. The Machine also spent the next ten days carrying out this process, however it was not multiple times per working day – it was every four hours without fail, day and night, Monday to Sunday.
Based on the data consumed by Machine and the subsequent learnings, the Machine made a range of optimisations from delivery to targeting:
· Performance of individual interests within an ad set were analysed; those weaker interests were removed and interests similar to the top performers were introduced
· Budgets were dynamically optimised between ad sets
· Age and gender targeting of individual ad sets were modified in line with performance
· Adverts were paused during lowest converting hours to optimise spend during most profitable time periods
Despite our Human making every effort to beat the Machine, she was simply no match for the level of data analysis and consistent, real-time optimisations performed by machine learning. The Machine produced an undeniable win with a 37.7% higher Return on Investment at a 20.9% lower Cost per Acquisition than that of our Human; achieving the best results to date for the client.
With the assurance that our tools can optimise more effectively than even our most intensive human campaign management, it’s a no-brainer that we should be utilising our home grown machine learning tools across all campaigns. With our Machine working to derive the maximum value from our live campaigns, our human employees can reinvest their time in more complex and strategic tasks for our clients.
Although machine learning is offering extremely impressive efficiencies in media planning, buying and even creative, the contextual and emotional understanding of a human is still vital for brands. An example being that accessible AI would not think to pause a promotion for an event in an area that has just experienced a destructive earthquake, as it stands.
To avoid the fate of Albert’s agency predecessor, it’s crucial that agencies work with AI and Machine Learning to ensure they are generating competitive performance results for their clients. Likewise, brands need to not shy away from these tech solutions but embrace them or lose out to their more adaptive competitors.