In a recent poll of marketing professionals, iProspect found that although 79% of people know the difference between Artificial Intelligence (A.I.) and Machine Learning (M.L.), less than 25% are actively using these technologies in their current plans. This is despite 61% wanting to apply them to their 2018 marketing strategy.
To address the research and outline exactly how performance marketers can make the most of the opportunity today, iProspect experts gathered on 29 November for our latest Breakfast Briefing on A.I. and Machine Learning. Alongside Google and Eurostar, we explored the possibilities, put the principles into practice, and unveiled our proprietary machine learning tool, iProspect CORE.
As Stefan Bardega, iProspect CEO UK & Ireland, who opened the event explained, our role as performance marketers is to generate transformative business outcomes using the most advanced techniques and disciplines. A.I. and M.L. are two of the most powerful developments to impact performance marketing in recent years so it is imperative to be prepared.
To demonstrate how Machine Learning can be put into practice, Sophie Wooller, Director of Data & Technology Products, and Josh Carty, Data Scientist, used crowd-sourced decision-making to predict the fate of passengers on the Titanic. Tapping into the principles of Machine Learning, the audience worked together to plug different criteria into a custom-built algorithm. With each iteration, the model’s accuracy improved, bringing to life the learning element of M.L. and highlighting the need for human involvement.
iProspect’s Director of Performance Media Products, Caroline Reynolds explained how effective CORE had been for high-speed travel service Eurostar. A Machine Learning-powered optimisation engine that automatically analyses, activates, and optimises media based on our bespoke algorithms, CORE processes data at a rate of 6bn rows per minute. With speed, efficiency and effectiveness at the centre, CORE delivers 2x the impact of performance media with unprecedented speed and accuracy.
Taking a deeper look at its infrastructure, Caroline overviewed the tool’s four stages: 1) Data Collection, 2) Data Storage, 3) Planning & Prediction, 4) Activation. It’s at the Planning & Prediction phase that Machine Learning comes into its own as algorithms evaluate decisions and adapt to create the optimal activation plan for performance media. This simultaneously activates CORE’s learning system through platforms such as Google Adwords, Bing, Facebook Business Manager, and Google DoubleClick Bid Manager to optimise innovative strategies and implement changes automatically in real time.
In only five weeks, CORE yielded impressive results: a 12% reduction in Cost Per Click, a 26% increase in conversion rate and, crucially for Eurostar, a 30% improvement in Cost Per Acquisition. Eurostar’s Head of Marketing & Brand, Guillemette Jacob, said “iProspect are key to driving our digital transformation journey. CORE is the latest example of how the agency collaborates to provide innovative solutions that address both our business and digital needs.”
Rahul Parmar from Google’s Cloud Platform then joined iProspect’s Head of Global PPC for a ‘fireside’ chat discussing the latest innovations in A.I. and M.L. He reflected on the democratisation of the technology, explaining that we are now seeing it become more accessible. While much of it remains high-end and requires skilled developers to build custom platforms, in the short-term there are APIs and products that enable brands to leverage M.L. quickly and effectively.
Over the past few years, Google has been using A.I. to tighten up the feedback loop in G Suite to implement quick and effective channel adjustments, and automatically identify the biggest cross-selling and upselling opportunities within Soothsayer. On the optimisation side, Deep Mind uses algorithms to optimise Power Usage Effectiveness and improve efficiencies in data centres.
Jack Swayne, Managing Director of London and Stafford, closed with the three key takeaways for brands to prepare for the Machine Learning opportunity.
Define the challenge: clearly articulate the business challenge, start with a number of different hypotheses, and develop a Machine Learning research roadmap now
Understand your data: audit the data you currently have and that you can access, understand your data governance, warehousing and QA capabilities, and start smaller analysis projects as soon as possible
Get the technology and expertise you need: consult with experts early, and keep talking. Explore data management and analysis platforms to find the right solution for your business, and create training plans to develop your internal knowledge and skills around M.L.
Find out what some of our speakers thought were the key takeaways from the event in the video below: