Harnessing the Artificial Intelligence frenzy

The world is in an Artificial Intelligence frenzy. On one side are the cheerful optimists like Mark Zuckerberg who believe AI will be a force for good in our future lives. On the other side there are the nay-saying warnings of Elon Musk about the possible threat AI poses to civilisation. The hot debate underscores the fact that AI is already with us in a big way. Not the Terminator-esque super-computer Skynet kind of AI, but rather weak AI (or Machine Learning) designed to specifically perform narrow tasks. Netflix is an example of how weak AI can support marketing. Netflix employs machine learning to recommend new content most effectively, continuously refining its suggestions based on what consumers have viewed in the past. The exponential growth in data, along with rapid developments in our ability to automate, collect and interpret this information, makes marketing analytics and AI natural allies. One of the benefits of this kinship could be more immediate marketing measurement. That’s an enticing prospect we here at MediaCom are exploring.

A new way to immediate answers on media return

For a quarter of a century MMM (marketing mix modelling) has been the gold standard evaluation methodology of marketing analytics, and it will remain so for the foreseeable future. Against its many benefits though is one key flaw – the time lag required to report on performance that can make it difficult to use in provoking quick behaviour change. To tackle this problem, we currently have a solution in beta that will give planners an early, accurate view of campaign success or failure. Appropriately enough it is codenamed Project Canary – after the famous canary down a mineshaft.

Our solution will combine historical MMM insight (controlling for underlying factors such as consumer confidence) with automated data sources (eg. Google Trends) to build an estimation of future media performance based on business leading indicators (eg. web visits). It will then be able to compare performance versus forecast in real-time, allowing us to infer campaign success or failure against expectations. What makes our solution industry-leading is the implementation of machine learning functionality. Usually, real time forecasts become less accurate over time as underlying factors and market dynamics change. In contrast, as our solution collects more data, it learns how movements in these different factors influence the performance of these leading indicators, thereby evolving the accuracy of its forecasting over time.

Supporting balanced decisions in real time

The flip side to having more immediate and accurate insight is making sure we don’t make knee-jerk decisions. Setting out upfront how you will evaluate performance against the objectives is a discipline that needs to be stuck to, given you can’t make a satisfactory or robust judgement on success on one day, or one week’s output. Our solution, complimentary to MMM, will enable a more informed and consistent approach to our planning that works outside of the main modelling debriefs. Looking further ahead, we also see an important integration with Mediacom’s Budget Allocation tool to provide one overarching view on spend recommendations. The Rise of the Robots is here already. It is certainly set to improve the way we measure the performance of marketing communications, but it’s not quite time to call Arnie – yet.

Next
Real World Insight
Previous
Audio Visual