Machine learning: ‘A game changer, but not a silver bullet’

Machine learning technology will lead to one of the marketing industry’s next biggest transformations – but brands and agencies should not over-value its potential, a new report has cautioned.

According to a global study by iProspect, predictive algorithms have become a high priority for marketers, with less than two per cent surveyed claiming it will have minimal or no impact on their business this year.

However, while more than half of marketers agreed the technology would help them decipher bigger data sets and deliver personalised content at scale, iProspect’s report cautioned them against using machine learning as a “silver bullet” for other industry problems.

For example, if the data submitted is bad or of poor quality, it is effectively useless, the study claimed.

“Two aspects are fundamental for an effective learning process, whether the student is a human or a machine: valid initial input and time,” it said.

“Algorithms require valid input and are not yet capable of evaluating the intrinsic quality of an informational entry. In other words, algorithms are only be as good as the data they are fed. Poor quality input will lead to wrong conclusions, which could turn into wrong decisions.”

In addition, marketers who have pinned their hopes on machine learning solving problems such as ad fraud and viewability may be disappointed. In fact, human input and insight will remain crucial to solving these problems, according to the study.

“The idea that ad fraud will be easier to prevent, and ad visibility will be more reliably measured and reported is not wholly true,” it said.

“We will still need to actively exclude data from uncorrelated or suspicious sources such as bots and unexpected consumer behaviours. Only humans will be able to navigate and address these individual nuances to ensure clear campaign performance.”

Nevertheless, the technology will allow decision-makers to have a “deep and well-informed insight” into their overall communications strategy, which may help aid the reporting of fraud and ad impressions.

Meanwhile, 48 per cent of those surveyed said they believe machine learning will allow them to automate manual tasks – such as disseminating data sets – to help them focus on longer-term goals.

However, the report argued that marketers should see machine learning less as an automation tool and more about the “marketer’s ability to solve previously unsolvable problems using technology”.


Get the latest media and marketing industry news (and views) direct to your inbox.

Sign up to the free Mumbrella Asia newsletter now.



Sign up to our free daily update to get the latest in media and marketing