Industry figures share their views on the latest issues. If you have an idea for a guest column, email firstname.lastname@example.org
Ask not what machines can do for you, but what you can do for machines
AI is emerging as a must-have tool for marketers of all kinds, but can brands and advertisers really rely on technology alone to solve their problems? Not quite, says MiQ South East Asia managing director Marcus D’Souza.
The rise of AI provides the modern marketer with the kind of data-based efficiencies that only existed in their wildest dreams a decade ago. And with the right know-how, these new technologies are transforming the way information is interpreted and unlocking untold consumer value that had previously been unobtainable.
By way of explanation, today’s AI in a marketing and advertising paradigm is mostly defined by machine learning. That is, utilising algorithms to sift through massive data sets to identify patterns, predict potential high-value clients or revenue streams, and bid on the inventory - and at a vastly quicker rate than any human could accomplish the same task.
By repeating this process, AI programs can not only churn through the avalanche of data marketers are faced with, but improve incrementally as time goes by. This is where machine learning thrives.
But AI technology on its own is not a marketing solution, no matter how seductive the idea of leaving the heavy lifting to a machine or algorithm may seem. In truth, AI has a series of drawbacks for marketers to contend with.
Where it doesn’t thrive is when human intuition is called for. By its very nature, machine learning is purely a mathematical solution and doesn’t understand the data beyond the need to sort it.
In other words, AI is only as good as the tasks it has learned, and at its worst, can provide a customer experience that is potentially infuriating, or a brand safety liability.
Scenarios like this invariably lead to a business facing a PR disaster which requires urgent addressing. Similarly, an AI glitch that affects the customer’s pathway to purchase could result in a dramatic drop in its operating revenue. In these instances, machine learning is powerless to assist and only the intervention of human reasoning can avert a crisis.
AI is also a reactive technology, meaning it will use past experiences that it has encountered to predict future outcomes. And as any hedge fund manager will tell you, this is fraught with danger.
And while there are no issues of latent malignance - the stuff of Terminator movies - in AI technologies today, it does unquestionably have no morality. By this I mean the lack of conscience can cause machine learning programs to cut corners to get to the desired result. This means that if a single goal is a paramount objective for an algorithm it might shortcut past other outcomes that are relevant, meaning the full picture isn’t presented.
Commonly in a marketing context, this is seen with businesses’ use of chatbots. As a consumer, you’ve probably experienced it with bots that have an apparent one-track mind, re-routing to a core response as soon as things get a little complicated.
Lastly and perhaps most pertinently, there’s the issue of making the processes intelligible. It’s all very well for a marketer to be presented with the results of a data set analysis, but understanding why a machine drew the conclusions it did can be like snorkelling through muddy water.
Without the requisite level of expertise in AI programs, it’s entirely plausible that the recipient of the results ends up with the conclusions sans the methodology. This is not a good thing.
Once again, the counterweight to these challenges is the application of human knowledge and experience.
And the first step to achieving positive business outcomes (and getting the best from your AI programs) is to ask the right questions. This means identifying the challenges that need to be solved, establishing the best ways to measure them, and discovering the metrics that will predict success for outcomes that are difficult to measure.
When these considerations are combined with a brilliant human insight, the results can be truly impressive in a marketing sense.
Take the Snickers Hungerithm for example. This ingenious marketing campaign applied an algorithm to the general level of anger online, dropping the prices of the chocolate bars as the metaphorical temperature rose.
In tandem with Snickers’ “You’re not you when you’re hungry” tagline, the brilliance of the insight saw sales spike by 67% and social traffic rise 1740% over the five-week campaign period.
Another memorable example was Bonds’ ‘The Boys’ campaign, which used real-time meteorological data and outdoor advertising to adjust the height of two men in hanging chairs as the weather warmed or cooled.
Combining a humorous take on underwear marketing, it again showed that combining a clever idea with rich data could yield big results. In this case, generating over six million video views and lifting Bonds underwear sales by 161%.
One last factor to consider, circling back to the issue of AI comprehension, is how you make sense of your data in the first place. After all, before brands can create marketing magic - like the examples above - they need to have refined data that provides a field of reference.
To this end, it pays to have data scientists and analysts who can interrogate the methodology and understand the results. Their expertise is the literal intersection of human capability and machine learning.
So, can AI help solve a marketer’s problems? Absolutely, but only with the aid of human intervention. And in the quest to achieve marketing alchemy, you can’t have one without the other.
Marcus D’Souza is South East Asia managing director of MiQ.
Have your say
Do you have a strong opinion on a topical industry issue? To submit a comment piece, please send a short summary of your idea to email@example.com. Views of writers are not necessarily those of The Drum.