When the hype around programmatic display advertising peaked early this decade, journalists and bloggers hit back with lots of articles about how marketing will never be an exact science.
Digiday’s Brian Morrissey wrote: “The ads, we’ve been told for years, would become so relevant that they’d be ‘like content’. I don’t know any normal person unaffiliated with this industry who would believe this to be the case.”
‘Ad Contrarian’ blogger Bob Hoffmann chipped in: “We certainly have more precise tests, and more precise data, and more precise measurements. We use the language, the methods, and the tools of science. But have we made advertising work better? Not that I can see.”
And perhaps most eloquently, Julia Kirby wrote in Harvard Business Review: “It would be easy to conclude that advertising has flipped to all science and no art. But then along comes fresh creative to show us what really sells.”
There have been many direct response campaigns that have been thoughtful (think The Economist or O2) and let’s not mention Trump’s election campaign, but the creatives were right – targeting doesn’t mean much if you’re serving people crap. Users will tell you they don’t give a stuff about display advertising, whatever format it takes or whatever jargon we drape it in.
Fast forward to around 2015 and machine learning technology for marketers began to spring up at the same rate ad tech firms had five years before.
But now the reaction from the press seems to me to be subtly different. Though the broader significance of artificial intelligence is many orders of magnitude greater than that of programmatic display ads, the creatives seem suspiciously quiet.
It’s as if the complexity of big data analysis and machine learning (which puts even demand-side platforms in the shade) is enough to subdue many marketers who are perhaps starting to think they have had their chips.
The debate about art versus science (or art and science) needs to be reignited. We need to once again champion the power of the human creator and curator.
Machine learning for personalisation is not a panacea
Machine learning powers some aspects of merchandising on big pure-play ecommerce websites like Amazon, and increasingly on smaller websites thanks to vendors such as Alfred, which is working with lingerie company Cosabella. However, the websites that impress (and persuade) me are still those created by strong brands with clever curators – think Lush.com (which I wrote about in Marketing Week recently), Topshop and Ikea.
Machine-learning personalisation may well drive an increase in conversions, but a website’s ability to drive lifetime loyalty will take longer to measure, and offline impact may never be satisfactorily attributed.
Marketers should make their websites useful and inspirational destinations, not simply well-honed sales machines. Combining the two is, of course, the ideal.
Just like the programmatic scenario, marketers can tweak what website content users see to a seemingly infinitesimal degree, but unless that content and the underlying product and brand are compelling, it matters naught.
Brand specialists are becoming more important than ever, simply because media has proliferated. The need for consistency is growing.
Airbnb is a good example of a fine balance. The website is brilliant at using algorithms to give pricing tips to hosts and surfacing recommended listings to guests, but the consumer is likely to be most swayed by photographs, a host’s description, glowing guest reviews and Airbnb’s impressive branding and advertising – all conspicuously human and difficult for an algorithm to interpret.
As Airbnb seeks to expand the services it offers, its experience listings, too, feel much-loved and hand-curated.
I’m not underplaying the usefulness of machine learning – to name just two applications, image search is massive, as is predictive analytics. However, I think it behoves marketers to get even better at the human side of the craft.
Marketers need to integrate this tech sympathetically and with great user experience if it is to succeed. It should be the rocket fuel for great creative.
Machine learning in the context of the experience economy
Jim Sterne, author of Artificial Intelligence for Marketing: Practical Applications, points out there’s a difference between shopping and buying: “I don’t go shopping on Amazon, that’s a terrible experience. I go to Oxford Street. But when I want to buy something – stores on Oxford Street don’t have it in the right size, or it’s not in stock. If I go to Amazon and I know what I want, Amazon says you’ll have it tomorrow.”
The point is that though Amazon is incredibly adept at encouraging purchase and providing convenience, creating an environment that consumers go to for inspiration is a very different kettle of fish.
And even if in the future we may direct all our wants through AI, Sterne says it “listens to my opinion, formed by advertising and marketing, and that makes the creative stuff even more important”.
“Marketing makes my wife want a particular soap,” he continues, “then Amazon gives me a subscription and that convenience is valuable to me.”
Machine learning should be a guiding hand for creativity, focusing the efforts of marketers on the areas where they can best make an impact.
Ben Davis is editor at Marketing Week’s sister title Econsultancy.