Scary clown in woods_Crop

OPINION20 March 2019

Clowns scare me too

Data analytics Opinion Technology Trends

Ryan Howard explores how deep learning algorithms offer us a peek into our unconscious.

It was an online video ad – the kind you have to watch before watching the thing you want to watch. Clowns knocking about, slapstick, white gloves aloft, fiery hair, wild eyes and silly enough that I remembered seeing it before. On its fated return to my laptop screen, weeks later, the sound was muted. Out the corner of my eye, a chilling series of bizarre, macabre scenes and a reminder that something, without context, is a wholly different proposition.

This is true of a lot of good advertising which draws upon higher order devices such as humour, irony, misdirection and reasoning. These are the necessary ingredients to draw us in, entertain and persuade us, if only to prolong our attention. These same devices however, are out of the reach of our unconscious. Our fight-or-flight unconscious does not intellectualise or evaluate information in the same way our analytical minds do.

Equipped with a different set of evolutionary filters, the unconscious is poor at interpretation; it deconstructs and grasps at momentary associations. You may testify to this if you’ve had too much cheese with your bedtime telly. It never sleeps, loves repetition and with a hunger for patterns, takes it all in. The same is true of computers, and it is down this rabbit hole where data science takes us next.

When Google shared its own deep learning framework, it triggered the largest wave of commercial innovation we’ve seen in a very long time. This darling of Silicon Valley is soon to be everywhere and in everything. Twenty years in the making, it is the bleeding edge, wrapped in a big beautiful bow. This year, one would be hard pressed to find a marketing sciences team not investing ambitiously. Make no mistake, the race is on.

Deep learning is a machine learning algorithm arranged into many layers (hence ‘deep’). Each layer generates its own representation of the data it receives before firing it into the next. It learns when its most recent representation is either rewarded or punished. Connecting equations behave like real neurons and together learn like an animal brain – this is slightly more than an obliging metaphor. In this way, deep learning works intuitively with audio, image, and text problems, allowing machines to gauge, relate and categorise a world of stimuli. If you fancy seeing it in action, visit here.

Oh yes, this is starting to sound very much like the future of market research. Science fiction fans will have already accepted the inevitability of the decisive war between cyborgs and humans. But before these dark days arrive, deep learning will gently transform this industry unrecognisable.

In theory, it is now possible to coax out the linkages between the signs and signals of brands, campaigns and messaging with traditional measures like recall, engagement and ROI. We’re already at the stage where collateral can be tested and quantified. It is a counting exercise, creative by creative, video frame by video frame. Even this, the driest of measures, raises extraordinary questions as it reveals both calculated and accidental strategies within competitive sets.

For example, we can compare brands that typically portray a lone customer interacting with a product against those that show families or friendships. Who showcases their product indoors versus who hugs neon cityscapes? Which brands are text heavy? Younger or older? Harder or softer? Smooth jazz or punchy electronic? Emotionally charged words or neutral, matter of fact language? Moreover, which cues are returned to, time and time again, and how do they play out across the course of a 30-second ad? Down this rabbit hole we race.

While a part of us may muse over the deliberate narrative of a campaign, computers are ready to track the stream of cues and associations that, through repetition, are registered by the deepest parts of our psyche. This is our creative unconscious, which is continually and automatically priming the decision sets, upon which we deliberate and ultimately act. If you were to make peace with these Orwellian ramifications, then you would forgive me for never again venturing near a circus, panto or my nephew’s birthday party – perhaps forever haunted by flashbacks of gurning clowns never intended to be seen without silly music.

Ryan Howard is director advanced analytics at Simpson Carpenter