I went in my first Amazon Fresh Store recently. If you haven’t done so yet, I heartily recommend it, as it is a fascinating experience. To gain entry, you swipe a QR code in your Amazon app, and once you are in, you put stuff in your bag and as the PR says, ‘just walk out’.
That’s it. A masterstroke in taking the pain out of physical shopping, no queues, no payments (at the time), and I expect many of the major supermarkets will swiftly follow suit. Naturally, I couldn’t help but wondering how it worked exactly. The official website lists the technology involved as a combination of cameras, sensors, RFID readers and deep machine learning.
The receipt then comes through to your inbox a few hours later. The timing of the receipt after you leave the shop is rather variable, and hints at the element that isn’t listed in the official list, which is of course the human element, which in technical terms can be considered as training the deep learning algorithms. I would imagine the deep learning algorithm produces a recommendation based on the activity of the store visitor, which is then reviewed and ratified by the human element, and slowly, over time, the guesses become more and more correct, and at some point, more accurate than the human, at which time the human is retired (not in the Blade Runner sense).
As an aside, when the stores came out, a surprising amount of (more or less) upstanding people of my acquaintance immediately started wondering whether it would be possible to trick the system by putting things back unexpectedly, bagging up quickly – as a thought experiment I’m sure, but it is amusing to me that even larceny can contain an element of gamification.
Machine learning and marketing automation
It got me thinking about the world of marketing automation and machine learning (ML). In our world, all the major players have a machine learning component, able to graft vast tracts of data to propose insight to the modern marketer. And that’s key to how ML can be well-used in the marketing world, not as a replacement for the human element, but as an adjunct. Used properly, ML in marketing can suggest segmentation, appropriate messaging avenues, and allow you to draw links between seemingly unconnected groupings in your response data.
It is often commented on how much time I spend on my mobile phone, but as anyone who truly knows me could tell you, my interaction is much more around setting reminders, updating to-do lists, writing collateral such as this – my phone, and more importantly Google Keep has become an extension of my mind, existing in silicon form, keeping me on track to be the most productive I can be, at whatever I am currently turning my hand to, which has always been marketing, but occasionally filmmaking, scribing popular science books, currency trading, anything to keep from being bored really.
The fusion of person and machine is often touted in the movies as a bad thing, where a person can lose their moral compass, and often their physical wellbeing through greater interaction with the much-vaunted artificial intelligence – the utter non-existence of which is a hill I will die on.
Machine learning however is here to stay, guiding us, making us the best versions of ourselves, and just think of all the cool marketing stuff that best version of yourself can achieve.