The adoption of automated buying in two phases

For these past few weeks I have been posting a lot about personalized AI platforms and automated buying. As you know I strongly believe that this is a pivotal moment for brands because they still have the time to devise a strategy in order to avoid becoming “filtered out” in the age of platforms. Once this happens, it will be almost impossible to fight back. But for now, there is still time.

 

Fully automated buying will happen. Once the algorithms prove that they work flawlessly, we’ll simply trust them, just like we do today with our GPS-system. Not just for the low involvement products like toilet paper, either. But for other higher involvement purchases too, like insurances or even regular travel plans. But lately I’ve been thinking a lot about this form of buying that a lot of companies are currently cashing in on and that’s very difficult to predict and automate: impulse buying. Without it, Asos, Zalando, Amazon, Alibaba and many others would be a lot less successful. Difficult, but not impossible. That’s why, in the future, I see the automation happen in phases, according to these steps.

Automated needs

Automation works best for repetitive tasks in a data-rich environment. It’s one of the reasons why repetitive jobs – manual, but just as much cognitive – will be the first ones to disappear under the influence of technology. The same goes for the process of buying. You brush your teeth twice or more every day with about the same amount of paste, which means that you buy toothpaste roughly in the same intervals of time. You maybe travel each year to that same marketing congress in February. You probably renew the ink in your printer about every 3 months. You could hail an Uber every first Thursday night of the month because that’s when you go out drinking with your colleague friends. These actions all repeat themselves, and are quite easy to duplicate for smart algorithms, taking the human intent and action of buying completely out of the equation.

 

Automated wants

Now, impulse buying is a completely different affair. You might buy an extra-large pack of salt and vinegar chips because you had a rough day at work and you’re stress eating, even if your usual diet is quite healthy. Or you might buy a new pair of sunglasses, even if you have one because you saw that popular actor wearing them. Or you might buy a new lipstick for no reason at all, even if you have 10 pairs of that same type of red already. How do you predict, and automate that? If youdidn’t even know you were going to buy these things an hour ago, how could the algorithm? And a lot of buying in fashion, books, make-up and even small electronics is based on this type of buying. These are all merely based on a sudden “want”, rather than a “need” and therefore a lot harder to predict and automate.

 

But I don’t believe that they cannot be automated. I’ll always remember how Nadira Azermai, CEO of ScriptBook, once told her audience “Humans are a lot more predictable than we like to think”. Her company scans scripts and then predicts the odds of them becoming a success. You’d think that with the changing fashion in movies, or the fact that we are dealing with art, this would be excruciatingly difficult. But their success rate is very impressive.

 

That’s why I have no trouble believing that, as long as we have enough contextual data, what we call “impulse” buying can be predicted, too. If you know that Jane likes to “impulse” buy clothes when she is stressed and you know what stresses her and are able to measure it (phone calls with her mother in law, doctor’s visits, working late, traffic jams between work and home,… whatever) and you know her favourite color, type of brand, skirt length, etc. from past buying behaviour, then you could predict her next action in the matter.

 

So, the full and frictionless automation of buying will probably happen in phases. First, the repetitive “needs”, based on buying behaviour, but then quite possibly the impulse buying of “wants” will follow, based on a combination of buying behaviour and contextual data. In case of that last one, the question will of course be: do we wantthese things to become automated? Because this could very well open up what I like to call a temptation island for marketing experts, with strategies that could surpass the boundaries of ethics. These are things that we need to think about now, before they happen.