When algorithms go shopping
How will you advertise to a fridge?
I had the most terrible morning today: I ran out of coffee. And I had no idea it was going to happen. I have an expensive coffee machine (my wife reminds me about it every day), with plenty of sensors, displays, and buttons. But — as smart as it is — the machine cannot do the simplest thing in the world (at least these were my thoughts in the morning) — remind me to buy beans. BEFORE. IT. RUNS. OUT. OF. THEM.
My expensive “smart” coffee machine. Eyes added for emphasis.
And it shouldn’t be hard, in the 21st century, should it?
I buy my beans from the same place, all the time. I have precisely three strong coffees a day, every day. It’s not hard to calculate how long a bag of coffee beans will last for, roughly. The coffee machine doesn’t even need to know my average consumption — it knows exactly how many coffees I am having. A truly smart coffee machine should be able to remind me to buy beans. Or better: buy them for me. Online.
And save me from having terrible mornings.
There are businesses that think it’s a great opportunity. Just like we have had office printers that can automatically request maintenance or toner change — making our offices “self-driving,” we can expect our homes to become “self-driving” too, soon. And while our home devices, in most cases, cannot yet perform physical tasks on our behalf (such as the “smart” fridge below, waiting for a human to close its door), they are becoming quite good at representing us online.
What we consider “smart” might differ, depending on who you ask.
If you live in the US, you can buy dishwashers and washing machines that automatically reorder detergent. Smart fridges will soon be able to automatically reorder beer and ice cream, among other products. Whatever rocks your boat, the algorithm in your fridge will make sure you stay happy (or healthy, if you prefer)!
Welcome to a world where algorithms go shopping.
Welcome to a world where retailers soon will no longer sell just to “human-customers.”
For the first time in history, we have given algorithms — robots, appliances, applications — the capability to make purchasing decisions on our behalf.
This new world creates an entirely new set of challenges for retailers. As I explored together with my colleagues Michael Rosemann and Paula Dootson in our prediction of future of retail, these shopping algorithms (embodied in various devices) will be smart enough to take into account our preferences. They will not just buy any beer, or any ice cream, or — my pain today — any coffee. These algorithms will understand whether we prefer to save or splurge, buy local or perhaps premium products, or how quickly we want them delivered.
Coffee? Drone-delivered, in the next 5 minutes, please.
But will these algorithms behave in the same way every time, or will they sometimes surprise us? When I buy my coffee, I like to experiment. Can you imagine a “hipster” algorithm, buying coffee that no one else buys?
There are three main stages that shopping algorithms will go through as they evolve their independence: hard-coded shoppers, preference-based shoppers, and inspiring shoppers.
The three independence stages of shopping algorithms (shoppers).
Hard-coded shoppers
Currently, most known algorithms that independently make purchases are considered to be in the first stage: hard-coded shoppers. They make purchases of a pre-defined product (often called a replenishment). The dishwashers I mentioned before will always buy the same dishwashing liquid. From the same seller. From the same manufacturer. That’s what they’re told to do.
But soon, regulators will realize that — with a growing market — such an approach leads to monopoly-like situations. Whoever sets the predefined list, unless it’s the human customer itself, will be in a position of restricting access to retailers, and preferring certain other retailers. We have already seen examples of it, even when merely differentiating channels (and not yet fully automating the purchasing process).
Preference-based shoppers
While current approaches involve a closed ecosystem (e.g., Alexa only buys Amazon-sold products or products listed by Amazon), we should expect that at some point regulators will step in to ensure equitable access for other providers.
We have seen such interventions before — in the computer software space — so it should not be a surprise in the shopping-algorithms’ space. And we can expect pushback not only from governments but also from customers (we want real choice!).
This pushback will fuel evolution to the next stage: preference driven shopping, where purchasing algorithms will take preferences into account, and choose from a broader, possibly practically infinite set of goods or services. We have seen a similar development — from restricted to open markets — with existing product categories such as coffee machine brands opening up to competition in the coffee pod market. As you — I am sure — gathered by now, I follow the coffee markets quite closely.
The preference driven algorithms will consider priorities of the human consumer: you and me. They will perform a broad search for goods or services that match our preferences. Just like information retrieval, where the best web search engines return the results most relevant to the specified query, the algorithms will aim to deliver the most relevant good or service. In other words, I’ll get the best coffee my coffee machine can find for me. Sounds great?
Inspiring shoppers
Will preference-based shoppers satisfy everyone? Just like a typical information retrieval system, that only gives you answers; a preference-driven algorithm only gives you what you already know you want. A cheap coffee. A strong coffee. An organic coffee. No surprises there. Can a coffee machine surprise or inspire me? Perhaps it can!
The third stage of evolution of purchasing algorithms: _inspiring shoppers_, will bring the first algorithms that suggest unexpected — or non-obvious — purchases, effectively surprising (hopefully, in a good way) us. Imagine a purchasing algorithm, aware of your preferences (as in the previous stage), offering you a seemingly unrelated product. It could be the coffee purchasing algorithm inspiring you to try kombucha as your daily drink — just like a good salesperson would. Or — bear with me here — will my coffee machine realize I am getting a kick out of playing computer games, and suggest the newest first-person-shooter game, instead of a coffee in the morning? Sounds ridiculous? A lot of great ideas do, initially.
What does this new world mean for us?
If algorithms start making purchasing decisions, and their choices are not hardcoded, we need to start treating them for what (or who) they are: customers. And how should retailers advertise to algorithms? Yes, you heard it correctly: how should retailers advertise to a coffee machine?
How about a few more questions:
Can you convince an algorithm that it should buy from you?
Can you convince an algorithm that it should buy now and not later?
Can you negotiate prices with algorithms?
Does the speed of my Internet mean I can offer my products to a fridge-customer faster than my competition?
Shall I hire experts to reverse-engineer algorithms of smart fridges?
Could I hack into Marek’s coffee machine and change the software to buy only from me? This is not different from many marketing campaigns turning humans into loyal followers of particular brands.
How will retailers compete and make their product seem favorable to an algorithm driven by business rules and possibly not influenced by any form of emotions?
Will the rapidly rising discipline of behavioral economics lose its relevance in a world where algorithms go shopping?
A new market, effectively parallel, will emerge, in which retailers will operate. Anna Oberlaender and her colleagues are exploring how this leads to the emergence of new partnership models — business to the algorithm (B2A).
Isn’t it creepy?
With the assistance of the right sensors, an algorithm running inside a dishwasher could be ordering its detergent when it knows it is running out, which completely removes the human from this stage of the process.
A coffee machine will order coffee pods. My car will never surprise me with a flat battery on a Monday morning. Instead, it will request a replacement at the end of life of the old battery (I’ll remain suspicious and think the service station is just trying to make an extra buck on me). A bottle of milk will be waiting for me in front of the door every morning I need it. Nappies delivered to the door. Grass mowed.
Wait. Won’t it be creepy? How did you know I was running out of toilet paper?
The good news is: the creepiness factor can be managed. To avoid potential creepiness (a parcel surprisingly appears on your doorstep), some existing purchasing algorithms require us to configure them first. For instance, GE’s connected dishwashers come with a mobile phone app, which is used to set the purchasing behavior of the dishwasher. Not that they don’t have the required information — they’re just ensuring we think we are in control of the process.
Is it good news, or bad news?
The emergence of algorithms that act on our behalf will create a lot of opportunities. We need to make sure we are ready. Manufacturers will need to learn how to build devices that order products and services on behalf of humans. Retailers will need to learn to sell to algorithms. Customers will need to learn to work with algorithms — imagine how much time can we save if we get rid of boring activities such as buying a dishwashing liquid.
There will be new jobs: fridge campaign managers (fridges as a target group), proactive delivery couriers (I need one to deliver my coffee before I run out of beans), and many others.
Yes, some jobs may disappear too — perhaps there will be a lower demand for shopping assistants? We need to understand and prepare for it.
And I would do anything to make sure I never run out of coffee.