Vending the Right Stuff
How our Vending Machines are Integrating with AI
Vending machines are all around us. You can find them at the train station, by the local school or even inside your apartment block. Vending machines provide us with the items we need on demand in a completely automated way. However, the design has stayed relatively unchanged since they were first released in the 1st century CE. With the age of machine learning upon us, it’s time we improved upon the design.
Teching-Up
For starters, all vending machines should be outfitted with a digital screen and camera. In fact, many already are for security purposes, but we have something different in mind. With a camera, it would be possible to detect age, gender, race and other demographic features which would allow for the display of custom items.
Initially, the displayed items could be set manually. For example, a drink vending machine could display healthy drink options for the young millennials or more traditional tea for older viewers. While manually set to begin with, these options could be optimised throughout time for higher conversion.
Optimisation
What metrics can we use to optimise the purchases on a vending machine? Well, the simple one is purchases of items, but with some smart video analytics, we can do much more. For example, user attention can be tracked by recording the eyes and where exactly on the screen they are looking — maybe they’re more interested in one of the items you didn’t highlight? Furthermore, the linger time of the user could be measured, that is, how long they spend in front of the machine.
Actually, both of the metrics described above are well known in the e-commerce space under different guises; cursor tracking and time-on-site measure almost the same thing in a digital space. If we’re doing it in the digital space, what’s stopping us from applying the same principals to the physical commerce space?
Getting Personal
Taking all this a step further, it’s possible to customise what a specific user sees by recognising their face and tying it to a particular purchase history. Now this is a little creepy, and may very well require opt-in, so ways of encouraging this may be through discount, e.g.
“Want a unique experience? We’ll remember you and at the same time save 5c on your next purchase!”
With proper encouragement, an assurance of security on data and the option to delete; adoption could be reasonable. With this feature, you can make suggestions to the user on new things to try, and all the standard recommendation algorithms can apply.
The beauty of this is that as the number of vending machines adopting this process grows, the data can be shared across the network which can allow for even greater product recommendations as the data is not restricted to just one location.
What about other profitability methods? The food & beverage advertising spend in the US in 2016 was 200bn. Imagine that instead of television ads, instead of display marketing, you could target users right before they purchase the item and display content or coupons — powerful stuff.
A B2B Alternative
Vending can also be thought of in the B2B sense. Large companies all have an IT department, which, among other roles, is responsible for handing out IT equipment such as keyboards, mice, carry packs, etc. In some cases the IT department will document each request made, getting a signature and being very formal, but in other cases, they might not bother at all, and employees have free range of the equipment.
The use of a vending machine allows for accountability as you must swipe to get equipment, better inventory management as you can program alerts for when stock is low and a reduction in manual labour costs. Companies that create the IT equipment can leverage this convenience by offering a vending machine that stocks all their products in one simple package, along with a refill service. In this way, they can provide added-value to companies.
Current Usage
Now, that’s a lot of talk about vending machines, especially since they have such few applications, right? Not quite. Vending machines whether assisted or fully automated are making their way into more and more settings. Many McDonalds across the world now use automated purchasing machines; the convenience of replacing the ~10 clicks I currently do with just 1 based on my historical orders (that feeling of walking into the restaurant and asking for ‘the usual’) would be terrific. And it’s not just McDonald’s, some vending machines can now prepare a fully cooked meal and more.
There are various companies currently working on solutions including Silkron, Nagomi from Japan, SmartRetail in Asia as well as many, many more. Penetration is still low, but the current global market value is in excess of $1.3bn, which is expected to grow to $3bn by 2025.
Conclusion
Vending machines are ripe for innovation and the space for applying the new technology is growing — I look forward to seeing a simplified and improved process.
