Join our team. Are you ready to change the game? Find out more →

News & Blog

5 big questions I have about Amazon Go’s cashier-less grocery store

News & Blog

I’m all for tech innovation, but when I heard about the Amazon Go store concept today — the first retail store could debut in Seattle early next year– I had to wonder.

You use your phone to check-in at the front of the store, grab the items you want, then leave. Sensors on the shelf and motion detectors powered by machine learning and computer vision know when you pick up an item and put it in your cart. In the explainer video by Amazon, someone picks up a cupcake and then puts it back. The narrator explains that the “store” knows you don’t want that item.

I get how this all works, and the comparison to the sensors in your car makes sense. For a quick refresher on those sensors — they work a bit like Doppler radar. A signal is sent out, it measures an object, then it comes back. This process is nearly instant and happens multiple times per second, and technology has advanced to the point that a Volvo S90 knows the difference between a duck and a moose. In the most advanced cars — say, a Tesla Model S — the car seems to be driving within an invisible shield, protecting you from the world outside. The Amazon Go store, it seems, is also in a bubble.

Except for one simple problem. When you drive a car, you are trying to avoid dangerous situations. For the most part, it’s a solo effort — there is only one driver. You’re on a fairly predictable path — it’s called the highway. Shopping is an incredibly random activity. My weekly pilgrimage to Walgreens or Aldi is fraught with indecision and unpredictable behaviors. The Amazon Go store would likely not have the best prices or the widest selection of products, given the small size and the fact that it will be a Tesla with groceries you can buy with your phone. I wonder about someone taking my phone and buying a month of groceries. And, I can see where this concept might present a few additional problems once it starts rolling out to the real world.

1. How will it catch thieves?

My biggest question is about shoplifting. People are crafty. They are also incredibly fast and subtle. How do you stop them? Also, hackers will probably target Amazon Go as the place to figure out how the sensors work. I’m talking about detailed schematics of the cameras as a way to circumvent any security measures related to Eggo Waffles and most energy drink brands. I’m not a fan of any criminal hacking whatsoever but I might become interested in what they come up with in this case.

2. How will it spot fast shoppers?

Speaking of Amazon Go hacking, what about people who are incredibly fast at grabbing products? And how do you prove they meant to buy eight different can of beer? Computer vision technology is good, but they have never seen me shop before. It is a blur of motion. Amazon Go will need to work incredibly fast for some shoppers who know what they want, move quickly, grab multiple items in one pass, push aside other shoppers, or unknowingly game the system in other ways.

3. What if you consume items in store and put them back?

And, what about downing a bunch of Red Bull in the store? I can only assume Amazon has thought about this scenario already, but if you drink a few cans quickly and put them back, then leave the store, are you charged or not charged? If you are charged, you could go back and argue that you never actually meant to purchase the drinks.

4. Can you return items as quickly?

This is a serious one. I really hate shopping. What I hate even more is returning stuff. It is stressful and unpredictable. To be honest, I don’t care about automated shopping as much as automated returns. If you buy a bunch of fruit snacks and then immediately go back into the store and set them back on the shelf, that would be ideal but I doubt that’s how it will all work. Again, it’s a hacker’s paradise and an easy target.

5. What if I put an item in someone else’s cart?

My last concern (if you can call it that) has to do with the confusion of multiple shoppers in the store at one time. You might be in a group of three or four people who all grab items and put them into the wrong cart. Would the computer vision system really be able to keep track of that? My guess is that everyone would have to behave perfectly when they shop, moving around in an orderly fashion. Like that will ever happen, right?


We take processes apart, rethink, rebuild, and deliver them back working smarter than ever before.