Amazon Go is both a data mine and a tech testing laboratory

Amazon recently opened a brick-and-mortar grocery store, Amazon Go, that “has no checkouts and instead works by tracking what users buy with AI-powered cameras.” Why would they make this move to brick-and-mortar? I can see four reasons.

  1. As of September 5, 2017, Amazon “controls 460 Whole Foods locations across the U.S.” which means that Amazon Go would make sense simply as a retail experiment to increase efficiency. E-commerce is growing rapidly, but a huge percentage of retail sales (around 90% in the US) still happen in the “real world.”
  2. Also, as Juan suggested in our February 19 PUB 802 class, it could be that Amazon is trying to change our idea of what shopping is. After getting used to shopping in a store without queues, maybe we’ll find it even more unbearable to wait in line for our groceries. Amazon could be trying to tip our preferences toward shopping online even more than we already do. Along that line of thinking, Amazon Go is a way for them to collect data on the physical shopping experience; data they would otherwise have to pry from their competitors’ cold, dead hands.
  3. In fact, collecting data is a good enough reason in itself for the move to brick-and-mortar. Data has been called “the oil of the digital era.” Why would an online marketplace care about data from physical marketplaces? Companies need accurate, up-to-date data from all areas of consumer behaviour in order to predict trends and preferences. If you’re selling shoes, you need to know more about the consumer than their current shoe preferences; you need to know who they are, what they will like in six months, and why. I can’t begin to tell you exactly what data they are collecting, but I think Amazon Go is a great example of what Evgeny Morozov calls “data extractivism.” Morozov quotes Andrew Ng saying that “at large [tech] companies, we often launch products not for the revenue but for the data … and we monetise the data through a different product.” Whatever the data is, I think Amazon Go is less of a store than a product; and it’s of a new class of product designed not to collect revenue, but data.
  4. Perhaps most importantly, though, all that data about consumer behaviour and preferences needs to be monetized, often using AI. That AI needs to be tested on a large scale, in a realistic setting, and Amazon Go is the company’s live tech laboratory. Take Amazon’s Echo Look as an example. While most people still prefer to buy clothes in person, Echo Look will, as Juan predicted, teach them to buy online. It takes a picture of you, analyzes it, and makes recommendations; and “it’s always learning.” Amazon Go will always be learning, as well; it’s the perfect lab for Amazon to test new technologies in “computer vision and machine learning.” Vice President Gianna Puerini says, “This technology didn’t exist [before Amazon Go]. It was really advancing the state of the art of computer vision and machine learning.” Amazon Go allows Amazon to test and develop their computer vision and AI technology at a much larger and faster scale. (Especially considering that consumers will take time to adapt to letting a cloud-enabled camera watch us change—although less time than we would have taken in the past). 

Conclusion: An example of evolution

I think Amazon’s main reasons for opening a brick-and-mortar grocery store are to collect and monetize data, and to test new technologies. The move into brick-and-mortar is an evolution of their strategy, not a devolution. It’s a strategy that creates for Amazon an opportunity to access up-to-date real-world data, and to test new AI technology with real consumers. These are exactly the things that a big tech firm that runs on data should be thinking about.

Leave a Reply