Amazon Prime Air testing drone delivery
Without question, Amazon and the holidays have become synonymous. With over 100 million Prime members enjoying free one-day shipping this year, every move the e-commerce giant makes garners a lot of media attention.
Yet Amazon is still a relatively small player when looking at the global retail market. With $232 billion revenue in fiscal year ending 2018, Amazon was less than 1% of the $25 trillion global retail market and just 15% of the $3.5 trillion global e-commerce market. By contrast, Walmart, the largest retailer in the world, is nearly twice Amazon’s size at $514 billion in revenue.
That said, Amazon is investing more in R&D than any other publicly traded company in the S&P 500. Since acquiring the mobile robotic fulfillment startup Kiva Systems in 2012, Amazon has invested over $100 billion in R&D compared to Alphabet ($87 billion), Intel ($84 billion), Microsoft ($83 billion) and Apple ($58 billion). During the same period, its stock price has soared by 843%.
Kiva robots working the warehouse
Many attribute Amazon’s rise in dominance to its fleet of 200,000 packing and shipping Kiva robots. Coupled with 30-minute drone deliveries by Prime Air, Amazon is seeking to lead the way in leveraging technology to get customers products faster, cheaper and in a more sustainable way.
Kiva robot close-up
I spoke with former Amazon executive Vince Martinelli who worked at Kiva with founder Mick Mountz. The two were MIT classmates and have since enjoyed a friendship that has spanned 35 years. They currently work together on piece-picking robots at RightHand Robotics where Mountz serves on the board and Martinelli heads up product and marketing. Founded in 2014 by Harvard roboticists, Yaro Tenzer and Leif Jentoft, the Cambridge-based startup has raised a total of $34.3 million with last year’s Series B round led by Menlo Ventures Mark Weiner. What follows is an edited transcript of our conversation.
How did Kiva solve the problem of fulfillment at Amazon?
Prior to Kiva, Mick Mountz headed up business process logistics for the online grocer Webvan. When it failed because of fulfillment problems, he went on a mission to redesign warehouses at fulfillment centers to bring the cost down for shipping a bag of groceries. Having previously worked at Apple engineering, he had given some thought to automation and built a massively parallel processing machine whose software could dispatch robots to fetch shelving pods of products and move them to central workstations for packing. The ability to gather the products and sort out the orders turned out to be a huge advantage over other systems. It eliminated whole steps and zones in the warehouse. The parallel processor was really quite a powerful idea. When Amazon acquired Kiva, all it needed to do was plug it the Kiva processor into its motherboard to add shipping to the system.
How do the pick-and-pack robots work?
Items from totes are grasped by the picker and sent downstream in a conveyor to a pack out station. The grasping system has fingers that conform around the item and can pick up items that weigh between two kilograms and five pounds, down to the size of a lipstick which is a good fit for the majority of fast moving products that sell well online.
Our computer vision system helps the picker distinguish one item from another. If there are several million skews in the warehouse, the robots can’t possibly be trained on all of them and next year there may be new sellers in the marketplace. There’s no way to know what products are going to be flowing through the building on any given day so the system needs to be very flexible in how it perceives and grasps the items.
Our system doesn’t have to learn the item set before picking, it’s model free. What that means at the AI and machine learning level is that the picker is figuring out what’s important in the data it’s capturing so it can learn from it. Then it runs it through a neural match to build classifiers that makes the machine better. This saves tremendous time skilling up an employee to get comfortable with the task.
As for our tech stack, there’s quite a lot out there that didn’t exist five years ago and we use open source third party tools when available, like from Google.
Who are your competitors in the picking space?
It’s a young market but we have a pretty good lead. Our picker can integrate with a wide variety of third party systems. It works just as easily in the back of a grocery store as it does in a large warehouse. We use data to build an intelligent picking platform with flexible scalable automation that works with a variety of software environments and allows for predictable fulfillment capacity.
When are we going to see drones or sidewalk delivery robots dropping packages at our doorstep?
It could take longer than people expect because there are safety and liability considerations, like if a product doesn’t make it to its endpoint.
Friends from MIT worked at Segway and despite the fact that they could be ridden safely on public sidewalks they were banned in many places. Scooters too. This is the risk that sidewalk delivery robots face.
These robots can prevent injuries of workers by taking over physically demanding pick-and-pack tasks, but by doing this, doesn’t it mean they are taking jobs from humans?
Amazon employs more than 650,000 people and Jeff Bezos started as one guy, so that’s a pretty good rate of growth over 25 years. He’s obviously hiring people and the robots are helping him win in the marketplace. They literally can’t find enough people within driving distance of these facilities which is why they’re going deep into the labor pool to find anyone who’s able to work a few hours and is reliable. Robots help significantly with the labor shortage.
Will one day warehouses be fully automated with no humans onsite?
For retail there’s an equation, if the facility is too far away, there are transportation costs. Additionally, today’s automation still requires people to operate machinery, manage software updates and provide preventive maintenance. Teleoperation is not a great substitute for these tasks.
Is the death of retail greatly exaggerated?
People still like to go into stores and browse but new models are needed. Perhaps stores can serves as entertainment with a showroom where people can hang out with friends and order what they like from kiosks.
What industry shows will you be attending?
In January, we got to NRF to check out front of store, point of sale and supply chain logistics. We’re going to be in the innovation area and likely in at least one partner booth. In March, we’re at a materials handling show called Modex in Atlanta and at a fulfillment design show called Logimat in Europe.
6 River Systems integrates Chuck robot with Soft Robotics’ Superpick
6 River Systems
How would you advise retailers trying to navigate the Amazon effect? Adapt or die?
If you’re a company that moves a lot of product and you’re not thinking about this, your competitors are and you’re going to fall behind. Bake automation into your strategic plans upfront.
In the retail wars, will one ring rule them all?
History has shown that whoever can figure out the model and scale it fast can own a market for a long time. As sales move online, there are major logistics issues that need to be solved.
Much of physical retail is still based on the self-service model which supermarket chain Piggly Wiggly instituted over 100 years ago. In this model, employees work a register and customers pick products and transport them home for free.
In Amazon’s direct-to-consumer model, picking and shipping reverts back to the retailer. They can’t charge twice as much for the product so they need to figure out how to get items to people cost effectively, quickly, and in the proper condition.
Right now, there are many disparate pieces coming together with robots across the spectrum. Everyone is trying to figure out what to do next.
Amazon always starts with what’s the problem that needs to be solved and then works its way backwards to find the right technology. If Amazon doesn’t win this market, it won’t be for lack of trying. They’re a formidable player in this space. Looking at their span of investments from their fully automated Amazon Go stores with no checkout to their last mile drones and extended fleet of airplanes, they’ve also just announced a brand new $40 million state-of-the-art Amazon Robotics Innovation Hub being built in Massachusetts that will add 200 new technology jobs. [To date, Amazon has invested over $3 billion and created more than 4,000 technology jobs in Massachusetts.] Last year they bought a full service pharmacy, PillPack.Not sure what their plans are but they’re certainly are moving fast in multiple directions.
Frontline producers visited us last year to do a story about who’s going to be the dominant country in AI and robotics: China or the U.S. Im watching the show, I was stunned by the scale of Alibaba’s operations, how far along they are, and how much their warehouse robots look like Kiva’s. China has a lot of cash and scale so it may not be a foregone conclusion that the winner will be the U.S.
What’s the outlook for 2020?
This coming year is really pivotal. Some of my former Kiva colleagues are at 6 River Systems which just got bought by Shopify to build an automated e-commerce marketplace. They’re integrating their Chuck robot with Soft Robotics Superpick. Walmart is working with Alert Innovation on building out a fully automated Novastore in New Hampshire following its pilot with grocery picking robots, Alphabot. Kroger is working with Ocado on an automated warehouse for groceries. Fabric (fka Commonsense Robotics) is offering micro-fulfillment for scalable automats, and Takeoff Technologies is figuring out the click-and-collect model for grocers to test. There’s such a great pool of technology right now, I think we’re going to see a number of exciting announcements over the next year.