Introduction
We hear it over and over again – “It’s the retail apocalypse!” “The store is dead!” However, here at RevTech, we do not think that stores are dead – they are very much alive, just maybe in different capacities than we traditionally think of stores. Amazon has paved the way to a new store experience with the un-manned, checkout-less Amazon Go stores. Countless digitally native DTC brands – Bonobos, Away, and Warby Parker, to name a few – have begun opening stores that resemble showrooms. So, perhaps we should change our refrain to, “The store is different!” Whether that means repurposing physical space to an experiential showroom, a micro fulfillment center, or something else entirely, we are on the hunt for innovative, game-changing retail concepts. RevTech invests in retail concepts that have not yet been done, or done well, and show promise for re-shaping the store experience as we know it.
Exploring new retail concepts
In the larger spectrum of new retail concepts, we see three specific concepts that are taking off and may be viable investment targets. In this report, we will cover:
- A “mall without borders”
- Small retail footprint + micro-fulfillment center in store
- Contactless retail in high-traffic areas
- Automated showrooms
For each of these, we will describe what they are, what the market opportunity is, any relevant case studies, and the investment landscape.
“Mall without borders”
What: Repurposing the mall as we know it while staying true to the needs a mall serves. This concept will be a large physical space with many different brands, stores, places to eat, services, etc., just like you find in a mall today. However, the checkout experience will be seamless – e.g. no individual store checkouts, just one contactless or smart checkout when the shopper leaves the mall. The “mall without borders” allows customers to have an easier shopping experience while providing all tenants with actionable insights and data supported by the in-store tech. This technology will include checkout data, shopper path data, engagement data, and will leverage AI, computer vision, machine learning, location tracking, among other emerging technologies.
Market Opportunity: In 2019, according to Green Street Advisors, there were about 1400 malls in the US. [1] Currently, as of June 2020, there are closer to 1000 malls in operation. [2] Malls are rated on a scale of A++ through D-, with A grades being the best, or “top tier.” According to Green Street Advisors, the top 20% of malls are responsible for about 75% of total mall revenue, meaning that top tier malls are benefitting the most from foot traffic and have the most sales per square foot. [3]
Over the past few years, prior to the COVID pandemic, Class A malls saw increases in sales per square foot while Class B and below saw decreases. [4] In order to take advantage of this growth, malls that do not meet the Class A criteria need to invest in remodeling and redeveloping their spaces. This involves updating buildings, adding new brands, including regional ones to cater to local shoppers, introducing additional services more lifestyle-oriented businesses – e.g. fitness centers, co-working spaces, or hotels, and amenities. A Placer AI report found that the decrease in sales per square foot for Class B and below malls would have been even worse if you were to exclude those malls that renovated to become Class A. [5]
All of this is a positive indication that the “right” malls will continue to thrive. Even with the coronavirus pandemic, we still see mall traffic recovery gaining momentum among the top malls. [6]
This concept therefore has two opportunities on which to capitalize in the US: 1) building upon existing Class A malls and enhancing the technology (which is already in progress), and 2) repurposing viable Class B and below malls with improved offerings and technology. Looking at the Westfield Century City mall in LA as an example, YOY traffic increased 92.7% after a $1B renovation [7] (from about 500K monthly visitors to 1M). Just considering the top 20% of malls as an opportunity for this concept, that could mean a possible impact on 200M monthly visitors. Assuming a conservative 20% conversion rate and an average transaction value of $50, that is $2B in revenue per month that could be captured. Factoring in the growth of traffic and engagement that this concept could lead to, this number becomes even larger.
Case Study: Westfield Century City
In 2017, Westfield Century City underwent a $1B renovation that resulted in its ranking as the “number one retail center experience” by Chain Store Age. [8] Westfield VP of leasing David Ruddick calls it a “modern retail resort,” . [9] In addition to brands and fashion stores, the mall has fitness centers, including Equinox and Gloveworx, health and wellness clinics, restaurants and eating/drinking experiences including Eataly, VIP services including private lounges and elevators, and an outdoor atrium for concerts and events. On the technology side, it offers app-based self-parking and VR experiences.
The parking upgrade executed with the renovation is a good example of how technology can lead to a much smoother customer experience at malls – especially considering how painful parking at a mall can be. Before even looking at the tech side, one of the main things Westfield did are part of the renovation was to increase the number of available parking spaces from 2400 to about 5000, as well as increase access points with additional entrances, allowing for more visitors and less traffic. [10] Additionally, they started offering two types of technology-enabled ticketless parking: Smart Parking and Reserved Parking. For Smart Parking, customers download and app and register their license plate and car in advance. The system will then recognize the car upon entry and exit and simply charge a card on file, eliminating the need for paper tickets, waiting in line to pay, and other inconveniences. With Reserved Parking, customers can pay extra to reserve a secure parking spot online and will be shown where to park via a digital screen with the customer’s name and parking space upon entry.
In the first year following the renovation, traffic increased 92.7% (as mentioned previously). However, what is perhaps more noteworthy is that traffic didn’t level off after the first year – from July 2018 to July 2019, traffic continued to increase 7.9% YOY . [11] Westfield Century City is now in the 99th percentile in terms of traffic nationally.
With the rise of the pandemic, Westfield Century City has begun offering additional services to ease the customer experience. For example, they are partnering with Happy Returns for contactless returns via a curbside concierge service. [12]
Westfield Century City is a prime example of how transforming a traditional mall space into something engaging, unique, and technology-enabled can lead to immense success in terms of foot traffic and revenues.
Investment Landscape: B2C Real Estate Services & Retail
Since 2011, there have been 26 deals across 203 companies, meaning that many companies see more than one deal [13]:
Most deals are M&A / corporate deals, with a few years seeing a higher amount of PE and VC deals and 2016 having two large IPOs. Geographically, most deals happen in North America. Top investors are PE firms or real estate investment firms. The deal with the most capital raised was $98.22M in the IPO of Mexican real estate development company Planigrupo, which develops shopping centers.
Recently, there has not been as much deal activity in this space as prior to 2018, but we are seeing more investors buy Class B and C malls cheap even as total investment volume has slowed. Often, these deals may be more of a play for the land itself rather than the actual mall. [14]
Small retail footprint with “micro-fulfillment” center
What: This concept involves repurposing exiting retail real estate as warehouses and micro-fulfillment centers. The concept is already gaining traction with grocers and drugstore retailers as perishable items and other high-turn items are difficult to ship from a centralized warehouse efficiently. Micro-fulfillment centers can increase delivery speeds to end customers and make inventory management more efficient. Furthermore, brick and mortar retailers can leverage their existing real estate, including closed stores, in order to execute this, thereby creating additional value out of an existing asset. For this concept to be successful, it requires investment in technology in addition to converting retail spaces into warehouses and fulfillment centers. Automation, AI, and robots are a few of the key technologies that are enabling retailers to execute this. [15]
Market Opportunity: In grocery alone, US online grocery sales are expected to reach $60B by 2023. [16] COVID-19 has only exacerbated this already prevalent trend, with sales hitting $7.2B (a record number) for the month of June 2020. [17] This increased demand for online grocery delivery and pickup has highlighted the need for improved last-mile delivery logistics, which is where the micro-fulfillment center comes in.
In 2019, there was an estimated 40,000 grocery stores in the US as well as an additional 155,000 convenience stores. [18] Looking at just large-format supercenters that are larger than 50,000 square feet, we are looking at 220M+ square feet of retail space. This is space that at least part of (e.g. 15,000-20,000 square feet) can be repurposed as small-format warehouses and micro-fulfillment centers focused on inventory management and last-mile delivery.
Case Study: Sedano’s Supermarkets and Takeoff Technologies
In October of 2018, Hispanic grocer Sedano’s Supermarkets and Boston-based startup Takeoff Technologies announced a partnership for an automated “hyperlocal” fulfillment center in Florida that would serve 14 Sedano’s location in the Miami area. [19] The micro-fulfillment operation runs from one of Sedano’s stores in an 11,000 square foot space (much smaller than a typical warehouse). [20] The partnership allows for fulfillment of pickup and delivery online grocery orders. The solution allows Sedano’s to pick 60 items in five minutes or less and fulfill customer orders in three hours or less [21], leading to decreased wait times, increased efficiency, and increased revenue as the volume of orders can be much higher.
Takeoff Technologies calls themselves both “hyperlocal” and “automated,” in contrast to Instacart, whom they consider a competitor, being just “hyperlocal”. [22] Both are essentially ”fulfillment as a service.” Takeoff mainly develops the software and has partnered with Knapp for the hardware piece to assemble their micro-fulfillment centers. Each micro-fulfillment center then has 10-15 staff members, which are provided by the retailer. [23]
Therefore, the benefits to the retailer are increased sales due to faster fulfillment (Takeoff says their solution can fill orders 10x faster than Instacart) and increased inventory turns. The costs are additional labor for the micro-fulfillment center and costs associated with implementing the solution. If the micro-fulfillment center is placed within an existing store, the retailer can likely leverage existing labor with a few additional people as staff can easily switch between store work and fulfillment work. [24]
If an entire store is converted to the micro-fulfillment center, becoming a “dark” store, then the labor needs are much less for the micro-fulfillment center than they would be for a store of comparable size. Furthermore, the costs of the lease may be lower because there is less of a need for “premium” retail space. [25] In either case – integrating the fulfillment center into part of an operating store or taking over a store completely – the volume of orders expected must be high enough to justify the additional labor and vendor costs. According to Takeoff Technologies, 700 pickup orders per week or 4500 delivery orders per week needs to be the volume for a strong business case, and anything less than 300 orders per week should still be done manually. [26]
On the real estate side, leveraging a micro-fulfillment center can cut real estate costs for retailers due to the need for large store spaces decreasing when fulfillment is condensed into smaller spaces. Retailers may need less space overall for their stores as well. Therefore, we can expect to see an impact to the real estate industry, especially among developers of large format real estate. A typical micro-fulfillment center only takes up about a third of a supermarket’s space, providing grocers, for example, with a great opportunity to optimize their space usage. [27] According to real estate investment firm CBRE, availability of small warehouse space has dropped 4% over the last five years and rents have increased 30% (compared to 16% increases for large format warehouses). [28]
Investment Landscape: Micro-Fulfillment
Much of the investment in this space is centered around enabling technologies that partner with retailers to execute the micro-fulfillment concept (as in the case study above). The exception would be a few large retailers that are able to implement this themselves, such as Walmart which has piloted pick-up only locations using existing real estate. [29]
Major startup players include:
- Fabric, valued at $56.59M, Series B stage
- Takeoff Technologies, valued at $500M, Late Stage VC (post-Series C)
- Dematic, acquired by German KION group for $2.1B in 2016
- Ohi is an emerging player that is valued at $10M after raising a $2.75M seed round in 2019.
On the real estate side, demand for smaller warehouse space has increased over 30% in the past five years, indicating the shift in trend toward smaller, micro-fulfillment centers. [30] However, the aforementioned technology companies often partner with retailers to implement their solutions in existing retail space, or in their own warehouse space.
Contactless retail in “high-traffic” areas
What: A “re-imagined” convenience store, fast casual lunch spot, or vending machine is the main idea behind this concept. Basically, everyday places that get a lot of foot traffic due to their location – e.g. a vending machine on a college campus, a sandwich place in Midtown Manhattan, a shop in an airport or train station – are great candidates for this concept. As we dig into deeper in the “recommendation” section that follows, a number of trends have exacerbated the need for easy, contactless retail in high-traffic locations such as these. Namely, the COVID-19 pandemic has increased the desire for contactless checkout and additional health measures. In addition to this, customers are demanding faster, easier, more seamless retail experiences that were already started to take hold prior to the pandemic (e.g. self-checkout at the grocery store, electronics vending machines at airports).
Having a small format, contactless “store” or “walk-in” vending machine in a high traffic area will lead to increased revenues due to high foot traffic, decreased wait times, and seamless checkout.
Market Opportunity: The Amazon Go store in Brookfield Place in lower Manhattan saw, on average, 50 people enter the store every 15 minutes. [31] The estimated sales for each Amazon Go store is $1.5M per year. [32] Assuming an average square footage per store of 1800 square feet (based on the range of store sizes from 1200 to 2400 square feet), that is $833/square foot annually. Average in-store sales for a convenience store in the US is also around $1.5M [33], but with an average square footage of 2425 square feet [34], making the national average $618/square foot.
Therefore, we can assume that a contactless store in a high-traffic area (such as locations where Amazon Go stores are placed) has the ability to make more revenue than convenience stores in their current status. Furthermore, in-store sales for traditional convenience stores are continuing to grow (growing 22% YOY from 2017 to 2018), so the contactless model can continue to build upon this organic growth.
Furthermore, the global contactless payments market is expected to reach $18B by 2025, and retail is expected to be the market leading vertical for contactless payments. [35]
Case Study: 7-Eleven in Taiwan
Taiwan’s 7-Elevens are already different than the ones we are used to seeing in the US – they have localized offerings such as bubble tea and rice balls, they often have seating where people can enjoy meals, and they offer services such as utility bill-pay, banking and more through in-store kiosks. [36]
However, in addition to these features, 7-Eleven has also rolled out “smart vending machine” stores and “X-Stores” leveraging AI and biometric facial recognition in Taiwan. These stores are unmanned convenience stores (similar to an Amazon Go concept) that allow people to “face in,” “face pay,” and “face out,” using facial recognition technology. [37]
7-Eleven has 60.5% market share of the convenience store market in Taiwan and continued growth in sales that surpasses the average retail sales growth rate in Taiwan. [38]
Investment Landscape: Contactless Payment & Computer Vision
In a recent survey, 22% of retailers indicated that they would invest in contactless payments post-coronavirus and 2% said they would invest in computer vision and in-store robotics. [39] Coronavirus has been the “inflection point,” according to NRF, for retailers to start investing in these technologies. [40]
While the demand for these technologies is growing, contactless stores, autonomous checkout, and compute vision is a crowded space. Retailers such as 7-Eleven (discussed above), Amazon, and Ahold are already testing and rolling out unmanned stores leveraging these new technologies. [41]
For retailers that can’t build this technology internally, Amazon is now selling its Amazon Go technology to other retailers. [42] Additionally, the following US-based startups have launched autonomous checkout tech:
- Accel Robotics, RevTech portfolio company, Series A Stage – $30M raised for Series A
- Zippin, Valued at $37M, Series A Stage
- Standard Cognition, Valued at $535M, Series B Stage
All three of these companies raised their most recent round in 2019.
Automated showrooms
What: Automated showrooms would be the natural expansion of showroom retail. While fashion showrooms pioneered by Bonobos and others often come to mind, other retail industries can be just as effective with the physical presence of touch and see before you buy. These can include fashion, furniture & appliances, and household goods. Specifically, these showrooms would use QR codes as means a build a basket on the customers phone and, upon leaving, enable 1-click checkout and shipping. Image trying on a piece of clothing in a fitting room, snapping the QR code on the tag immediately, and having it show up on your doorstep a few days later. The online ecommerce site turns offline.
Market Opportunity: The trend of inventory lite showrooms and pop ups has evolved over the years through D2C brands trying to expand via a physical presence without losing their ecommerce heritages. Additionally, to incentive and draw customers, brands will have to invest more in immersion rather than sales capacity, e.g. investing more in lighting and fixtures to create an “Instagrammable” moment. The gap here is how to maximize conversion. If they cannot supply immediate gratification, then effortless ecommerce is the next best alternative.
Additionally, the commercial real estate landscape changes dramatically. Where, previously if you help stock, you needed a large space in, often prime, real estate to have a presence. Not only was this costly, it limited the reach to only customers in the nearby geography. Instead, without the inventory burden, you could open several small spaces, spread out across the city, and capture significantly more potential customers. Automation only drives the feasibility further by decreasing the dependence on human capital.
Case Study: Nordstrom Local
Nordstrom local tested the traditional retail model – can a brand like Nordstrom be successful with a store that sells nothings? In short, after their initial openings in Los Angeles, customers in Nordstrom Local stores spend more and return goods faster, giving the retailer a better chance of reselling returned products. [43]
The Local store is the size of a Starbucks – where you can get fitted, personalized style advice, alterations, do pickups and returns, and order online. More specifically, customers work with personal stylists to put together ensembles, using tablets or phones. The outfits are usually requested from a nearby full-size Nordstrom store and delivered — sometimes within hours — for customers to try on in Local’s dressing area. [44] As other competitors such as Saks are investing into their flagship stores, Nordstrom is taking the showroom effect to boost its sales.
But critics are also skeptical of the benefits these added services bring. “This is just a slower, more expensive version of online,” said Bob Phibbs, chief executive of The Retail Doctor, a consulting firm. This highlights the importance of creating an ease for impulse purchase at the showroom, e.g. effortless and immediate purchases via QR codes as described in our concept earlier.
Investment Landscape: Traditional retailers and D2C are embracing showrooms
Younger customers are becoming more amiable to and driving growth in the showroom concept. 55% of millennials surveyed want showroom stores vs. only 28% of boomers. [45] A Wharton study, described the supercharged customer, i.e. the customer post visiting a showroom. These supercharged customers will spend up to 60% more on average on a particular order. They also show up at a higher velocity, so the time between purchases is reduced by 28%. And they tend to buy 20% more categories, so they will expand [their purchases] into new categories. [46] As showrooms expand, new technologies and concepts enabling their success will become excellent investment opportunities in the future.
Wal-Mart was one of the first to rethink its traditional strategy with the purchase of Jet.com, Bonobos, and various other ecommerce players. Nordstrom is testing the concept with Nordstrom Local – where customers can work with personal shoppers, get shirts altered, pickup online orders, or get a manicure – all separate from the large department store nearby. And Target is experimenting with Target Open House – a specialty store for selling smart home deices.
D2C ecommerce players continue to move into showrooms with no signs of slowing down. Beyond fashion, eyecare companies such as Warby Parker and Clearly Contacts both have showrooms with try on, fittings, and eye doctors. Mattress companies like Endy are enabling customers to try before they buy. And makeup companies like Glossier are using showrooms to market their products can provide customized support.
Recommendation for contactless retail in high-traffic areas
Our recommendation emerges from a larger discussion of current day technology vs. the next best alternatives. More specifically, we explore the unit economics of a computer vision AI based retail (using Amazon Go as a model) and extrapolate those conclusions onto market trends.
Unit economics of computer vision AI retail
As a disclaimer before starting, our assumptions and conclusions are drawn from a limited dataset of available data from Amazon Go (presented earlier in this report), the current pioneer and large implementer of the technology. Additionally, rather than to try to analyze the full unit economics of an Amazon Go like store, we only look to understand the additional uplift over traditional convenience store. In other words, what do we need to believe about the market to justify the added CAPEX for the hardware and software solutions vs. traditional retail with a cashier and checkout lane.
Additional unit economics for Amazon Go vs. regular convenience [47]
To understand the benefit of this technology (and its associate cost), we focus on four metrics:
- Additional topline
- Reduction in shrink
- Reduction in labor (cashier)
- Additional CAPEX
Two assumptions drive the topline uplift: (i) foot traffic or persons/hr and (ii) average basket size. For this model, the average basket size is constant at $10/person. Instead, the uplift comes from being able to service more customers, more quickly without queues for check out. This is especially important for morning and/or lunchtime queues where lines are often hinderances.
An added benefit of the automated, computer vision-based AI is the reduction of shrink – largely from theft. Average retail shrink amounts to approximately 1.44% of retail revenues. If we eliminate shoplifting and employee theft that saves approximately 0.96% of revenues over the standard convenience store. [48]
The reduction in labor comes from no longer needing a manned checkout lane. The model supposes one a single manned cashier lane being replaced. We assume other personnel will still be needed, e.g. for stocking shelves or assisting customers.
Lastly, a significant amount of upfront CAPEX is needed to build out the camera, weight sensors, gates, and other infrastructure for the store to operate. Additionally, it is likely that some recurring investment will be needed annually to maintain the CAPEX – we estimate this to be about 10% per year. We are not including software development in these estimates – while this would be overhead, it would not affect the unit economics of a single store. Additionally, if this modeled is scaled outwards, the development costs begins to marginalize across the stores.
With the example modeled (store size of 1,500 sqft, operated for 14hrs/day for 350 days), the Amazon Go looks to be a very attractive investment. What is absolutely critical to these numbers, however, is the ability to drive the additional foot traffic to justify the investments.
Implications for recommendation
As shown with the unit economics, the recommended opportunity needs to have the appropriate foot traffic to justify the investment on CAPEX. More so, the traffic needs to justify it over the next best alternative. These include pre-existing and widely used technologies such as scan-and-go and self-checkout.
With this, our recommendation focuses on high-traffic retail. But more specifically, we are focused on high convenience built into everyday pathways. For example, this could be in a train station shortly after ticketing where passengers have to cross through to get to their trains, very similar to the mandatory duty-free paths at certain airports. The difference here will be contactless retail will enable complete impulse purchases in a single flow of traffic – everything from a beverage to a full pre-prepared meal. We will remove all barriers of purchase of the customer: inconvenient location, speed of purchase, lines and barriers.
References
[1] Green Street Advisors, Mall Insights Report, April 2019
[2] CNBC, “A third of America’s malls will disappear by next year, says ex-department store exec,” June 2020
[3] Footwear News, “These Customers Are Giving Shopping Malls a Fighting Chance,” October 2019
[4] Placer Labs, Inc., “The Demise of the Mall?” December 2019
[5] Placer Labs, Inc., December 2019
[6] Barron’s, “Shopping Malls Are Reopening, but Visits Are Still Way Down. Here Are the Numbers.” May 2020
[7] Placer Labs, Inc., December 2019
[8] Chain Store Age, “Top 10 Retail Center Experiences: No. 1 Westfield Century City,” July 2018
[9] Chain Store Age, July 2018
[10] L.A. Biz, ”Westfield Century City introduces ticketless parking,” August 2017
[11] Placer Labs, Inc., December 2019
[12] Westfield.com
[13] Pitchbook
[14] National Real Estate Investor, ” Mall Investment Sales Volume Is Down, But Investors Are More Willing to Buy Lower-Tier Assets,” October 2019
[15] Retail Wire, ”Is the future of retailing going dark?” June 2020
[16] Statista
[17] TechCrunch, ” US online grocery sales hit record $7.2 billion in June,” July 2020
[18] FoodIndustry.com, ”How many grocery stores are there in the United States?” November 2019
[19] DC Velocity, ”Takeoff Technologies partners with U.S.-based Hispanic grocer to launch robotic supermarket,” October 2018
[20] Grocery Dive, ”Sedano’s debuts new look at 35th Florida store,” November 2015
[21] Sun Sentinel, ”Robots are helping Sedano’s Supermarket gather your groceries faster,” February 2019
[22]The Robot Report, ”Takeoff Technologies CEO discusses developing grocery microfulfillment market,” July 2020
[23]The Robot Report, July 2020
[24] Forbes, ”The Future Of A Profitable Online Grocery Service Is Here,” May 2020
[25] Forbes, May 2020
[26] IGD, ”Responding to ecommerce growth: a conversation with Takeoff Technologies,” May 2020
[27] FMI, ”The Small, But Mighty, Micro-fulfillment Center, August 2019
[28] Investable Universe, ”Micro-Fulfillment: Venture Cap’s Big Warehouse Play Is Playing Small,” October 2019
[29] DSN, ” Retailers embrace micro-fulfillment centers,” August 2019
[30] Medium, ”Micro-Fulfillment: Venture Cap’s Big Warehouse Play Is Playing Small,” October 2019
[31] Forbes, ”New York Proves Amazon Go Works, And An Even Bigger Rollout Is Only A Matter Of Time,” June 2019
[32] Supermarket News, ”Amazon Go makes New York City premiere,” May 2019
[33] PR Newswire, ”Record U.S. Convenience Store Food, Merchandise Sales in 2018,” April 2019
[34] Statista
[35] Research and Markets, ”Global Contactless Payment Market Worth $18+ Billion by 2025 – The Impact of Blockchain, 5G and Augmented Reality on Contactless Payments,” March 2020
[36] Forbes, ”A 7-Eleven Where You Can Work Out? Why Retailers May Want To Study Taiwan’s Convenience Stores,” February 2019
[37] Canadian Grocer, ”Check this out: 7-Eleven tests unmanned store in Taiwan,” February 2018
[38] Forbes, February 2019
[39] RIS, ”Investments in Home Delivery & Contactless POS Will Continue Post-Coronavirus: RIS Poll,” April 2020
[40] NRF, ”COVID-19 boosts the prospect of contactless commerce,” May 2020
[41] Supermarket News, February 2020
[42] TechCrunch, ”Amazon is now selling its cashierless store technology to other retailers,” March 2020
[44] https://www.nytimes.com/2017/11/17/business/retailers-showrooms.html
[45] https://www.retaildive.com/news/why-retailers-are-trying-on-showrooms/439990/
[46] https://knowledge.wharton.upenn.edu/article/digital-retailers-opening-physical-stores/
[47] https://www.vox.com/2019/1/4/18166934/amazon-go-stores-revenue-estimates-cashierless