The number of increasingly fascinating, yet powerful applications of computer vision is making it one of the areas to watch in data science. With an estimated Compound Annual Growth Rate (CAGR) of 19.6% forecasted between 2023 - 2026 and a market size shy of $17 billion by the end of 2023 [1], the ability to track customer moods, analyze shelf space, and offer virtual fitting rooms is paving the way for a new era of retail.
Turning to the modern era of retail: a combination of carefully calculated visual merchandising and an obsession with spaghetti junctions of data. When thinking of futuristic retail, who comes to mind? Amazon? Walmart?
Let’s take Amazon Go: the trendy teen on the block curated by Amazon [2], keen to take their e-commerce empire to new heights by tapping into the $1.5 trillion (and increasing) grocery market [3]. Despite continuous macroeconomic and supply chain pressures, US grocery dollar sales growth is expected to hit shy of $1.74 trillion by 2027 [4]. Commercially, it’s an ice-lolly sweet and cool space when AI technologies are fused in.
Computer vision and the future of retail
Aside from the post-COVID-19 wave Digital transformation has had to ride, the future of retail has created healthy debates between the turbocharged state of E-Commerce vs. the future of the traditional high street store, sometimes referred to as the phygital experience [5]. Let’s take Amazon’s Go Store as an example: a chic convenience store on the outside but packed to the gills on the inside with Amazon’s own “just walk out” shopping technologies on the inside [6].
The diagram below summarises the customer experience:
In summary:
1. Customer walks in and scans items with their phone.
2. Sensors around the store detect items that have been both picked up and put down.
3. Customer completes shop and can simply walk out of the store: their cart total is charged to the Amazon account in their name and a receipt is emailed shortly after.
Although diagrammatically simple, Computer Vision technology of this nature requires an extensive number of considerations, for example:
- Having the correct number of cameras within the stores and knowing their exact location.
- The ability to identify the continual rotation of products on the shelf, any applicable sell-by dates, and when items have been sold or put back.
- Recognizing obstructions in the store, whether it be through lighting, objects obstructing/ partially obstructing items on the shelf, or people standing close together.
- Being able to identify subtle similarities in items, for example, differing flavors of drink.
- Identification of which customers took what items.
These are just some of the examples of the in-store challenges faced.
In addition, user experience testing of how the QR code should best be scanned is another, as detection of customers coming in and out, repeat custom - the list goes on and isn’t including the continuous stream of video being sent to the cloud combined with the amount of training data being processed. Heck, there could be several months' worth of training data being processed in a day. Serious stuff.
Computer vision: Future or a fad?
Despite the extensive number of considerations with a Cashierless store listed above, computer vision in retail isn’t a fad: With the correct use case it can provide significant insights traditional streams of data can’t capture: Crowd Analysis, for example, can provide insights into where customers are congregating in store, why they’re choosing that retailers product specifically and thus provide visual merchandising teams with new perspectives on how to display certain brands at key pillars in store.
In addition to the customer journey angle, the utilization of Computer Vision within Inventory Management allows retailers to identify gaps on shelves, and conduct stock audits to monitor and forecast pricing trends.
As a result, the technology can help retailers not only better understand their customer's buying habits, but work towards a more efficient set of operations, allowing the staff to focus on helping the customers instead of worrying about what’s not on the shelves.
Swish technology aside, looming over these technologies and use cases brings in the debate of privacy: Consumers are becoming increasingly savvy about where and whom they share their data with, so it’s a delicate balancing act of giving the consumer choice in sharing their data whilst giving them a personalized shopping experience and respecting data protection legislation.
For example, what proportion of customers would be happy using or being exposed to such technologies when they’re already having to share chunks of data? How Amazon, for example, will shape the future of the Go Stores with increasing regulatory pressure on all sides will result in many fascinating fireside conversations.
In addition, as policymakers and practitioners work to understand the increasing maze of ethical considerations within Generative AI, the more it’ll help carve a path in understanding what broader Data Science methodologies and tools really have the commercial longevity to keep the retailer a household name of the high street.
Conclusion
The future is going to see more incredible innovations within Emerging Technologies, but global macroeconomic conditions are still delicate, so taking steps to minimize choppy waters when consumer spending is still fragile is key.
Computer vision applications have the power to revolutionize entire industries with big bucks behind them, but conceptualizing something new with solid buy-in is challenging and architecturally roadmapping it to see a return on investment is another beast altogether.
Aside from the usual methods in road-mapping the potential for Computer Vision in a Retail environment, if you have a competitor utilizing the technology you’re thinking of deploying, go and test it with your own eyes and hands. Walking the journey as a customer can open your eyes to opportunities, pitfalls, and risks that can’t necessarily be observed through a glossy buzzword-filled brochure.
Bibliography
[1] Computer Vision Market Size: Market Research Reports & Consulting | GlobalData UK Ltd. (n.d.). Computer Vision Market Size, Share, Trends, and Analysis by Vertical (IT and Telecom, BFSI, Healthcare, Government, Retail, Manufacturing, and Energy and Utilities), Region and Segment Forecasts, 2023-2026. [online] Available at: https://www.globaldata.com/store/report/computer-vision-market-analysis/ [Accessed 6 Sep. 2023].
[2] US Grocery Market Size: Supermarket News. (2023). U.S. grocery retail to grow 6% this year. [online] Available at: https://www.supermarketnews.com/retail-financial/us grocery-retail-grow-6-year [Accessed 4 Sep. 2023].
[3] Amazon Estimated Revenue 2023: companiesmarketcap.com. (2022). Amazon (AMZN) - Revenue. [online] Available at: https://companiesmarketcap.com/amazon/ revenue/.
[4] US Grocery Market Size: Supermarket News. (2023). U.S. grocery retail to grow 6% this year. [online] Available at: https://www.supermarketnews.com/retail-financial/us grocery-retail-grow-6-year [Accessed 4 Sep. 2023].
[5] Phygital Retail: Hughes, M. (n.d.). The Future Of Retail Is Phygital. [online] Forbes. Available at: https://www.forbes.com/sites/mikehughes1/2022/10/31/the-future-of-retail-is phygital/.
[6] Amazon “Walk Out Technologies”: Amazon Web Services, Inc. (n.d.). Frictionless Checkout – Just Walk Out technology – Amazon Web Services. [online] Available at: https://aws.amazon.com/just-walk-out/.
[7] Amazon Go Store DeepDive: Gross, R. (2019). How the Amazon Go Store’s AI Works. [online] Towards Data Science. Available at: https://towardsdatascience.com/how the-amazon-go-store-works-a-deep-dive-3fde9d9939e9.
[8]: Future of Retail with AI: www.v7labs.com. (n.d.). 6 AI Applications Shaping the Future of Retail. [online] Available at: https://www.v7labs.com/blog/ai-in-retail.
[9] Amazon Store Closures: Malik, A. (2023). Amazon to close eight Amazon Go stores in Seattle, San Francisco and New York. [online] TechCrunch. Available at: https:// guce.techcrunch.com/copyConsent?sessionId=3_cc-session_a4fa1ae0- a8bb-48fc-8951-2fd5568171ef&lang=en-US [Accessed 11 Sep. 2023].
