What’s your average out-of-stock (OOS) rate?
For many retailers, the number hovers around 8%, with estimates of up to 15% on promoted items. Knowing how much product is out of stock when a customer places an order is important to ensure that customers get what they need on time and will be satisfied with the shopping experience.
Industry research shows that retailers carry up to $1 trillion in losses linked to low on-shelf availability when people cannot find the products they need. Since COVID-19 hit, the situation with OOS rates became even worse. During the pandemic, uncertainty and panic shopping contributed to shortages in all product categories. A total OOS rate in bath tissue and paper towels alone accounted for a $1.5bn missed opportunity in U.S. supermarkets over nine months. Inventory distortion in 2020 caused losses of up to $670m for manufacturers and $580m for retailers worldwide.
In this blog post, I’ll share my ideas about how retailers can cut costs related to high OOS rates with the help of computer vision and machine learning. You’ll find out how retailers apply the technology and what benefits they receive.
Inventory management system (IMS) is about retail supply meeting customer demand without tying up too much stock that could go to waste in warehouses, or conversely, running out of stock.
On-shelf availability (OSA) in inventory management systems considers a bunch of factors that include, but are not limited to:
OSA implies that a product is visibly accessible on a shelf for customers at the right time in the right store. It’s the percentage of products available from the full potential of stock. OSA mismanagement brings losses to all parties involved: customers leave stores or switch to other retailers, and stores, in turn, lose fragile customer loyalty and sales in the long run.
In the early stages of the pandemic, demand fluctuation, with consumers stockpiling essential products and shifting to larger sizes of stock-keeping units (SKUs), along with shocks to the supply chain, led to an OSA drop from 92% to 60%, or even lower in some categories. OSA is a blind spot between inventory coming in and point of sale (POS). Industry leaders realize that after COVID-19 hit, OSA was no longer a constant variable. Retailers needed to evolve IMSs not only for offline purchases, but also for online retail, and to make sure pick-up options satisfy customers as those sales rise.
For years, computer vision has been transforming business solutions across industries, from self-driving cars to healthcare. Advances in this kind of artificial intelligence, which relies on machine learning (ML) models for object detection and recognition, have made computer vision competitive with human vision.
The global computer vision industry will increase fourfold over the next 10 years, seeing a growth in value from $9.45bn in 2020 to $41.11bn by 2030, according to Allied Market Research.
In OSA tracking, computer vision and ML challenge the status quo 8% OOS rate by monitoring and presenting opportunities through inventory visibility at any point in time.
So how exactly can retailers use computer vision to optimize OSA tracking?
Computer vision enables real-time data collection via images and video collected by drones, phones, robots, and fixed cameras on shelves, in stores, and in warehouses. Computer vision-enabled software tracks stock, detects damaged or mislabeled items, analyzes and forecasts peak- and off-peak demand for particular items and stores, and even orders product from suppliers.
CV technology has significantly matured over the past few years, thanks to the appearance of higher-quality cameras and sophisticated ML programs that rely on huge amounts of data to train computers to recognize patterns in a host of datasets and make inferences.
“[Computer vision] is a science that has existed for decades, but it’s exciting now because of deep learning. When you have a visual problem or an auditory problem or speech, deep learning has automated a lot of these processes,” said Tuong Nguyen, an analyst at Gartner.
Integrating computer vision into the OSA workflow gives retailers the ability to track information about product availability in real time and react accordingly, offering a competitive advantage. Industry research shows that for manufacturers, a 3% increase in OSA leads to a 1% increase in sales, and for retailers, a 2% increase in OSA leads to a 1% increase in sales.
Let’s take a look at some successful cases of retailers and consumer goods companies integrating computer vision into OSA tracking systems.
One of Europe’s biggest grocery retailers, Auchan deployed computer vision-enabled robots for shelf-monitoring in 34 of its stores in Portugal. The robots, which use computer vision and Internet of Things (IoT) technology, captured pictures of every shelf in each aisle of a store three times a day. Computers then digitized the data to generate reports with actionable metrics and insights. These reports were delivered via mobile app to managers, who used them to prioritize and resolve issues such as out-of-stock items and pricing compliance more efficiently than if the data were gathered manually.
In 2019, the online retailer Amazon changed their product tracking process by adding computer vision in 20 of its 175 warehouses. Instead of humans doing the work, computer vision-powered cameras and scanners read barcodes to track all inventory movement, including which bins inventory would land in. This technology had a significant impact on Amazon warehouse efficiency, according to Brad Porter, the company’s vice-president of robotics.
Amazon can be rightly considered a leader among retailers integrating computer vision into its workflows. And it goes beyond warehouses. The retail giant also uses delivery drones, internet-connected doorbells, and even an AI-powered fashion tool, all of which rely on technology that lets computers see and understand the world around them.
The consumer goods company Unilever also turned to computer vision to audit stores. It deployed a crowd-based workforce to a number of European stores selling its goods to grab pictures of store shelves. The pictures were then processed by a computer vision-enabled platform. The company’s goal was to provide its managers with an accurate snapshot of in-store shelf conditions. The analytics and insights received from shelf images analyzed with CV allowed Unilever to spot anomalies in store, respond accordingly, and measure results. Using technology freed up merchandisers and sales rep resources, allowing them to focus on higher value tasks, like expanding store coverage.
Computer vision-enabled software for tracking availability of products can help retailers eliminate the following challenges:
Stockouts can be linked to inaccurate records and poor forecasting that affect product availability. ML can detect early signs of increased demand and alert managers and other members of the supply chain when they need to purchase more product.
Overstock results from excess inventory due to inaccurate forecasting. Overstock increases storage costs and risks products going bad or becoming obsolete. This problem can be eradicated by deploying an ML model that relies on historical sales data to forecast demand for each product.
Misplaced inventory on store shelves or in warehouses leads to wasted time searching for products, and hence, increases business costs and causes delivery delays to customers. The current solution without the involvement of ML does not provide quality information on stock locations and mostly relies on placing goods manually. However, more and more warehouses are implementing barcode scanning technology and robots to locate misplaced products.
Of course, no technology is perfect until it matures and passes through many rounds of testing across several iterations. However, a meticulously-crafted ML pipeline that automates OSA-related tasks can reduce the number of human errors in the process to near zero.
Regarding the benefits of integrating computer vision-enabled solutions into OSA monitoring, the first thing that comes to mind is that it can optimize the process of providing retailers with real-time insights. This is extremely important in terms of keeping customer satisfaction rates high.
Another benefit is the potential to increase productivity by reducing manual labor, providing real-time shelf data for management, and managing inventory for online orders.
Moreover, computer vision allows for agile IMS with improved forecasting, accurate OSA for online ordering, and a quick notification system. Smart video analytics can also track and analyze customer behavior, helping improve sales strategies and boosting customer retention.
Using computer vision and ML also gives retailers opportunities to improve waste management in retail. For perishable products, where computer vision can help management know when to replenish displays just-in-time, preserving the quality and longevity of displays, rather than overstocking goods that can be damaged and lead to product loss.
Computer vision technology is constantly evolving, promising new and beneficial applications across industries.
At Postindustria, we offer custom AI development services based on computer vision and machine learning solutions.
If you’re looking for a technology partner for your computer vision projects, don’t hesitate to contact us. Our machine learning engineers will help bring your ideas to life.
Thank you for reaching out,
User!
Make sure to check for details.
