Grocery retailers and CPG manufacturers are increasingly evaluating new solutions based on Computer Vision (CV) and Artificial Intelligence (AI) technologies – the former being a subcategory of AI that enables computers to process, analyze and interpret visual data – to help staff better manage various aspects of retail operations. From improving demand forecasting to online product recommendations to in-store shelf management and much more, these powerful new technologies are driving a digital transformation at retail.
It’s a much-needed development, especially at the retail shelf. As we’ve noted many times, in-store stockouts alone cost brands and retailers upwards of $2T every single year. Why does this problem exist? Because to keep shelves stocked, the industry largely relies on either woefully inaccurate algorithmic approaches to understanding shelf inventory based on perpetual inventory and point of sale data, or similarly inaccurate, time-consuming, and cumbersome manual spot checks or legacy inventory solutions based on image matching or reference.
Those methods have failed – spectacularly so. On average, consumers can’t find the product they are looking for between 10-30% of the time. And now, in a world where omnichannel grocery shopping has skyrocketed, there are more “opportunities” than ever to disappoint customers. Most retailers today have no clue if the item they sold to a shopper online is actually on the store shelf for an order picker to retrieve. The result? The average omnichannel picker spends triple the time searching for items that aren’t available on the shelf, driving up labor costs, resulting in unacceptable substitutions and creating unhappy customers in the process.
Computer Vision and AI to the rescue – but buyer be aware
Fortunately, Artificial Intelligence in its different manifestations can be applied to solve the problems that inaccurate shelf inventory creates across a wide breadth of retailer and CPG use cases.
But, when it comes to in-store data collection and analytics, it’s critical to understand that not all CV and AI technologies are created equal, and that innovation is proceeding apace.
Pensa’s automated retail shelf intelligence solution digitizes the shelf using bleeding edge CV that rapidly distinguishes between huge numbers of different products – the average U.S. retail grocer might have 100,000 distinct SKUs in-store – at extremely high levels of data accuracy.
Unlike other retail Computer Vision that takes snapshots of the shelf, Pensa’s CV ingests a video stream that’s equivalent to a hundred pictures of each shelf item from different angles and perspectives. Unlike other retail AI, Pensa’s AI is not dependent upon planograms or reference product images to understand the shelf. Instead, Pensa’s AI visually learns the shelf as it goes automatically and accurately identifying restocking patterns and actual shelf conditions at an item level. Across a total category including private label, Pensa determines if products are in the right place, running low or out of stock. Take a look at our solution in action. Our AI quickly translates shelf condition data into business insight that is being utilized across a breadth of sales, supply chain, store operations, merchandising, category management and omnichannel use cases.
AI is about the brains, not the brawn
Many of today’s headlines related to retail Computer Vision and AI highlight the visible data capture techniques that elicit a compelling vision of the future (and corresponding clicks to the article). By that I mean it can be easy to focus on floor robots, drones, shelf sensors, or shelf or ceiling cameras because those are the shiny objects that staff and shoppers see in the store. But the real key to useful retail CV (data capture) and AI (analysis to derive conclusions from the data) is the “brains” not the “brawn.” (For a deeper dive into the patented Computer Vision, Artificial Intelligence and Deep Learning that power Pensa’s solution here’s a great podcast)
We get it. When Pensa launched, we highlighted our use of fully autonomous aerial drones as one method of data capture. Admittedly, our drones are super cool. But practically, we’ve found our mobile scanning options (via handheld IOS and Android phones, tablets and Zebra devices) to be the more utilized method for data capture in today’s market. Unlike drones, or robots or shelf cameras, handheld scanning options integrate seamlessly into existing customer processes, don’t require any CAPEX, and require minimal operational infrastructure. (For the record, Pensa drones remain an attractive, frictionless deployment option, and our technology works on fixed cameras and floor robots.)
The point is, while the capture device is an important consideration, the CV and AI are doing the heavy lifting to accurately capture holistic shelf condition data and turn that data into useful business insight. And every retailer and brand are at a different stage in their in-store digital transformation strategy implementation. For these reasons and more, Pensa’s approach to data collection is to be customer-driven, capture device agnostic and to focus our innovation primarily on Computer Vision and AI.
The immediate path to value is human-aided automation
What we’ve learned from our customers is that the most expeditious path to value for retailers and CPG brands is human-aided automation. By placing our technology on mobile devices in the hands of humans already in the store aisles, we’re providing a solution that integrates capture technology into existing processes and systems.
There’s no debate that fully autonomous store inventory solutions promise huge value in terms of “no human in the loop” effectiveness and efficiency, but many practical considerations still need to be worked through before that promise becomes reality.
For a good analog, think self-driving cars. Fully autonomous vehicles at scale are still years away, but there is useful “autonomous car technology” that’s adding immediate value as humans continue to take the wheel, such as automatic braking or blind spot detection when systems perceive a potential accident.
Similarly, fully autonomous solutions at scale in retail will take time as they require different thinking in terms of capital investment and operating expense. But, human-enabled shelf data capture, today, helps retailers optimize labor and improve store operations via lightning fast and highly accurate scans. That means more time for higher-value customer service tasks, together with increased omnichannel sales and higher customer satisfaction due to fewer stockouts. Retailers’ CPG brand partners benefit as well as real-time shelf data helps them optimize planogram compliance, improve category management and ultimately increase sales.
There’s currently a lot of noise in the market about retail Computer Vision and AI and corresponding data capture techniques like robots and shelf cameras. To make decisions that deliver the greatest business impact it’s important to separate the data capture technique from the data capture technology, and to understand that all retail Computer Vision and AI is not created equal. Want to learn more? Reach out. Pensa can help.