Image Recognition

Google Cloud’s Welcome Move To Evangelise Image Recognition for CPG Retail

Infilect and Google Cloud are partnering together to implement state-of-the-art Image Recognition for worldwide CPG Retail. The technology brings tremendous benefits to alleviate manual audits, replenish stocks in real-time, and optimize the retail supply chain.

By Vijay Gabale

Co-founder, Chief Product & Partnership Officer

Infilect Inc.

“A lack of timely, accurate information about shelf inventory is a major problem for retailers and it is so difficult to manage that it is an industry standard to just make guesses,” says Robert Hetu, VP analyst at Gartner Inc. 

The global retail market has reached a value of over USD 23 Trillion in 2022 and is expected to grow exponentially in the foreseeable future. The numbers are huge and the stakes are high. Out-of-stock costs over Trillion dollars every year to the worldwide retail industry. And therefore, relying on guesswork and accumulated data isn’t a viable option.

Right before NRF 2023, the retail industry’s biggest event, Google Cloud introduced a Computer-Vision based AI-powered technology to facilitate retailers to solve out of stocks in real-time. While the technology seems to be in the pilot phase, built on Google Cloud’s Vertex AI Vision, Google Cloud’s self-checking AI solution is intended to provide real-time insights to store associates on how well the shelves are stocked. The technology is touted to utilize cameras mounted on retail shelves. In my opinion, this is a great step forward for the retail ecosystem. This piqued interest by tech giants such as Google and Microsoft helps the entire ecosystem to reestablish the relevancy and urgency of the challenges faced by the retail ecosystem. 

How should a Retailer or a CPG manufacturer look at utilizing this advancement? As per WSJ, Graham Watkins (EVP of Supply-Chain Transformation at Giant Eagle where Google test-piloted its AI) has a first-hand take on the technology. Accordingly to him, the Google Cloud product has shown about 90% accuracy during early tests in a Giant Eagle innovation lab (the lab replicates store conditions). The supermarket chain mentions that it plans to begin piloting the technology in an actual store in the next several months, but a deployment across the full chain may take several years to come to fruition. “It could take three to six years for computer-vision shelf checking to become more mainstream”, WSJ quotes him.

This is where Infilect has been spearheading its R&D and vertical SaaS solution to make Image Recognition for retail a reality, today! With a team of over 30 Computer Vision AI engineers, Infilect has developed industry-first Image Recognition stack that works with over 95% accuracy to identify brands & SKUs from photos of retail shelves, identifies new SKU patterns in no time, and provides real-time insights within 60 seconds as soon as a photo or video is parsed. In fact, Infilect has been partnering with Google Cloud for the past 4 years to develop and deploy Image Recognition to over 400,000 unique stores worldwide. 

With the product, Infilect has mastered the art of collecting & processing photos & videos at scale. By partnering with Google, Infilect has built On-Device-AI that runs locally on Android and iOS devices to guide CPG Sales Reps or Store staff to rapidly capture in-store photos. This technology is available as part of an SDK and it is already embedded across numerous devices such as in-store robots and drones. 

In fact, Infilect AI runs completely on devices to identify SKUs as soon as a photo is captured. With the devices of the future such as Google Glass, imagine a store associate just walking down an aisle while the AI runs on her mobile phone to provide real-time insights on out-of-stock, overstocking, and promotion non-compliance. Now that is how AI is supposed to augment humans to optimize manual labor.

CPG manufacturers are already using Infilect to collect on-shelf data not just from one retailer but across banners, regions, stores, and channels. A fundamental difference to note here is the way in which a CPG manufacturer would utilize Image Recognition as compared to a retailer.

While a retailer is looking to solve out-of-stock only within its umbrella stores, a CPG manufacturer wants to understand shelf share, market share, competition, new product launches, promotion compliance, and numerous other Retail Execution metrics across multiple retailers, channels, and categories. 

In particular, how should a CPG manufacturer comprehend the buzz created by Google? How should a CPG manufacturer measure & fix in-store on-shelf retail execution metrics across a large number of banners and stores? Well, fret not. InfiViz already provides a turn-key solution to not just collect data from a large number of retailers but also do real-time Image Recognition to solve compliance problems within seconds. 

With ‘click-and-go’ audits, InfiViz solves numerous retail execution challenges by providing AI-driven sales recommendations, competitive pricing intelligence, and pricing & planogram compliance monitoring. By deploying InfiViz, CPG manufacturers are already able to skyrocket their sales figures by 2 to 5 percent per store. On the productivity spectrum of their reps and merchandisers, they are already seeing a substantial increase in their man-hours and a reduction in operational costs due to same-visit fixes using instant Next Best Actions.

On a final note, while Image Recognition in itself is not a new technology, it takes several years of Product, Tools, and R&D to develop a customized AI such as identifying SKUs from images. Several CPG companies have unnecessarily spent millions of dollars of budget on internal R&D to redo Image Recognition or reuse Google technology towards no outcomes. The problem of tracking SKUs is notoriously complex as over 10% of inventory on the shelf changes every quarter if not every month, and tracking new and old SKUs becomes a nightmare. 

Thankfully, specialized Image Recognition firms such as Infilect bring global best practices of tracking new SKUs on the same day, offering a turn-key and unified database of SKUs for a given category and making the solution affordable for just a few Thousands of dollars. Clearly, the industry needs a simple and cost-effective ERP system to measure and fix its in-store on-shelf visible inventory and products like InfiViz are leading this product category.

Having seen the evolution of AI from early 2010 to today, from my perspective, it is natural that whenever there’s an advancement in technology, there’s a layer of excitement that comes with it. But seeing past the period of attention and actually penetrating through the market requires a solid product-led technology foundation. By partnering with Google Cloud, Infilect is already making Image Recognition for retail a reality!