Retail Image Recognition getting Googly!!
Any retail business thrives on availability of SKUs (Stock Keeping Units) and accurate prediction of demand to match the supply. The cross talk of availability, demand and supply determine the success of the retail business and downstream for the FMCG companies. Irrespective of the complexity of supply chain, the buck stops at availability of SKUs in retail stores which will determine the purchasing behaviour of the consumers. There are ledgers and manual techniques to keep track of stocks in each retail store, but the advent of artificial intelligence and advancement in technology has brought the concept of image recognition into foray.
What does image recognition do and how does it help in maintaining availability of SKUs in retails stores?
Image recognition leverages sensors in cameras and data science techniques to identify the SKU present on the shelf and determine its availability. This dynamic nature helps enable a cross talk for forecast prediction at a store level on an ongoing basis influencing daily sales heavily. This has allowed FMCG companies to ensure that the right assortment of products is maintained at a store level to enable their sales to increase. Image recognition has been around for 5+ years with multiple companies providing the technology to aid in image recognition. Some of them like Vispera have been at the forefront for enabling image recognition in retail stores with much success. However, there have been 2 things that have constrained the concept of image recognition to be a gamechanger.
1. Uncertainty owing to external factors like lighting, colour, packaging changes and ambience of the SKU location. This has hampered the accuracy of image recognition leading to lower-than-expected adoption of image recognition among retailers and FMCG companies.
2. Absence of a major tech giant to embrace image recognition in retail space. This has resulted in many ideas being thrown on the wall, but the final validation would come when a tech giant ends up embracing this or puts its feet in the game.
January 13, 2023, came up with a news which will potentially redefine the image recognition in retail space as it is all set to become a Noogler. Google Cloud has announced that it has developed an AI tool to help retailers track inventory on their shelf i.e., stock availability in the retails stores. Its algorithm claims to identify the SKUs leveraging the cameras in the store with over 90% accuracy. This is a breakthrough moment as this will help transform image recognition in retail space and end up as a service which everyone can leverage amongst the retailers. This announcement and introduction of new tool indicates an important signal for the competitors, FMCG industry and retailers.
1. Ease of Access: With Google Cloud entering the space for image recognition, this could end up turning into a service on offer as a bundle with Google Cloud Platform. Instead of companies pondering on how to build this solution, the service may simplify their lives and save immense time and effort.
2. Cost becomes the king: Enabling this technology of image recognition can lead to companies incurring ~$1 million along with ongoing cost. The introduction of Google’s AI tool will lead to economies of scale as this could turn into a service leading to significant cost reduction for FMCG and retail companies to invest in other initiatives.
3. Consistency among competitors: There is no single source of truth for comparing stock availability or identifying the same across competitors. With Google’s AI tool, this will provide a consistent source of comparison with competitors and ensure that an apple-to-apple comparison is happening.
4. Potential advancement in data science models for image recognition: At present, CNN (Convolutional Neural Network) is the underlying technology for image recognition. Google’s foray into this vertical will lead to better data science models being introduced to have even more accurate recognition. A simple cue can be taken from how Microsoft’s introduction into data science led to the advent of Light GBM model which is commonly used across the world.
Image recognition is an interesting space and its Googliness will only make it a technology/product to watch out for as retail heads into a digitization era.
