In 2016, the Grocery Manufacturers Association (GMA), along with leading food & beverage and consumer product companies launched Smart Label – the initiative to enable consumers to gain instant access to information (including ingredients, usage guidelines, health impacts, safety tips and the like) about thousands of products on retail shelves. The vision was that a simple scan of a bar code or an online search, as indicated by the smart label, would lead to everything there is to know about a consumer product.
More than 30 major companies committed to participate and comply with the guidelines of the Smart Label program. For them, this means:
For the company, the process of aggregating and re-recording product information means making costly investments in several hundred man-hours of effort. And the output is likely to be of questionable quality – because human error is inevitable in records created by data processes that are managed manually. The real problem is to find a solution that mitigates this human effort while also guaranteeing complete accuracy and traceability of records.
The answer lies in intelligent automation. Leverage a platform that has Optical Character Recognition (OCR) capabilities to “read” the product labels and create accurate digital records. Then, develop algorithms to categorize the data meaningfully. The right workflow tools can streamline the whole process with minimal human intervention. See how the platform is a simple solution to the complex and labor-intensive process of managing Smart Label compliance:
Structured Data from Images: The system uses OCR capabilities to carefully scan the images of thousands of product labels to distill granular product attribute data. These include usage instructions, health and safety information, product benefits and claims, UPC numbers, ingredients, net content, expiration date, and the like, typically found on the labels. The system can zoom into a specific area of the label and also capture images of the labels from different angles. In instances where a particular image quality is poor, algorithms trained and perfected to arrive at data sourced from multi-directional images of the labels help to zero in on product information. This data is stored and managed in a highly scalable data lake.
Categorizing the Data Meaningfully: Machine learning can be easily leveraged to classify this data into relevant categories. We can build and train algorithms to automatically categorize the data as required by the Smart Label template, so uploading it to the gateway site, after due diligence, becomes a one-time, error-free process.
Automating the Workflow: Once the information is extracted and categorized for a particular SKU, use a workflow tool to share a sample set with the reviewer (an expert from the company who can certify the accuracy of the information). Should the reviewer find an error, the process of categorization is automatically redone. The algorithms continuously learn from these process re-runs to improve and perfect the categorization, so the same errors are not repeated, and the workflow progressively becomes error-free