One of the most important tasks for brand manufacturers' sales representatives is getting an idea of the situation on their outlets' shelves. During visits to outlets – for example, to a supermarket – the employee records important information on their laptop regarding the distribution of products, their placement on the shelf, possible out-of-stock situations or information about competing products. This information is transferred to a backend for analysis and optimisation purposes and is then sent to Key Account and Category Management. In the systems currently used for field sales management, the employee uses a matrix with a sort sequence defined by the office staff, which shows the items and results of the last survey. Category Management provides a planogram, i.e. a visual representation of how the articles should be placed on the shelf. The sales representative walks along the shelf, checks the current situation there and manually records changes that have occurred since their last visit.
The challenges faced by sales representatives may sound trivial at first, but they hinder efficient data collection and often take longer to complete than was scheduled for the outlet visit: The set sorting order of the products by Category Management rarely corresponds to the running direction, meaning that the survey is impeded by frequent scrolling and searching in the planogram. Entering information via a laptop keyboard is also a problem without a stable surface on which to place it. In addition, it is not possible to directly evaluate detected changes on site. As the employee has little time available for the shelf survey during their visit to the supermarket, a more efficient recording method would be a great aid. Therefore, how would the situation improve if an AI were to support the sales representative during the shelf survey?
With the help of an application developed by 4brands Reply and Go Reply "AI Store Check" allows sales representatives to automatically perform surveys – all the employee needs to do is create photos of the shelves with the app and upload them. This eliminates complicated hardware handling as well as the cumbersome scrolling to find the products in the software.
The survey results are compared in real time with stored target values according to the contractual conditions for the corresponding outlet. The app then presents the sales representative with a complete inventory and possible deviations from the target.
A clearly-laid-out frontend allows the employee to create a new survey with just a few clicks. To subsequently carry out a shelf analysis, they only need take a photo. Next, an analysis is performed entirely on the Google Cloud Platform before the employee is presented with the finished analysis. The photos contain all detected objects with coloured markers.
Processing in the cloud allows extremely powerful machine learning models to be used for analysis. The photos taken by the sales representative are automatically processed by image detection and optical character recognition (OCR). The first step, image detection, concerns identifying all the objects in a photo, such as products, brands and price tags. In the second step, the price tags are read via OCR. The analysis makes it possible to determine where each product is located on the shelf, to compare this information with the price labels (using brand names and price) and finally to compare the overall situation with the target values and to check for correctness.
Once the survey has been completed, the employee immediately receives a comprehensive analysis overview, product-specific detailed analyses and the evaluated images on their tablet. This makes deviations from the target values immediately visible, which enables the sales representative to contact the store manager regarding any anomalies and to advise them on short-term optimisation measures.
All survey data is provided on the Google Cloud Platform and is available for further analysis. The day-to-day use of SAP Cloud for Customer can be linked to the app in order to conveniently document visits in the existing system. The long-term collection of survey data in the cloud enables easy decentralised access, better evaluation of the survey history and, at the same time, further refinement of forecasts.
A photograph is all that is needed for efficient shelf analysis; the corresponding analysis is visualised in a way that is easy to understand and is directly available to other employees from the cloud. In addition, the collected information is more reliable than that collected during the classic processes frequently used today. At the same time, this also makes backend analyses more reliable; these are carried out by Key Account and Category Managers on the basis of the recorded data. The measures initiated at short notice by the sales representative increase sales in the outlet and product category; they also provide more reliable figures in Category Management.