From Vision to Value : The development of a Conceptual Smart Checkout Solution

Sammanfattning

The retail sector is facing operational changes, driven by rising labor costs, staff shortages, and high employee turnover rates. While traditional self-checkout (SCO) systems have been widely implemented to maintain high service availability and reduce customer waiting times, they introduce certain barriers, including perceived system complexity, frequent scanning errors, and technical friction. Computer-vision (CV) technology and Machine Learning (ML) offer an alternative through barcode-less product recognition, commonly known as smart checkouts. However, a defined research gap remains regarding how to design and scale a vision-based product recognition SCO within the Swedish convenience retail landscape, where retailer specific preferences, assortments complexities, and regulatory frameworks apply differently than internationally. This Master’s thesis, conducted in collaboration with ITAB Shop Products AB, aimed to identify the primary challenges within self-checkout customer journeys focusing on less than ten item purchases. Furthermore, the study researched a possible retail environment for a Minimum Viable Product (MVP) deployment on the Swedish market. Moreover, it explored how a vision-based checkout solution can be physically designed to be user friendly and fit retailer needs. Following the Double Diamond design framework, a qualitative and empirical methodology was employed. Data collection involved thirty-nine field observations and informal customer interactions across high-density retail locations in Stockholm, Gothenburg, and Jönköping. This was supplemented by semi-structured interviews with retail stakeholders, and a CV and ML engineer. The qualitative data was processed using an inductive thematic analysis to translate field visit observations and interactions into actionable design problems. The thematic analysis points out that poor visibility, cluttered SCO layouts and perceived SCO complexity can directly contribute to checkout queues and SCO underutilization. Staff shortages and high staff turnover rates were two large themes, rising from interviews and staff interactions. Based on observed product assortment, operative environments, convenience retail formats, and current SCO adoption, cinema retail emerged as the most viable environment for potentially developing an MVP. The final design deliverable presents a robust, elegant, and minimalistic product architecture featuring a stationary arch structure over an item placement area. The arch integrates a downward-facing item illumination bar and plexiglass to cover three cameras. Two of them are angled at 40 degrees toward the center of the placement area, and the third camera is placed in the center of the arch, angled vertically downward. Furthermore, the concept incorporates an interaction light to signal checkout availability, an ongoing transaction, or a staff assistance requirement. To guide system interaction, the design presents a screen placed on the side of the placement area to keep the area clear and ensure accessibility. The proposed design also has an integrated backup barcode scanner in case the vision system fails to recognize unlearned or hidden items. This combined system is designed to enable reduced checkout times and minimized queue friction, especially during peak hours in time-sensitive environments.