With food waste levels of about 30%, mostly caused by overproduction, reducing food waste poses an important challenge in the food service sector. As food is prepared in advance rather than on demand, there is a significant risk that meals or meal components remain uneaten. Flexible meal planning can promote the reuse of these leftovers and, thereby, reduce the sector’s environmental impact. We propose an innovative menu planning model, incorporating not only the traditional meal planning decisions, determining the meals offered per day and the ingredients to purchase, but also decisions related to the storage of surplus meals and their use on the following days. Demand uncertainty is a crucial aspect of this problem. Given that demand is generated based on human preferences, which are hard to model, we use a data-driven adjustable robust optimization approach and show how this can be solved efficiently. Our model then aims to provide a menu with the least environmental impact or cost while aligning with customer preferences. In this setting, we prove that demand uncertainty is the main driver of waste. By applying our model to a real-life case study of a food service provider, we demonstrate that reusing leftovers is a viable strategy for reducing purchasing costs by 4% and the environmental impacts of waste by 11-19%. Our findings suggest, furthermore, that food service providers should prioritize the production of the cheapest meals to minimize purchasing costs, while vegetarian/plant-based meals should be prioritized to reduce environmental impacts. Further analysis of additional case settings shows that increasing order frequency and making early purchases during the planning horizon allows for more flexibility in responding to fluctuating demand, thereby reducing waste. These findings offer actionable insights into promoting a circular (bio)economy in industries managing perishable products under uncertain demand.