Multiskilling is a workforce flexibility strategy where companies educate workers to perform a set of task types effectively. When the multiskilling plans are structured using k-chaining policies, it is possible to obtain the maximum flexibility to match the uncertain workforce demand. This work evaluates the potential benefits of multiskilled workers using a k-chaining policy with k≥2, considering the learning/forgetting phenomena to model a heterogeneous workforce. We propose a deterministic mixed-integer programming model to compute the level of required multiskilling. The mathematical formulation determines how many workers should be single-skilled and multiskilled, which task types they should be trained in, the allocation of working hours, and the expected productivity of each worker on each week of the planning horizon. We test our methodology on a case study using real, processed, and simulated data from a Chilean retail store. We perform three experiments, comparing them: zero multiskilling, k-chaining with k≥2 and homogeneous workforce, and k-chaining with k≥2 and heterogeneous workforce. We consider nine variability levels in the workforce demand for each experiment. The results show that modeling the workforce as homogeneous leads to underestimating the multiskilling level required to minimize understaffing. Incorporating heterogeneous workforce modeling through the learning-forgetting phenomena suggests more multiskilling to compensate for the lower workers’ productivity. We consider this solution is closer to the actual operation of the store. We also perform a sensitivity analysis on the learning rate parameter to evaluate the stability of the report solutions for each variability level.