César Augusto Henao

Multiskilling is a workforce flexibility strategy that enables employees to perform multiple tasks effectively. This paper addresses a personnel training problem by incorporating multiskilling through a k-chaining policy with k≥2, while explicitly considering three key employee training preferences. A two-stage stochastic optimization model is proposed to minimize total training costs and the expected costs associated with overstaffing and understaffing, while incorporating penalties and rewards linked to employee preferences. In this way, the model simultaneously seeks to reduce employer costs and enhance employee satisfaction. The proposed methodology is evaluated through a real-world retail case study using processed and simulated data from a Chilean store. Four experiments and nine levels of demand variability are analyzed. The results show that incorporating employee preferences, either partially or fully, does not significantly affect employer costs or overall multiskilling requirements, while substantially improving employee satisfaction. This suggests that employer and employee objectives are not necessarily conflicting, enabling the design of training plans that benefit both parties. Furthermore, sensitivity analyses on key preference parameters indicate that, although these parameters influence workforce composition, their impact on costs and multiskilling levels remains limited, confirming the stability of the proposed solution.

In this data article, we present and describe datasets designed to address multiskilled personnel assignment problems (MPAP) under uncertain demand. The data article introduces simulated datasets and a real dataset obtained from a retail store in Chile. The real dataset provides details on the structure of the store, including the number of departments and workers, the type of labor contract, the cost parameter values, and the average demand across all store departments. The simulated datasets, consisting of 18 categorized text files, were generated through Monte Carlo simulation to encapsulate information about the stochastic demand for store departments. These text files are classified based on: (i) type of sample (in-sample or out-of-sample), (ii) type of truncation method (zero-truncated or percentile-truncated), and (iii) demand coefficient of variation (5%, 10%, 20%, 30%, 40%, 50%). This categorization allows academics and practitioners to select the scenarios that meet with their specific research or application needs, increasing the flexibility and applicability of the datasets. In addition, researchers and practitioners can use these comprehensive real and simulated datasets to benchmark the performance of diverse optimization methods under uncertain demand, thereby ensuring robust multiskilling levels for similar MPAPs. Furthermore, we offer an Excel workbook with the capability to generate up to 10,000 demand scenarios for varying coefficients of variation in demand.