Latent variable regression model (LVRM) inversion has been demonstrated to be a valid tool to support the design of new products and processes and for process control. One of the basic assumptions of LVRM inversion is that the set of desired characteristics specified for the inversion must adhere to the covariance structure of the historical data used to build the model. When developing a new product, it is not unusual that the desired product characteristics are assigned or allowed to vary slightly so as to satisfy the end-customer or the downstream processing requirements. If these values do not obey the covariance structure described by the LVRM, a mismatch between the model estimates obtained from the inversion solution and the desired product properties is observed. In this paper we address the above-mentioned issue: starting from a set product characteristics that does not comply with the LVRM structure, we propose a strategy to assist the selection of the new product quality profile, which is most suitable for LVRM inversion. Two approaches exploiting the desired values for the product characteristics and the LVRM parameters are proposed to reconstruct a new product profile in order to minimize the mismatch between model estimates and desired product properties. The feasibility of the proposed methodology is tested in a pharmaceutical product development case study, where the product is obtained through high-shear wet granulation.