PurposeIn recent years, the literature on program evaluation has examined multi‐stakeholder evaluation, but training evaluation models and practices have not generally taken this problem into account. The aim of this paper is to fill this gap.Design/methodology/approachThis study identifies intersections between methodologies and approaches of participatory evaluation, and techniques and evaluation tools typically used for training. The study focuses on understanding the evaluation needs of the stakeholder groups typically involved in training programs. A training program financed by the European Social Fund in Italy is studied, using both qualitative and quantitative methodologies (in‐depth interviews and survey research).FindingsThe findings are as follows: first, identification of evaluation dimensions not taken into account in the return on investment training evaluation model of training evaluation, but which are important for satisfying stakeholders' evaluation needs; second, identification of convergences/divergences between stakeholder groups' evaluation needs; and third, identification of latent variables and convergences/divergences in the attribution of importance to them among stakeholders groups.Research limitations/implicationsThe main limitations of the research are the following: first, the analysis was based on a single training program; second, the study focused only on the pre‐conditions for designing a stakeholder‐based evaluation plan; and third, the analysis considered the attribution of importance by the stakeholders without considering the development of consistent and reliable indicators.Practical implicationsThese results suggest that different stakeholder groups have different evaluation needs and, in operational terms are aware of the convergence and divergence between those needs.Originality/valueThe results of the research are useful in identifying: first, the evaluation elements that all stakeholder groups consider important; second, evaluation elements considered important by one or more stakeholder groups, but not by all of them; and third, latent variables which orient stakeholders groups in training evaluation.