A great deal of evaluations are commissioned and conducted every year in social work, but research reports a lack of use of the evaluation results. This may depend on how the evaluations are conducted, but it may also depend on how social workers use evaluation results. The aim of this thesis is to explore and analyse evaluation practice in social work from an empirical, normative, and constructive perspective. The objectives are partly to increase the understanding of how we can produce relevant and useful knowledge for social work using evaluation results and partly, to give concrete suggestions on improvements on how to conduct evaluations. The empirical data has been organised as four cases, which are evaluations of temporary programmes in social work. The source materials are documents and interviews. The results show that findings from evaluations of temporary programmes are sparingly used in social work. Evaluations seem to have unclear intentions with less relevance for learning and improvement. In contrast, the evaluators themselves are using the data for new purposes. These empirical findings are elaborated further by using the knowledge form phronesis, which can be translated into practical wisdom. The overall conclusion is that social work is in need of knowledge that social workers find relevant and useful in practice. In order to meet these needs, researchers and evaluators must broaden their knowledge view and begin to include practical knowledge instead of solely relying on scientific knowledge when conducting evaluations. Finally, a new evaluation model is suggested. It is called phronesis-based evaluation and is argued to have great potential to address and include professionals’ praxis-based knowledge. It advocates a view that takes social work’s dynamic context into serious consideration and acknowledges values and power as important components of the evaluation process.