- This study provided an overview of the tasks that Dutch prisons have to perform. At many points laws and regulations clarify the tasks for which prisons are deemed responsible, although there is a lack of clarity at some points. Once a clear overview of the tasks can be provided, performance can get measured comprehensively. Performance indicators may then give a balanced view of the efforts made to achieve the organizational mission. When a set of performance indicators reflects (all elements of) the mission, the chance of perverse effects (like tunnel vision) may be smaller. That is important because perverse effects may interfere with accountability and performance improvement. To date, this condition of performance measurement is partly at hand in the Dutch prison system. Furthermore, a step-by-step method is provided to compare performance of organizations while accounting for the diverging ‘given’ circumstances. The exercise of ‘leveling the playing field’ has proved to be meaningful in the Dutch prison system. Performance comparisons therefore may lead to the matching of the right benchmark partners (superior performers can ‘teach’ and inferior performers can learn) to exchange experiences in order to improve performance. Next, accurate performance measurement ensures the comprehensive assessment of the sometimes complex and multifaceted goals that (semi-)public services strive for. In this study, we get to grips with the condition of having a measurement strategy: we assessed performance by including outcomes, mechanisms, and context measurements derived from several data collection methods. Together these measurements may comprehensively measure a specific (and multifaceted) task of Dutch prisons. After testing these considerations on empirical data concerning prison staff safety, it is concluded that performance phenomena should be modeled in a network and incrementally adjusted to reach a model in which the measurements are significantly but moderately related with the expected sign. In this way we may ensure that the measured elements are indeed related but also measure different parts of the same performance phenomena (so no element is overemphasized). Furthermore, the current study found evidence that performance measurements are related to factors a local prison management can influence. If that would not be the case, performance measurement may have a perverse impact on performance because managers cannot improve their performance scores and get demotivated. Overall, we found possibilities and developed methodological solutions to put the conditions of accurate performance measurement into practice in the Dutch prison system. Furthermore, eight respondents expressed in interviews that they expect the conditions described in this thesis to contribute to accountability and performance improvement. Simultaneously, these interviewees expect that such conditions may help prevent perverse effects of performance measurement as well. However, their assessment is that there should be ‘managerial conditions’ as well. For example, the national prison management and the local prison managers should uphold a non-punitive performance regimen and a managerial culture in which managers feel safe to ask for help to improve their performance.