Dikes are crucial for the protection against floods. One of the ways in which dikes can fail is by the instability of the inner slope. Credible probabilities of failure for slope stability are essential for the safety assessment of existing dikes and the design of dike reinforcements. This dissertation focuses on improving failure probability estimates for slope stability by using observed behaviour and performance of dikes. Examples of performance information are survived loads such as flood water levels or proof loads, and measurements during such loading conditions. The research uses Bayesian analysis to account for one or more performance observations or measurements in estimating failure probabilities. Using multiple case studies, this research identifies the observations and success factors leading to significantly lower failure probabilities. Furthermore, Bayesian decision analysis was used to consider the cost-effectiveness (Value of Information) of performance information, to determine which strategy of dike reinforcement and/or uncertainty reduction leads to the lowest overall cost to comply with a given safety level. It is found that incorporating multiple data of the behaviour and performance of dikes improves estimates of the failure probability for slope stability, leading to better safetyassessments and more efficient design of dike reinforcements. The cases considered in this dissertation suggest that savings of several million euros per kilometre dike reinforcement are possible (10-35% compared to the current dike reinforcement costs), for the Dutch situation with typically relatively high cost of dike reinforcements compared to the costs and risks of obtaining performance information. The use of performance information therefore contributes to improving the efficiency of managing flood risk in the Netherlands, and in particular the dike reinforcements in the Dutch Flood Protection Programme.