The Stockholm congestion charging trial was implemented with the aim of satisfying a number of transport policy objectives related to reducing car use in the central area of Stockholm. However, the multifaceted nature of congestion that the charges aimed to alleviate was not technically defined. The traffic impact evaluation efforts were based on survey results for chosen links without any scientific review of the dynamic character of congestion or any investigation of the statistical characteristics of the chosen congestion indicator. Neither was there any analysis of area networks nor any calculation of an overall indicator. Based on a literature review, approaches for defining congestion and a number of alternative congestion performance measures were selected. Data from the trial ‘before' and ‘after' surveys were used to estimate congestion levels and the accuracy of the performance measures. Specific issues and lessons regarding the quality of the data collection are discussed. Accuracy rankings are presented for different cases, concluding that some performance measures showed better reliability in all situations whereas others proved to always be less reliable. These findings are used as a basis for recommendations regarding the implementations of the congestion indicators studied.