Street pavements are subject to various types of distress which necessitate a cost-effective management approach. This paper presents the outcomes of a survey focusing on street pavement maintenance and the utilization of machine learning (ML) pavement performance models on a 320 km municipal street network in Skellefteå municipality, Sweden. The findings reveal that the most common types of distress on Swedish streets include potholes, surface unevenness and alligator cracking, while prevalent causes of these distress are pavement ageing, heavy traffic and pavement patches. The windshield method of assessment of street pavement is prevalent, but the use of pavement management systems (PMS) is limited and pavement performance models are rarely employed. The case study reveals that Random Forest (RF) models developed for non-residential streets perform better than residential street models. RF models based on the variables age (A) and traffic (T) emerged as the best models, with 84% prediction accuracy. However, the R-squared value for the RF model applied to residential streets was 0.53, slightly surpassing the values for all models applied to non-residential streets (0.31, 0.50, 0.49). Further evaluation of models is suggested by using additional data.