City maintenance is laborious and expensive. One of the main challenges is to monitor unpredictable situations like potholes, graffiti, broken footpaths, and so on. Once detected, government officials must be able to allocate the resources timely and prioritizing the issues most relevant for the citizens. The pervasive presence today of mobile internet connection in urban centers has enabled modern ways of interaction between the municipality and the population, resulting in the so-called Government 2.0. One of such way is crowdsourcing platforms, such as See-Click-Fix, FixMyStreet, CitySourced, OpenIDEO, and many others, which allow and stimulate collaborative par- ticipation by reporting urban issues. The importance of an issue can be endorsed via votes in the platform, meaning that issues with more votes represent the overall felling of the neighborhood that those should be solved first. This, in turn, constitutes important information to help organizing logistics, allocate resources, and fulfill citizen’s well-being feeling. The prob- lem is that may take time until collecting enough votes to be able to estimate the urgency of an issue. In this work we propose to estimate the number of votes an issue will receive using machine learning techniques. As the number of votes is a proxy to the urgency, we hope to improve city maintenance by providing in advance sensitive information.