Over the last decade the internet has been the source of a wide variety of democratic innovations, from the creation and maintenance of political forums and discussion sites to the establishment of e-petitioning platforms, voting advice applications, and social media enabled activist organisations and campaigns. One well known characteristic of these online participatory services is that contribution levels typically exhibit a highly skewed distribution, whereby the majority of people who make use of the service contribute only a little, whilst a small minority, often known as contribute a lot. These exceptional activity levels mean these individuals are extraordinarily important in terms of keeping a given community alive; it also means they wield disproportionate influence within it.Despite their importance and potential impact, these power users have received scant attention in the literature on online democratic participation. This article aims to fill this gap by addressing two key questions. First, we ask what explains why some users become power users whilst others do not. Second, we examine the impact of power users on site outcomes. Based on a unique observational dataset of hundreds of thousands of users of an electronic petitioning platform, we show that classical participation resources such as wealth and education appear to make little difference in terms of generating power users. By contrast having more time makes a real difference: with the unemployed particularly likely to become power users. We also show that initial experiences with the website make a lasting impression on users and influence their entire future participation in the site. Finally, we show that power users are both more effective than regular users, and also have interests which differ substantially from the rest of the crowd: meaning that these small groups have a powerful (and distorting) influence on overall outcomes for the site.