More than 50% of all survey data in the Netherlands are collected via Internet. However, these data may not adequately represent the views of the Dutch people. The majority of the Dutch people are not willing to join a web panel, and from the people that are in a panel the minority (20%) fills out the majority (80%) of the questionnaires (NOPVO, 2006). Therefore, the answers obtained from web panels can differ significantly from the general population. It is well known that panels contain too many (heavy) Internet users and too few ethnic minorities. So how can we get people into a panel that would normally not join and (hopefully) make the results more reliable? An unconventional approach is used for building this panel: via social networks. Traditionally one could make the distinction between probability and volunteer opt-in panels. Although most survey researchers agree that probability panels are needed for representativeness, the majority of web surveys is based on volunteer opt-in panels because of budget restraints. Volunteer opt-in panels are prone to selection bias, however. This new way of recruitment may increase representativeness compared to volunteer opt-in panels (recruitment is on invitation only; respondent driven sampling can be used for difficult to reach groups) while keeping the costs at a minimum. By asking respondents via friends and relatives to join the panel, respondents that are normally not willing to join a panel might be persuaded to join. The starting point of building the panel are administrative records of Breda University of Applied Sciences in the Netherlands (about 7000 students with a national spread). I will investigate response quantity, response quality, and costs and give suggestions about when to use this type of recruitment. Note that the Internet penetration rate in the Netherlands is about 90% in 2010.