Innovation diffusion theory suggests that consumers differ concerning the number of contacts they have, the degree and the direction to which social influences determine their choice to adopt. To test the impacts of these factors on innovation diffusion, in particular the occurrence of hits and flops, a new agent based model for innovation diffusion is introduced. This model departs from existing percolation models by using more realistic agents (both individual preferences and social influence) and more realistic networks (scale-free with cost constraints). Furthermore, it allows consumers to weight the links they have and it allows links to be directional. In this way this agent based model tests the effect of VIPs who can have a relatively large impact on many consumers. Results indicate that markets with high social influence are more uncertain concerning the final success of the innovation and that it is more difficult for the innovation to take-off. In addition, the article shows under what conditions highly connected agents (VIPs) determine the final diffusion of the innovation.