In this study, we compared the early fish colonization of three types of artificial reefs deployed in the coastal waters of Saba and St Eustatius in the Caribbean: reef balls®, layered cakes and piles of locally obtained basaltic rock. As an indicator of performance, three fish assemblage parameters (abundance, biomass, species richness) were measured using underwater visual censuses at 11 months post-deployment and 4 months after restoration from hurricane damage. All artificial reef plots showed higher values for fish abundance, biomass and species richness than control plots covered by bare sand, which shows that artificial reefs can locally enhance the fish assemblage. However, the effect differed among artificial reef plots. Fish abundance was 3.8 times higher on the layered cake plots compared to the reef ball plots, while fish biomass was 4.6 times higher. Rock pile plots had intermediate values. Species richness did not differ significantly among different artificial reef plots. Three-dimensional modelling revealed that layered cakes had a smaller gross volume, shelter volume and total surface area than reef balls. The availability of multiple small shelters in the layered cake design appeared to be more relevant than other physical parameters, as the layered cake plots had higher fish abundance than the reef balls plots. We concluded that on Saba and St. Eustatius, layered cake plots performed better than reef ball plots after one year of colonization. Rock pile plots, made of local volcanic rock, showed an intermediate performance, and were 4–10 times cheaper to construct. If observed differences are consistent with other locations and persist during further colonization, current efforts to deploy reef balls could better be allocated to deploy artificial reef structures with a higher shelter density.