Determining the causes of a failed eradication of a pest species is important because it enables an argued adjustment of the methodologies used and the improvement of the protocols for future attempts. We examined how molecular monitoring can help distinguish between the two main reasons for an eradication failure (i.e., survival of some individuals vs. recolonization after eradication). We investigated genetic variation in seven microsatellite loci in ship rat (Rattus rattus) populations from four islets off the Martinique coast (French Caribbean). In 1999 an eradication attempt was conducted on the four islets. Three years later rats were observed again on two of them. We compared the genetic signatures of the populations before and after the eradication attempt. On one of the islands, the new rat population was likely a subset of the pre-eradication population. A weak genetic differentiation was found between them, with almost no new alleles observed in the new population and moderate FST values (0.15). Moreover, assignment procedures clustered the two populations together. In contrast, on the other islet, many new alleles were observed after the eradication attempt, resulting in an increase in genetic diversity (from 2.57 to 3.57 mean number of alleles per locus) and strong FST values (0.39). Moreover, genetic clustering clearly separated the two samples (i.e., before and after the eradication attempt) in two different populations. Thus, to achieve long-term eradication on these islets, it seems necessary to redevelop the eradication procedure to avoid individuals surviving and to prevent reinvasion, probably from the mainland, by installing permanent trapping and poisoning devices and conducting regular monitoring. We strongly encourage wildlife managers conducting eradication campaigns to integrate molecular biological tools in their protocols, which can be done easily for most common invasive species.