The increased translocation of plant species for biodiversity restoration and habitat creation has provoked a debate on provenance and genotypic diversity of the used plant material. Nonlocal provenances are often not adapted to the local environmental conditions, and low population genotypic diversity may result in genetic bottlenecks hampering successful establishment. We tested provenance differentiation of four plant species used in agri-environment schemes to increase biodiversity of agricultural landscapes (wildflower strips). Provenances were collected close to the experimental field and at four further sites of different distances ranging from 120 to 900 km. In two of these provenances, different levels of genotypic diversity were simulated by sowing seed from a high and low number of mother plants. We found a large provenance differentiation in fitness-related traits, particularly in seedling emergence. There was no evidence for a general superiority of the local population. The productivity was greater in populations of high genotypic diversity than in those of low diversity, but the effect was only significant in one species. Productivity was also more constant among populations of high diversity, reducing the risk of establishment failure. Our results indicate that the choice of an appropriate provenance and a sufficient genotypic diversity are important issues in ecological restoration. The use of local provenances does not always guarantee the best performance, but a spread of superior alien genotypes can be avoided. A sufficient genotypic diversity of the sown plants might be a biological insurance against fluctuations in ecosystem processes increasing the reliability of restoration measures.