Since 2011, 189,417 Montbéliarde females have been genotyped at the farmers’ request. Farmers are interested in getting GEBV of their young females for selection and for mating purposes. However, more genomic information than only GEBV could be use at mating, such as genomic inbreeding and carrier status for recessive haplotypes or major genes. This study aims at comparing five different on-herd mating strategies for their capacity to maximize expected genetic gain while limiting the expected average progeny inbreeding and the probability to conceive a foetus homozygous for any of three genetic defects segregating in Montbéliarde breed (MH1, MH2 and MTCP). In September 2018, 160 Montbéliarde herds were selected for which more than 80% of the females to be mated were genotyped and where the actual reproducing strategy was known, i.e. the farmer’s choice of females to be inseminated with sexed, conventional or beef breed semen. The females of each herd were then mated to 54 purebred top bulls using random mating method (RAND) or two optimized mating methods (LP = linear programming; SEQ = sequential) with or without use of genomic information (PEDI vs GENO). All mating methods were constrained for required type of semen use (sexed or conventional) and for ease at calving for heifers inseminated with conventional semen. A bull could be mated to a maximum of 10% of the females of a herd. In total, 9,655 matings were planned per method. LP_GENO was the best method since it provided the highest average expected genetic gain, the lowest expected progeny (genomic) inbreeding and the lowest risk of conception of a foetus homozygous for any genetic defect. LP_GENO provided on average +12% of economic gain compared to SEQ_PEDI and + 40% of economic gain compared to RAND for the farmer, showing the high value of planned matings compared to random mating. In addition, expected genomic inbreeding was >1% higher when using PEDI rather than GENO coancestry. These results are very promising for an on-herd application since the mate allocation step with an R software was fast (<10 seconds per herd).