Québec artificial insemination data from 518 609 first inseminations performed by 304 AI technicians in 15 488 herds using semen from 1750 dairy and beef AI bulls were analyzed using a mixed linear model. The effects of month of insemination, age of cow, semen price, breed of service sire, technician and herd were included in an evaluation of service sires for 60- to 90-d non-return rate. Herd was found to be the most important factor influencing service sire mixed model solutions with a standard deviation of solutions of 12.2% non-return rate. Technicians had moderate importance with a standard deviation of solutions of 3.7%. Higher fertility solutions resulted for summer months compared with winter months with the largest difference being 5.18% between December and September. Fertility solutions were highest for virgin heifers and decreased with increasing age of cow. Semen from more expensive bulls generally showed lower solutions than for lower-priced bulls. Dairy breeds, except Jerseys, had lower solutions compared with beef breeds. Service sire solutions were only moderately correlated to unadjusted non-return rates (r = 0.58) therefore indicating the importance of using a linear model approach particularly when several breeds of service sires are represented. Based on correlations among three measures of non-return rate, it was recommended to replace 60- to 90-d non-return rate in Canada by 56-d non-return rate. The mixed linear model procedure used in this study has been adopted by CIAQ and disseminated to other Canadian AI centres for implementation. Key words: Non-return rate, AI, cattle, bulls, technicians