Targeted　marketing　is　a　low　cost,　low　risk,　and　profit-driven　strategy.　It　aims　to　recommend　right　products　to　right　customers　or　select　right　customers　for　right　products.　It　is　an　important　area　of　application　for　data　mining　and　Web　intelligence.　It　typically　involves　two　components,　customers　and　products.　This　paper　employs　a　probabilistic　inference　model　to　model　the　relationships　between　customers　and　products.　The　relationships　are　interpreted　by　precision-oriented　and　recall-oriented　interpretations　in　this　model.　According　to　available　different　information,　naive　bayes　and　the　latent　semantic　model　could　be　used　to　measure　the　relationships.　As　an　illustrative　example,　this　paper　present　inductive　methods　to　show　how　to　employs　the　probabilistic　inference　model　to　measure　customer　preference.　Experiments　on　real　world　data　show　that　the　proposed　methods　are　effective.　©　2008　Springer-Verlag　Berlin　Heidelberg.