What instruments are used in hiring procedures and how they are weighted affects validity and diversity. Moreover, given the validity-diversity dilemma, maximizing validity or diversity results in different hiring procedures. Many studies investigate how selection systems should optimally be designed, but hardly any investigate how practitioners actually design selection systems, and why. Therefore, in a pre-registered exploratory study, we used a between-subjects experiment to investigate how n = 1212 people with hiring experience designed a hiring procedure, with the aim to: 1) maximize predictive validity, 2) improve the representation of female employees or 3) improve the representation of Black employees. Out of a list of nine common selection instruments, participants chose which instruments they wanted to include, and assigned an importance weight to each chosen instrument. First, compared to having a validity only goal, people who had a female representation goal included and weighted biodata somewhat more and cognitive ability and integrity tests somewhat less. People with a Black representation goal included and weighted biodata moderately more, and included and/or weighted cognitive ability tests, structured interviews, integrity tests, and emotional stability somewhat less. Some of these choices are in line with the empirical literature on adverse impact and predictive validity, but others are not. To estimate the effects of these different choices on the predictive validity and adverse impact resulting from the chosen selection procedures, the predictive validity and adverse impact for gender and ethnicity was estimated for each participant's selection procedure, based on an empirical correlation matrix. The average composite predictive validity and adverse impact was virtually identical between conditions. So, even though professionals seem to make some different choices when designing selection systems for validity or diversity goals, they were not effective; they did not result in practically relevant differences in adverse impact or predictive validity.