Measurement is critical because the use of meaningless numbers will result in meaningless conclusions, no matter how complex the statistical analyses. Ultimately, there is no substitutes for high-quality measurement methods. In selection, if the number signifiesdegrees of differences in units that reflect nearly equal interval units, not just more than or less than (ordinal), then you have a quality predictor orcriterion.
Keep in mind that such a conclusion is grounded uponthe assumption that these numbers can be meaningfully interpreted as representing the unobservable construct underlying the number. If this is true, you can reasonably subject these numbers to useful interval scale statistical analyses (eg. use of parametric stats including means, standard deviations, person product moment correlations and F-statistics)