A 'good' correlation is data-context driven. For example, I did a project correlating sales volume per state and customer demographics (pretty 'hard' data) and got a .92 correlation. In customer satisfaction work getting a correlation of .5 relating overall satisfaction (or repurchase intent) with a specific attribute like 'timeliness of delivery' (soft data) is considered a high value. I did a study correlating net profit per location with expenditures on an in-house promotion, and got a correlation of about .15, and that was received very well. If you are following up your correlation analysis with, say, factor analysis of cluster analysis, then a .6 or better correlation makes these subsequent analyses more valid.