The R2 is the coefficient of determination and tells you how well the prediction equation fits the data. It is the ratio of the variation due to the regression, to the total variation. Ideally R2 = 1.0 and all the points lie on the fitted line. In practice R2 lies between 0 and 1. Multiplied by 100 we have R2 expressed as a % or 'the percentatge of total variation explained'.
There is no objective means of assessing what value of R2 suggests a good fit other than "close to 1.0". If you need a rule of thumb, think of it as a confidence level. How confident do you want to be? 90%? Then the R2 value should be .90 or higher.
Hope that helps.
Anyone else have guidelines they use?