Flatness is a curve, not a number. If you try to distill it into a number (by integrating the error for example), you can have two speakers with the same "flatness number" which cost about the same, yet sound very different.

This is not theoretical, I've experienced it - two professional speakers (or monitors as they are called) from different manufacturers, which cost the same and the frequency plots look the same to a human eye if you put them side to side, but sound very different.

Which is why it's preferable to listen them if you really care. But there are "safe" recommendations for every price range.