Can we just combine the dealers that have duplicated results? These are dealers from the pre-survey periods. However, by splitting them gave us absolutely no new information. The ones that are good stay good. The ones that are bad stay bad. By combining them gives us larger statistical base with larger sample size. The simpler it is, the easier we can handle.
The only exception may be A253 Lampang with total ownership change. The old score from the old owner can be dropped (but they are not that different yet anyway).