I am not sure if this link is different from the one Asrujit posted, but here is the document sheet for Detecting Feature Change in 10.4 (with an option to view the instructions on 10.3 by clicking on "other versions"):
I found the same thing as Holly: some of the 2679 linear features I'm running through the tool have been flagged both as D and as N.
I thought I would familiarize myself with the tool by comparing a feature class in a file geodatabase to itself. To do this, I copied the original GDB and then pointed the Update Features to the first GDB and the Base Features to the second GDB and ran the tool with a Search Distance of .01m (our established standard tolerance). That's it, nothing else, none of the optional parameters filled in.
I would expect that the tool would return every record as 'NC' (No Change) because I haven't changed anything at all but...as Holly pointed out, some of the records are flagged as both 'N' (New) and 'D' (Deleted). An examination of the optional output table when I reran it shows that for many of the SRC_FID values, there are -1 values in the TGT_FID and further down, those exact same SRC_FID show up in the TGT_FID with a SRC_FID value of -1.
Why are these records not being matched? The GDBs are identical and all the records should be 'NC'. What am I missing?
Did you ever have any success with solving this? I am running into the same situation - I have a feature class with 67767 lines, and I'm comparing against a historical version of the same feature class, and I have about 15 records that show up in the results as a "D"elete and "N"ew feature instead of matching, despite having identical geometry and all attributes also identical. If I can't trust this tool to match all the records then I can't reliably use it at all.