Also, once failures are found, you could distinguish actual '?' values from Unicode values and get the percentage of Unicode characters with ...

SELECT t.val

, UnicodeStringLen = len(t.val)

, UnicodeNonQmarkChars = len(replace(t.val,'?',''))

, NonQmarkChars = len(replace(convert(varchar, t.val),'?',''))

, UnicodeChars = len(replace(t.val,'?',''))

- len(replace(convert(varchar, t.val),'?',''))

, PercentUnicodeChars =

(LEN(replace(t.val,'?','')) - len(replace(convert(varchar, t.val),'?','')))

* 100.0 / len(t.val)

FROM #TMP t

WHERE t.val <> CONVERT(varchar,t.val) --FAILURES ONLY