100,000 may be too large a denominator for RANK to distinguish values, so you
might try 5,000 or 10,000 depending on your data - but you can figure that
out. Bernd's comment that RAND() does not guarantee uniqueness is correct,
of course, but 1) the likelihood of a problem is small and 2) you can include
a sum of your numbers and rerun if it's wrong. For example, with 100 numbers
the total of the ranked numbers should be 1 + 2 + 3 + ... + 100 = (N) (N +
1) / 2 = 5050. So if your total is not what it should be - hit "F9" to
recalc and try again.

"Victor Delta" wrote:

"dhstein" wrote in message
...
You can also add a small random number to each value -- "Your Number"
+
(Rand() /100000). As long as the denominator is large enough it should
have
no material effect on your data and ensure that each number is unique.

That's a clever idea. Thanks, I'll use it!

V