There's three issues.
The first, as pointed out; you're targeting innocent people due to solely superficial characteristics. You're saying that black guy looks like a criminal
because he's black. That's outrageously racist. That's the really obvious one.
The second, which isn't as obvious, is that the entire
idea of profiling is absolute hokum. The majority of whichever group you're looking at are innocent of any wrongdoing. And
no group is fully 100% free from offenders. Meaning you're making a judgement call that a certain rate of offense is "okay", and it's higher rates than that which are "bad" and you'll target those groups. That is an entirely
subjective line to draw, and in pretty much every single case, it's not drawn for any objective statistical basis, but merely to justify the outstanding racism.
The third issue, which is entirely damning of the entire concept, is that it ignores the population levels in question. Going by
this FBI data, ~2,600,000 African-Americas were arrested for something in 2011. However, ~6,500,000 Caucasian-Americans were arrested, about 2.5 times as many. Yes, there are significantly more Caucasian-Americans than African-Americans, population-wise, and the rate of offense is higher for the latter, but the
greater proportion of actual criminals is still white dudes. So if you're profiling African-Americans due to the
rate, you're missing out on the majority of
actual criminals. Which means profiling is failing to actually accomplish its purpose.