One bank was getting examined to determine if it was engaging in unfair lending practices. The regulators pulled 6 loans at random from the bank's portfolio. 3 loans were to men, 3 to women. Then they looked at the average interest rate on the loans to the men, the 3 on the loans to women. They found a .5% difference between the male and female averages (the men had the lower rate).
The regulators fined the bank several thousand dollars and required it to adopt new processes to ensure that future loans were "fair."
Had the regulators looked closer, they would have noticed that the three random loans to men all went to older men with high credit scores...and each male was in a white collar job (doctor, lawyer, etc.). The loans to women, which were pulled at random, were each to younger women, just starting their careers. These were auto loans to a first time car buyer, etc.
The regulators didn't care about the details and punished the bank anyway. You are guilty until proven innocent, and they don't expend a lot of effort trying to prove you innocent.
In this day and age of big data, a bank or credit union's best bet is to use the power of business intelligence to continually prove that it is lending fairly in the community and climate that it dwells.