|re: "This would just be a tool to help "flag" potentially erroneous data entry, which I currently do by hand, or eye as the case may be."|
It's probably a good idea that you automate this.
Coincidentally, on the way home from work today, I was listening to a report on the radio about a study that showed that if you are looking for something that is not supposed to be there, there's a decent chance that you won't find it when it is.
They talked about searching for things like signs of cancer on thousands of xray images or looking for guns or explosives in thousands of xrayed bags. Since the percentage of "bad" items is pretty small, and out the ordinary, there is a pretty good chance that you will miss some of them.
In one test, they put a fairly large sample of "bags with guns" in a stack of baggage images (something like 20%) and the participants missed about 7% of the guns.
They then increased the number of "clean images" so that the guns only made up about 5% of the images. The participants then missed over 30% of the guns.
The scientist who conducted the study said that you can't even talk your way into finding what you are looking for by saying "I'm going to look really, really hard and do a really good job." Even he, who knew the statistics, fell right into the middle of the pack when it came to missing the bad stuff.
Apparently the part of the brain that makes you miss things that aren't there most of the time doesn't listen to the part that says "I'm gonna find them, really I am!"