Overview

The previous four chapters discussed column properties, data structures, and data rules that provide a specific definition of what is valid and what is invalid. For "soft" data rules, the rule is very specific about what constitutes correct, whereas a violation may have been caused by someone making an exception to the rule. However, the data rule itself is a clear definition of what the data ought to be.

There are additional tests you can construct that point to the presence of inaccurate data that are not as precise in establishing a clear boundary between right and wrong. These are called value rules. You compute a value or values from the data and then use visual inspection to determine if the output is reasonable or not. You can easily distinguish between the extremes of reasonable and unreasonable but cannot be sure about the center, fuzzy area.

This type of analysis will surface gross cases of data inaccuracies. Surprisingly, this is possible even when all of the column values are valid values and there is no external reason to suspect a problem.