I review many data models throughout the year using the Data Model Scorecard technique, and over 95% of the entity and attribute definitions reviewed lack precision. There are certain terms in a definition that, if not explained, automatically make the definition imprecise and vague. The terms I look for include:
- “Generally”
- “Sometimes”
- “Normally”
- “Most of the time”
- “With few exception”
What terms in a definition do you look for that scream ambiguity?
Related Articles
5 Comments
-
Other – as a code/value in a list. I know sometimes you don’t know, but in a list of finite values “Other” is infuriating. I’ve even seen it in binary Y/N lists.
-
“Other” is frustrating but sometimes necessary: if your have DQ issues in a Fact table (blanks) but still need to include those rows and not automatically filter them out, you may need to transform the data to “Unknown”, “Not Stated”, or “Other” (depending on what your business user prefers). The decode should explicitly state “Does not contain a valid [Category entry]”; e.g. “Does not contain a valid State code”.
BTW, if you find that the number of “unknown” entries in a report is creeping up to 30%+, there’s a good chance that no-one is using this report… (true story!)
As for binary lists, I’ve seen numerous examples where the person entering data simply does not know if an answer is Y or N. If the business users decide that this is not a mandatory field at data entry time, they may relax the rule and allow blanks.
For example, their business process may be to pre-fill a questionnaire, then review the answers with a customer, at which point the blanks may be filled in. (Or they may not!) If this process occurs over several days, your data warehouse (both relational and downstream Fact tables) may inherit blanks in a supposed binary field – which you then have to deal with.
At least if this is built-in, you know what you don’t know – the “known unknowns”. (Thanks, Donald!) And your business users know what information they need to follow up on.
-
-
Always and never. Although they should scream confidence, I find that they shout the definers over-confidence.
Cheers, -
“Ambiguity in Definitions” is unavoidable once we understand a bit of how Humans interact with the ‘data’ around them, or the underlying way the human cognitive process works. Some data available is less or more ambiguous than the other and we need to manage them both in context, with different methodology and models. Also, for a more complete definition of ‘Definition’, we can check with wiki at https://en.wikipedia.org/wiki/Definition.
-
Those are all called “weasel words”. As in they allow people to weasel out of responsibility. 😉