Politics With Classification

A common "business logic" problem that creates a high complexity tax is fiddling around with classifications for political reasons. By re-classifying things for different viewers (users), one can manipulate the view and/or classification of information in order to affect perception.

For example, one manager wanted two reports: a year-to-date (YTD) comparison to prior years' YTD, and a "last 3 months" comparison. For example, if it's the start of April 2008, then the YTD report would compare the total of Jan-2007 thru Mar-2007 to the total of Jan-2008 thru Mar-2008. The last-3-month report would compare Oct-2007 thru Dec-2007 to Jan-2008 thru Mar-2008.

The manager would select one or the other to present to outsiders in order to show the maximum improvement (which may be an increase or decrease, depending on the metric). The manager would look at both reports and select the one that painted the best picture to present. The end result from an IT perspective is that 2 kinds of reports had to be maintained. (It's even more involved than this. This is merely a simplification to keep the example concise.)

Another more public case is the time a prominent politician tried to get restaurant cooking re-classified from "service" to "manufacturing" in what appeared to be an attempt to make the manufacturing picture look better (offshoring of industry being the worry). One could argue for either classification, but the change request itself looked like numeric hanky-panky to many in the public.

http://www.cbsnews.com/stories/2004/02/20/politics/main601336.shtml

Discussion: This kind of historical reporting is actually quite common, even if there is no CherryPicking of comparisons. Many financial statements show both ReportYear?/ReportQuarter? vs. PriorQuarter? and ReportYear?/ReportQuarter? vs. PriorYear?/ReportQuarter? comparisons. Similarly, economists often use both ReportPeriod?/ActualData? vs. PriorPeriod?/ActualData? and ReportPeriod?/SeasonallyAdjusted?Data vs. PriorPeriod?/SeasonallyAdjusted?Data.