Spotfire Ideas Portal
Spotfire infers column types by examining the values in (the first 1000) rows of the data set being loaded. This generally works well but makes mistakes often enough that manual correction becomes laborious.
A case in point is when real-valued data are inferred to be currency data when in fact they should simple by reals. There are many other ideas describing this pain-point. Elsewhere it has been suggested that the currency data type be eliminated (or at least be ignored during the inferencing process).
Data type information is often present in the data set, whether it's an information link connected to a database (the table definitions specify the data types) or an MS-Excel spreadsheet in which cells have data types (formats). This information should be exploited to improve Spotfire data type inferencing. E.g. currency would only be inferred if a currency type were explicitly specified in the data source.
This is extremely annoying behavior in Spotfire, and should be easily fixed by incorporating header information (as suggested here) or just tweaking the algorithms to make Real more of a default and Currency more of a special case. Two years in, and it's only "under consideration"....