2-8.5 Refine and Classify

Because data are typically collected from multiple sources, they must be reviewed to eliminate duplicates. To this end, the Postal Service rationalizes and classifies spending data elements and attributes.

Typical cleansing needs that should be noted include:

• Supplier names are spelled differently and thus treated as different suppliers (e.g., GE, G.E., Gen. Elec., and General Electric).
• Parent-subsidiary relationships are not captured, and the company is treated as different companies (e.g., Time Warner vs. America Online, Inc. vs. Netscape).
• Recent mergers or acquisitions among suppliers or company name changes are not captured and thus treated separately (e.g., Southern Bell Company vs. GTE).

Tasks that are evident in this process include compiling data into database tools, categorizing purchased items, identifying gaps and inconsistencies, and producing high-level spend baseline and priority listings.

Each system containing purchasing data likely uses different data structures and nomenclature by item, commodity, category, and supplier. As a result, after the data are refined, they should then be classified to remove errors and inconsistencies and to create a standard language for useful analysis.

Examples of classification schemes include:

• Aligning the dates of spend data.
• Converting to common currency.
• Assigning and/or correcting common category classifications and associated commodity names.
• Accounting for divested or discontinued businesses.
• Accounting for future capital expenditures and other investments.

This analysis requires consolidation of data from multiple sources and varied formats into structured and manageable database(s).