One process that increases the quality and usability of an organization’s mission-critical data is standardization. Any high-quality, high-value data processor employs some collection of methods by which various sources of information are converted into common and consistent formats to facilitate data processing, reporting, and analytics.
To standardize, each element of structured data is subjected to examination and restructuring to achieve physical uniformity. Name and address information can be used to illustrate. For example, a New York City address containing "100 East 4 Street" can appear in many ways. On inspection, there is recognition of what is commonly referred to as “address line 1 .” Within address line 1, there is a number component (100) and a street component (East. 4 Street). Within the street component, there is direction prefix (East.), street name (4), and a street type (Street). Here a just a few ways in which this address could appear in a data source:
100 E. 4th St.
100 E 4th St
100 E. 4th Str.
100 East 4 Street.
One Hundred East 4th Street.
One Hundred E. 4 Str
The United States Postal Service (USPS) takes this input on its website as follows:
USPS takes this input and makes several intelligent standardizations. Based on the information entered, the results are as follows:
The street number remains as-is.
The prefix “East” is changed to “E.”
The street is changed from “4 Street” to “4TH ST.”
City and state remain as -is.
Zipcode (5 + 4) is added.
Also, note what happens when the street number is spelled out as “One Hundred.”
The standardization process cannot process this information because it does not recognize “One Hundred” as a street number. To eliminate this type of error, a standardization engine must be able to reliably recognize this form of street address or address line 1 to the exclusion of all other forms of this data element to produce the correct result. Without this processing, an information worker would not be able to identify profiles with a common address, on the same street, in the same city or state, and so on. Most regions and countries have standardized address services that align variations into an agreed-upon format and style. If standardization did not occur, there would be no way to associate these various attributes with other profiles in a system. This would prevent the ability to recognize that several profiles share the same exact attributes.
Once data is standardized it can be submitted for matching. This is where external profiles are aligned against the customer master data galaxy to determine if a profile is new to that galaxy or should be associated with an existing profile in the galaxy. In order for "new" or "unknown" profiles to be matched to existing profiles or enumerated as new profiles, there should be a concept of an authoritative source in the master data system. This will be covered in Managing The Razor’s Edge – Part 5 Matching.
コメント