All of the more than 2.1 million organizations in the NCCS databases are classified in NCCS's master list of NTEE codes. These codes come from a variety of different sources.
Since the late 1990s, the IRS has classified new organizations as they applied for exemption. As of 2006, more than 320,000 organizations have been classified by the IRS "determination specialists," who are responsible for reviewing organizations' initial applications for recognition of their exempt status. After they have reviewed the lengthy and descriptive IRS Form 1023 or 1024 application, they are required to assign a single NTEE code to the organization. This code is included in the IRS Business Master File, which NCCS downloads at least twice per year.
Approximately 106,000 organizations were classified by NCCS, then part of Independent Sector, based on an assessment of program descriptions contained in Parts 3 and 8 of the Form 990 between 1994 and 1996
More than 90,000 organizations have been classified since 2004 by NCCS staff and consultants using the web-based EON (Exempt Organization NTEEs) program. This program displays Part III information and a range of other items to provide classifiers with all the information they need to make accurate decisions.
More than 120,000 records were manually classified in the late 1990s by NCCS using only the organizations' names and IRS activity codes for guidance.
In 2007, NCCS completed work on a sophisticated automated classification tool that incorporates information from the organization name, activity codes, and descriptions of organizational purpose and programs. This program incorporates more than 1,500 complex classification rules to identify possible NTEE codes and assign the most likely code to every organization in its database. Each rule is tested on a sample of records and a confidence level is assigned to the rule to determine how reliable it is likely to be.
The final NTEE selection for an individual record depends on what sources for the NTEE are available -- e.g., an IRS-assigned code, an EON-assigned code, etc. -- and the level of confidence assigned to each code. For example, an organization classified manually by NCCS using EON would have a confidence level of "A" if the classifier felt confident that the information on the organization's purpose and programs pointed clearly to a single category. If the program and purpose information is missing or vague, then the classifier might assign a confidence of "C". In this scenario, the NTEE code generated by the automated program, if it gets a confidence level of A or B, would take precedence over the manual code.