Pages 133-134 of the First Edition User’s Guide describes how matchcode components are defined. Page 50 of the First Edition User's Guide describes what some of the effects are when changing the number of characters within the matchcode that are being compared.
By setting the number of characters to [the maximum of] 18 characters increases the “sensitivity” of the detection. For example, if you have 20 people with the name John Smith in Los Angeles, you are more likely to cause a duplicate detection on a new subscriber named John Smith in Los Angeles unless the number of Duplication Check Characters (this option is set in Setup, General, General tab) is set high (so it can also compare parts of the street name). But this means that data entry must be consistent and accurate, the first, last, and street names must be spelled exactly, and Zip codes must be accurate.
Conversely, setting the number of characters to a lower number increases the possibility of duplicate detection (there may be more instances of duplicate detection, and these may include more instances of false duplicates) because it is not checking all characters of the matchcode. You will get more duplicate detections for Bob Smith in Los Angeles because it will not be comparing parts of the street address.
Duplicate detection is not a perfect science, and does not have a perfect solution. One good example would be two siblings residing at the same address, have the same initials, and each have their own subscription. Both of these siblings would have the same matchcode. Fortunately, this scenerio is unlikely to occur very often.
It is recommended that data entry procedures for consistency be defined and implemented for all First Edition User's. Examples include spelling out Avenue instead of the appreviation "Ave", Suite versus "Ste", etc. Either spell it out or use the appreviation, that doesn't matter, just be consistent. These data entry policies should also apply to all other sources of name and address data such as your web site and telemarketers.