ID926 compares data between two Customer Records looking for identical or near-identical data. If it finds two records that are ‘enough alike’ on the initial run, both will be flagged as potential duplicates. If it finds a new record later only that new record will be flagged. It essentially assigns ‘points’ for each matching field and looks for at least four points total.
**Please note this alert will only notify you ONCE of possible duplicates unless changes are made to the customer record matching the alert triggers. This alert looks at ALL Customer Records regardless of relationships to eachother (i.e. BillTo, Parent/Child).
**Please note this alert does not show the identified duplicate(s) because duplicates is a training issue. The effort of tracking down the original should help the agent learn how not to make the same mistake next time. The initial output does show both the original and any dupes, because at that point it's a clean-up issue.
Run Schedule: Daily
Type of Output: Email
Here are the criteria:
Phonetic match of CustomerNames – this drives the process overall and is double-weighted so (worth 2 points)
Numerical match of non-blank primary Phone # - either identical or one ‘containing’ the other (for instance 682-555-1212 contains 555-1212) and ignoring any punctuation (worth 1 point)
Numerical match of secondary Phone # - otherwise identical to above (worth 1 point)
Close edit-distance match of non-blank e-mail address (worth 1 point)
Matching City & State or matching ZIP code (either scores 1 point but both does NOT score 2 points)
NOTE: We discussed having the alert look at street address for more points of similarity, even built out alpha logic to do it, but decided against it for three reasons. The first is technical, the SQL load added by twice-comparing address fields (you have to compare once by stripping out all alpha characters, then again by stripping out all numerics for a phonetic check) is substantial. Second, customer-moved-to-new-address is the #2 reason causing duplicate customer records in our experience (#1 being poorly-trained-agent) so different address does not have as much weight as you might expect. Third was feedback from our beta testers, who indicated that comparing address was not a dealbreaker for them although they did ask us to display it in the output for a quick mark-1-eyeball check on the initial run.