When a Customer Take-on System cannot find a match, there is a good chance that the operator will not perform any further searches, and simply add the "new" customer as a new record.

Even when the system finds an existing record, if that record is not identical or not easily visible in the list, a new record will often be added.

The important consequence of missing a match, if there was one, is not the duplication in itself, not the extra disk space that duplicate records use, nor the increase in candidates returned in future searches, but that the new customer record will be "unaware" of the existing one. Therefore, in future transactions it will often be random as to which customer duplicate will be used or updated. Such unlinked duplication is a major risk to the integrity of the database. It is a risk to the business processes which expect to find only one record per customer, or at least to find all records relating to a customer together.

Duplication can be tolerated provided that the duplicate records are linked. Resolving duplication with merge/purge can cause data corruption and data loss.

Provided that duplicate records are linked, and systems are built to recognize the links, the decision to merge or purge duplicates becomes one of housekeeping rather than absolute necessity.

The real problem with duplication is when systems which use the data cannot resolve it, resulting in duplicate or unintended mail and even duplicate product being sent to customers, as well as a distorted view of the customer base.