CRM lead test cases, where the lead quietly disappears

Duplicates are the defining problem of lead management, and they are a concurrency problem more often than a matching problem. Two form submissions arriving within the same moment both check for an existing record, both find none, and both insert. Any deduplication that runs as a step after insert will simply create two records and leave them. The fix is a uniqueness constraint at the storage layer, and the test is to fire the submissions simultaneously rather than one after another. The matching rule itself then needs its own cases for case differences, plus aliases and shared phone numbers across a company.

Assignment is the second failure, and its worst form is silent. A lead that matches no routing rule and lands with no owner is not visible in any pipeline view, so nobody knows it exists. It generates no error and no alert, and it is discovered weeks later if at all. Every routing configuration needs a defined fallback queue, and that queue needs a test. Round robin adds a concurrency case of its own: a position counter read and written without a lock will hand the same rep two leads and skip the next entirely.

Visibility is routinely implemented in the wrong layer. Territory and hierarchy rules that filter a list view but return the record on direct identifier lookup or through search are not access control, they are presentation. Any rep who learns a record identifier reads a competing pipeline. The manager case is the mirror image: the rule has to walk the whole reporting subtree without exposing peers, and it has to do so on the server.

Finally, imports and consent interact badly. A bulk file that resets an opt out flag turns a compliant database into a regulatory problem, and the record then enters the next campaign legitimately as far as the system is concerned. The related case is partial failure: an import of a thousand rows where forty fail needs per row reporting and needs to be safely rerunnable, because the corrected file will be uploaded again and it must not duplicate the rows that already succeeded.