PHANTOMBUSTER ALTERNATIVE: INTERRUPTED-AUDIENCE HANDOFF EXERCISE Created 13 October 2026. All identities, employers, timestamps and approvals are fictional. Every person row is synthetic=TRUE and do_not_contact=TRUE. Do not visit or contact these invented profiles. These are teaching inputs, not exports from either vendor or a reliability benchmark. Goal: restart an interrupted profile-input job without treating repeated/held records as new accepted prospects. Batch A contains 10 rows; seven are manually approved, three held. Batch B contains 10 rows; two match the approved A reference. The workbench retains eight rows; two also appear in identity review. Four of the retained rows pass the fictional manual audience checks, four remain held. Final approved audience: eleven unique person-shaped keys. URL normalization alone does not verify a real identity. Replay in /tools/prospect-list-workbench.html?mode=exclude 1. Primary input: phantom-handoff-batch-b-raw.csv. 2. Reference input: phantom-handoff-batch-a-approved.csv. 3. Matching columns in both: LinkedIn URL (column index 1). Key type: LinkedIn profile URL. 4. Run Exclude. Default CSV downloads include UTF-8 BOM; Excel-safe formula protection off. Compare downloaded output/excluded/review with the three expected files. 5. Retained output is not an approved list. Join record_id to phantom-handoff-batch-b-manual-decisions.csv and inspect every HOLD. 6. Final approved file contains the exact seven approved A raw records plus B03-B06, retaining source scope, batch and date. B02's changed title is not applied to A04 by exclusion. Keep both raw records if you need an actual refresh decision. Expected counts: A raw10; A approved7; A held3; B raw10; B excluded2; B retained8; B identity-review2 (a subset of retained); B manual accept4; B manual hold4; final approved11. Never sum retained+review as separate people. Do not rewrite an opaque Sales Navigator URL into /in/. The eight-row blank workflow-status file is a planning document, not a configured scheduler/importer or vendor API schema. The expected files replay the site's unchanged local workbench. No PhantomBuster launch, Leadsforlinked extraction, enrichment delivery, CRM import, account change, or scheduled run was performed for this exercise.