The research question
A folder contains a quote for Northstar Services and an invoice from North Star Service Ltd. A name comparison can make this pair worth examining, yet the names alone cannot establish that the organizations are identical. One could be a subcontractor, a trading name or an entirely unrelated business operating in another region. Entity-resolution research separates selecting candidate pairs from deciding whether they identify the same object. That distinction makes the evidence and reversibility of a CRM identity decision central to this design question.
A concrete design choice
A proposed customer card should preserve both original names, the documents they came from and any registration identifiers. The review screen can show matching and conflicting attributes together. Confirming a relationship should create a recorded decision, while leaving a candidate unresolved must remain a normal, usable state. Later evidence should be able to reopen that decision without erasing either source history.
The review should include contradictory attributes rather than merely listing similarities. A different registration number is often more informative than several matching words. Keeping rejected links also matters: otherwise the same ambiguous pair can reappear after every import and consume the reviewer’s time. An explicit rejection should say which evidence was considered and remain revisable if genuinely new material changes the interpretation.
What would demonstrate usefulness
A useful evaluation includes deliberately similar names belonging to different companies and genuine aliases with different spelling. Measure mistaken merges separately from missed matches, and ask whether the reviewer can explain each accepted link directly from the documents.