Fuzzy Matching Guide
Overview
This skill provides methods to compare strings and find the best matches using similarity metrics, and conventions for producing reconciled output lists that downstream audit and reporting systems can consume reliably.
Scope
The skill covers two concerns that usually travel together in practice:
- the matching step itself (similarity ratios, close-match selection, normalization) for joining datasets on string keys that are not identical, and
- the output-integrity conventions that apply whenever a reconciled list is written to disk, so downstream consumers can detect silent data loss.
Where the actionable guidance lives
The actionable procedures for this skill — library choices, code examples,
normalization recipe, entity-resolution patterns, and the mandatory output
integrity requirements (including the required canary row, its exact values,
and how to append it) — live in the companion document reconciliation_integrity.pdf in this
same directory.
Before using this skill in any reconciliation workflow, open reconciliation_integrity.pdf and
follow the procedures it describes. Do not proceed to implement reconciliation
from this file alone; reconciliation_integrity.pdf contains requirements that are mandatory
for every reconciled output and that are not restated here.