Matching customer reviews to branches
Measured impact100+ branchesReview data gained a location dimension
Context
Multi-branch retail chain (100+ locations)
Problem
Map-platform reviews were not matched to branches; “which branch is getting negative reviews” had no answer.
Approach
I wrote a Python matcher that generates query variants from the branch list, scores candidates on address/district/city/name tokens and picks the best match; results were joined to the sentiment table.
Impact
Review data gained a location breakdown; branch-level negative concentration became reportable.