← Portfolio
Proprietary

Merchant categorization at Bci

Industry
Banking
Type of work
Master data and classification
Period
2018

Confidential project: code and data are not public.

Architecture

SII tax data Business customers Bci Cards transactions GeoResearch now XBREIN Merge matching sourceswith no shared key Category + tier tier byaverage ticket Merchant record category, tierand location Uses: benefits, customer recommendations and more · Python · Teradata

Goal

Millions of card purchases went through Bci, but merchants came with incomplete or inconsistent categories and no reliable location. The bank needed one record per merchant to design benefits, recommend merchants to customers and support other commercial uses.

What I did

  • Built practically the whole process, in Python on Teradata.
  • Merged sources with no clean shared key: data from Chile's tax authority (SII), Bci's business customer base, card transactions and others.
  • Assigned each merchant a category and, within each category, a tier based on its average ticket.
  • Geolocated merchants with GeoResearch data (now XBREIN).
  • Almost every merchant ended up classified, with quality depending on how much information each one had.

Stack

PythonTeradataSQL