Measuring Overlap Between Digital Subscribers and Print Home-Delivery Households
A Daejeon publisher asked us a question that sounds simple: how many paying digital subscribers also receive the Saturday print edition at home? The difficulty was not visualization — it was address matching discipline.
Why the CRM export was not enough
The subscription database listed billing addresses for digital accounts and a separate delivery file for print routes. Roughly 14% of records had formatting differences — apartment block numbers in different fields, romanised versus hangul street names — that prevented naive joins.
Matching protocol
We normalised addresses through a three-pass process: standardise hangul formatting, extract building identifiers, then fuzzy-match remaining pairs above a 92% similarity threshold. Uncertain matches went to a manual review queue of 847 records; 312 were confirmed overlaps, 89 were rejected, and 446 lacked sufficient address detail.
Validation survey
We then sampled 400 households stratified by match confidence and ran a short phone survey asking whether they held both subscriptions. The survey-adjusted overlap estimate landed within 1.8 percentage points of the matched figure — close enough for the product team's cannibalisation debate.
The board extended the print sunset timeline by 18 months. The chart type did not matter; the address protocol did.