Audience question: splitting a document God object into separate Invoice and Receipt models is good in principle, but the business often wants to see 'all my documents' in one table with pagination, sorting and search. How to handle that? Chikahiro answers that read and write concerns should be separated — write operations need consistency and integrity (drive the modeling), while read operations may need denormalisation via materialized views or Elasticsearch. The cross-document list is a read-side concern, orthogonal to the domain modeling.