A startup is using Cloud Spanner for a global user base. They need to design a schema that minimizes interleaved table joins for common access patterns. Which schema design principle should they prioritize?