Journey Builder performs best when the heavy lifting of segmentation happens before contacts are evaluated for entry, not inside the journey itself. Performing large-scale segmentation in Automation Studio (B) offloads CPU-intensive filtering to SQL Query Activities, which are optimized for bulk processing across millions of rows, rather than relying on Journey Builder ' s per-contact entry evaluation engine. Creating a prefiltered, dedicated sendable copy of the source data extension for each journey (D) avoids lock contention and re-evaluation overhead that occurs when multiple journeys share a single large entry source, and it ensures each journey only processes the exact population it needs. Option A is a valid best practice for decision-split performance generally, but Attribute Groups govern contact-level data joins for Contact Data-based splits and does not address entry-source scale at 10 million records specifically — it ' s a secondary optimization, not one of the two primary levers for this scenario. Option C describes filtering directly on the entry source Data Extension, which still forces Journey Builder to evaluate the full underlying population against the filter at entry time, rather than pre-segmenting it, and is less efficient than doing that work in Automation Studio beforehand. Reference: Journey Builder Performance Best Practices, Automation Studio Segmentation.