Enriching before sequencing, not during, halved my errors

Doing all enrichment up front (vs. mid-sequence) removed a class of 'wrong person' sends. Batch the data step; keep the outreach step clean.