Audience asks how often modifying structural attributes (e.g. randomizing counts) becomes a problem for downstream data analysis. Sergyenko: the obfuscation is mostly static — done once, maintained via deltas, not re-randomized on every run — so outlier masking is a one-time effect. Analytics at a high/'helicopter' level (e.g. 40–50% fewer therapy sessions in December due to holidays) stays accurate; precision only degrades for very fine-grained reports, and you shouldn't blindly rely on the exact quality of obfuscated data for fine-grained analytics.