# Virtual Data Modeling — Zone Decision Matrix (July 2026)

| Need | Prefer | Avoid |
|------|--------|-------|
| Ad-hoc / bursty lake queries | Athena views + workgroup scan caps | Stacking 5+ nested Athena views |
| Concurrent BI, tighter latency | Redshift late-binding views / data sharing / Spectrum | Athena-only for always-on dashboards |
| PII-safe broad analyst access | Lake Formation column/row filters on governed tables | PII masks only in view SQL |
| Sub-100ms app reads | DynamoDB / Aurora / cache (not VDM) | Athena cold starts |
| Persist transforms for downstream | Glue / dbt materializations | Deep view-on-view chains |

## Cost levers (Athena)

- Parquet/ORC + partition pruning
- Workgroup per-query scanned-data limits (~$5/TB scan — re-quote)
- QuickSight SPICE for stable dashboards

## Counter-case

Unpartitioned multi-TB curated tables make VDM a bill amplifier — fix storage first.