# Lake tier decision matrix (S3 Tables + Iceberg 2026)

| Scenario | S3 Tables (managed Iceberg) | Self-managed Iceberg on S3 | Legacy Hive-style S3 + Glue crawler only |
|----------|----------------------------|----------------------------|------------------------------------------|
| Need automatic compaction + snapshot expiry | **2** — default 512 MB target, sort/z-order optional (Jun 2025) | 1 — you run Spark jobs | 0 |
| Glue 5.1 ETL with Lake Formation write FGAC | **2** — native table bucket support | 1 | 0 |
| Athena + Redshift Spectrum on same tables | **2** — Iceberg 1.x catalog | **2** | 1 |
| < 500 GB, batch weekly loads | 1 | 1 | **2** — defer S3 Tables until query pain |
| Multi-team write with column-level ACL | **2** — LF + Glue 5.1 Spark writes | 1 | 0 |
| Strict on-prem export only | 0 | 1 | **2** |

## Opinionated default

**S3 Tables** when Iceberg is already chosen and you lack a dedicated platform team to run compaction jobs. Compaction pricing dropped up to **90%** effective **July 1, 2025** ([What's New](https://aws.amazon.com/about-aws/whats-new/2025/07/amazon-s3-tables-reduce-compaction-costs/)).

## When NOT to adopt S3 Tables yet

- No Iceberg consumers — fix catalog + format choice first
- Single analyst team on CSV in S3 — Athena external table may suffice
- Regulatory requirement forbids managed maintenance deleting snapshots — tune `MinimumSnapshots` / `MaximumSnapshotAge` before cutover