# Partitioning
<a name="msk-data-delivery-s3-template-partitioning"></a>
+ **By Kafka partition** — include `!{partition-id}` as a path segment, for example `!{topic-name}/!{partition-id}/...`.
+ **By time** — include time variables as path segments, for example `!{yyyy}/!{MM}/!{dd}/!{HH}/...`. Avoid minute-level (`!{mm}`) granularity for high-throughput topics — it creates many small objects and can slow delivery and downstream queries.
+ **By topic** — include `!{topic-name}` as a path segment.
+ **Unique object name (required)** — each delivered object must have a unique key, so the final path segment must include a batch token: `!{sequence-number}` or `!{kafka-offset}`.