Merge pull request #1827 from segmentio/evanSegment-clarify-object-throughput

Clarify object throughput logic

Author: markzegarelli

src/guides/usage-and-billing/mtus-and-throughput.md

@@ -62,7 +62,11 @@ An event is a data collection triggered in response to a user action: a [Track c
 
 This is in contrast to "Objects" which represent a single thing that persists over time and can be updated. Objects have "traits" (instead of properties) which record information about that object, and which can change over time. For example a "user" object could have a trait of "email" which doesn't change often, but could also have a [computed trait](/docs/personas/computed-traits/) like `logged_in_last_7_days` that changes between `true` and `false` based on how much they use your site.
 
+## How is object throughput calculated?
 
+[Object Cloud Sources](/docs/connections/sources/about-cloud-sources/#object-cloud-app-sources) retrieve records from integration partners on a scheduled basis. Segment processes these records before writing them out to connected [Storage Destinations](/docs/connections/storage/). Segment counts one object for each record retrieved from a Cloud Source. The number of objects ingested during a billing period has a direct impact on [throughput](#what-is-throughput), which is calculated as `(objects ingested + API calls received) / MTU allowance`.
+
+Depending on the capabilities of the partner's API, Segment may need to retrieve all available records and then deduplicate them prior to writing them out to a connected storage destination. In such cases, all retrieved records are still counted as ingested objects, even if the same records are retrieved multiple times in a given billing period. If you experience overages due to high object throughput, contact [[email protected]](mailto:[email protected]) to request a less frequent sync cadence.
 
 ## MTUs, object throughput, and Cloud sources