Everyone in the company draws from one shared credit pool, and admins can’t enforce how much any department uses. Credit Utilization is a reporting view, and dataset frequency permissions only limit how often something runs, not what it costs.
We’d like the ability to:
- Allocate credits to departments or cost centers for the contract term
- Charge usage to the owner’s bucket
- Stop scheduled executions when a bucket runs out, until an admin reallocates credits
- Alert owners based on projected usage, so they know early if they’ll run out before renewal
Scenarios
- Build and forget. A use case that uses about 3,000 credits a year is built for one person. Later it’s no longer needed, but nobody deletes it because keeping it costs the owner nothing. We’re hesitant to automate cleanup, since one dataset could power 25 dataflows and hundreds of cards.
- Old lineages. We have dataflow lineages up to 10 levels deep that have run for years. Many could be consolidated, but the only time anyone reviews them is when I schedule a meeting to walk through each one.
- Limiting builders isn’t enough. We restricted building to about a dozen super users, but they still build every request without weighing the real value, and rarely delete anything.
- Cost lands on everyone. When someone builds for another team, the credits come from the shared pool, so the requester never weighs cost against value. Later there could be technical debt because no one deleted anything when the use case changed.
- Frequency permissions are limited. They limit how often a dataset runs, not what it costs, and they don’t cover storage, dataflows, workflows, or AI Pro. One large dataflow and five smaller ones can run on the same frequency and produce the same result, but the large one uses 1 credit and the five smaller ones use 5.
