The task
Scenario. Your PR renames user_id to customer_id in dim_users. The diff looks small, but the real risk is the views, dbt models, and dashboards that read dim_users and never show up in the diff. You want the full blast radius while you are still in review, not after the merge.
The ask
- Every downstream view, table, and dashboard that references the table
- Each one's owner and certification status
- Enough to decide whether the change is safe to merge
Hero prompt
I'm about to rename the user_id column to customer_id in the dim_users table. Trace the downstream consumers of dim_users — start with the direct ones and expand into the branches that matter — so I can assess the impact before merging.
Step by step
-
Find the asset you're changing and confirm it has downstream lineage.
I'm renaming <column> to <new name> in <table> — pull <table> and check what depends on it.
-
Trace its downstream consumers.
Trace the downstream views, tables, and dashboards that read <table> — direct consumers first, then expand into the branches that matter.
-
Return the impact list with owners and certification.
For each impacted asset, include its owner and certification status so I can decide what to test and who to warn.
What you'll give it
Required context
- The table or column changing in the PR (user_id in dim_users)
- Lineage populated for that asset
Optional context
- Whether to include indirect (multi-hop) consumers or only direct ones
- The certification bar that matters to you, so certified consumers stand out
What Atlan creates
Outputs
- Every downstream view, table, and dashboard that references dim_users
- The owner and certification status of each
- A checkable impact list to paste into the PR
Summary. "I think this is safe" becomes a list you can check off. You see who to warn and what to test before merging, not after a dashboard goes blank.
Nothing is written. This recipe only reads lineage and metadata, so it is safe to run against production from inside any review.
Follow-up prompts
Keep going once the first answer lands.
Draft the warnings
For the certified consumers on that list, draft a short heads-up I can send their owners.
Focus the testing
Which of these consumers are certified or heavily queried, so I test those first?
Widen the change
I'm also dropping legacy_flag in the same PR. Add its downstream consumers to the impact list.
Tips & troubleshooting
Paste this into your PR review. It turns "I think this is safe" into a list you can check off.
- Match the tool to the work. Run it in the AI tool you already review PRs in, so the impact list lands right next to the diff. Ask for the immediate downstream first and expand as needed; a full multi-hop pull can be huge.
- Automate or chain it. Chain it into CI: on a PR that touches a modeled table, post the impact list as a comment automatically.
- It only reads. Nothing is written back, so run it as often as you like, on any branch, without a change window.