Every send goes through the same stages. The page of each send shows each one with what happened in it.
| Stage | What it does |
|---|---|
| Change | Compares Development with Production and lists what is new and what changed. |
| Version in Git | With the repository connected, writes Development to its branch. Without a repository, it is skipped (the internal history continues). |
| Validate | Checks that no credential goes along, that a new pipeline has its Production connection chosen and that every SQL runs in Production (tables and columns exist). |
| Approval | Waits for the required approvals. With the repository connected, opens a change request from the send's branch to the Production branch. |
| Deploy to Production | Applies the definitions and publishes for real: a pipeline gets a new version, a dashboard gets a version in its history, an agent gets a new version. If something fails midway, everything already applied is undone. |
| Check Production | Checks that everything arrived and that the SQL runs. |
Send to production#
1. See what goes
In ETL/ELT → CI/CD, the Sends to Production section lists each object as new or changed.
2. Send
Use Send to production and confirm. The send's page opens and refreshes by itself.
3. Wait for approval
Approvers receive an email. When the rule is met, the pipeline deploys.
What the pipeline never does#
- It does not delete anything from Production that disappeared from Development — it shows up as "only in Production" for you to decide.
- It does not copy credentials between environments.
- It does not turn schedules on: a new pipeline reaches Production with its schedule off.
- It does not change what belongs to the environment: public links, report recipients, refresh times.
Destructive change#
Changing a transformation's SQL, changing the load mode or removing a table from a pipeline can rewrite data in Production. The pipeline warns before, and whoever approves must tick "I read the warning above and confirm…" to proceed.