Key Concepts
A handful of terms show up everywhere in Nimbus. Learning them once makes the rest of the documentation — and the product — much easier to follow.
Business Unit (tenant) #
A Business Unit is a single SFMC account that Nimbus monitors, identified by its MID (Marketing Cloud ID). Inside Nimbus a connected Business Unit is called a tenant. Everything you see — events, queries, data extensions, rules — is scoped to one tenant, and you switch between them with the tenant selector in the sidebar.
Automations, events, and activities #
An automation is an SFMC Automation Studio workflow. Each time it runs it produces events — automation.started, automation.completed, automation.errored, automation.stopped, or automation.skipped.
Within a run, each step is an activity: a SQL query (activity type 300), an import (type 43), a file transfer, and so on. Nimbus records every activity with its own status and duration, so you can see exactly which step in a long automation was slow or failed.
The CloudPage snippet #
SFMC's native webhooks tell you an automation finished, but not much else. To get per-step detail, Nimbus deploys a small AMPScript snippet into a CloudPage in your account. When an automation finishes, Nimbus calls that CloudPage, the snippet reads the run's internals, and posts an enriched result back. The snippet is versioned — Nimbus redeploys it automatically when a newer version is available.
ENS #
The Event Notification Service is SFMC's official event feed. When you connect a Business Unit, Nimbus registers an ENS callback and subscription so SFMC pushes automation events to nimbus-ingest the moment they occur.
Monitoring rules #
A monitoring rule tells Nimbus which failures to act on. A rule targets either a specific automation or a folder of automations, and decides whether a failure should trigger an automatic rerun — and whether AI recovery is allowed.
Reruns and rerun attempts #
When a monitored automation fails, Nimbus creates a rerun attempt. It waits for the original run to finish, builds a trimmed copy of the automation containing only the failed and unstarted steps, and fires it. Attempt #1 is a plain rerun; attempt #2, if needed, is an AI-assisted rerun.
AI remediation modes #
When AI recovery is enabled on a rule, it operates in one of three modes:
| Mode | Behavior |
|---|---|
| **Advisory** | AI diagnoses the failure and records a recommendation. Nothing is changed. |
| **Assisted** | AI applies a fix to a staging copy and reruns it. A human approves before it reaches production. |
| **Autonomous** | AI applies, reruns, and — if successful — promotes the fix to production automatically. |
See AI-Powered Remediation for details.
Data extensions, staleness, and conflicts #
A data extension (DE) is an SFMC table. Nimbus inventories every DE and tracks when it was last written. Staleness grades each DE as fresh, aging, or stale based on per-tenant thresholds. A conflict is when two jobs read and write the same DE in overlapping time windows — a common cause of query timeouts.
Organizations and roles #
An organization groups users and the Business Units they can reach. Organizations are either agency or client type. Within a tenant, users hold a role — owner, admin, operator, or viewer — that controls what they can do. See Team & Roles.