How-to · Step 4

Action glossary

← All docs

Action glossary

Every action type the platform can propose, what it means, and what to check before approving.

pause_campaign — Pause campaign

What it does: stops a Meta or Google campaign from spending. Reversible.

Typical reason: the campaign has spent significant budget with zero or near-zero results over the last 30 days.

What to check:

  • The spent and results numbers match Ads Manager.
  • The campaign isn't being throttled by something fixable (tracking gap, audience too narrow, creative fatigue) — sometimes the fix is upstream, not "pause."
  • Is this the client's primary revenue driver? If yes, escalate even if numbers look bad.

Risk: Medium if AED at risk < 5,000. High if ≥ 5,000.

adjust_budget — Adjust budget

What it does: raises or lowers a campaign's daily budget. Capped at ±30% per step by house rules.

Typical reason: scaling a winner up, or pulling a moderate loser down to test.

What to check:

  • Direction makes sense (don't raise a campaign with poor ROAS).
  • Step size is within ±30% of current.
  • Campaign has been stable for at least 3 days before scaling.

Risk: Medium. Always escalate.

refresh_creative — Refresh creative

What it does: flags that we need new ads for a campaign — does NOT auto-create them.

Typical reason: frequency > 2 (creative fatigue) or CTR declining over time.

What to check:

  • Frequency genuinely is > 2 (check Ads Manager).
  • We have ad assets available from the client. If not, escalate to Thageeshan to request them.

Risk: Low. Just a flag — no live change.

flag_for_review — Flag for review

What it does: raises a structural issue the operator should look at. Not a live change.

Typical reasons: missing WhatsApp track, tracking gap, naming convention broken, deprecated campaign sitting around, audience overlap.

What to check:

  • Read the issue field — what specifically is flagged.
  • Triage by impact: tracking gaps are urgent (we're losing attribution), naming is cosmetic.

Risk: Low (no live change), but the underlying issue may be high impact.

launch_creative — Launch creative (future)

What it does: creates a new ad with given copy and creative IDs.

Typical reason: the agent identified a winner pattern and wants to scale it.

What to check:

  • Copy follows Clozer voice rules (no em-dashes, "dirhams" in voice scripts, comparison/curiosity hooks).
  • Naming starts with CLOZER.
  • Targeting is correct for the client's market.

Risk: High. Always escalate.

send_email — Send email (future)

What it does: sends a Klaviyo campaign to a defined segment.

Typical reason: behaviour-based trigger fired (cart abandoners, recent purchases, etc.) or a calendar campaign is due.

What to check:

  • Segment size is reasonable.
  • Subject line and copy pass the Clozer voice rules.
  • This isn't a duplicate of an existing flow.

Risk: High. Always escalate.