Safeguards
How to use it
Put guardrails on automated ad spend before you scale:
- Set a dry-run window so new plans record (don't spend) at first. (Onboarding → Safeguards)
- Set a daily spend cap you're comfortable seeing on your card. (Onboarding → Safeguards)
- Know where the kill-switch is — one toggle halts all automation instantly. (Orchestration dashboard)
- Reconnect promptly if you see a re-auth banner.
⏱ ~5 min · 💳 Pro+ · 🎯 Automate ad spend without the "$10K overnight" risk
Why safeguards exist
Automated ad orchestration is only useful if you trust it with spend — and you can only trust it if you can stop it, preview it, and cap it. These four safeguards are the trust layer: they sit in front of every action that moves money, so you can hand off ad operations without handing off control.
The global kill-switch
One switch on the orchestration dashboard halts every automated plan step across your store, immediately. Crucially, it does this without mutating your plans or enrollments — nothing is cancelled or reset. When the switch is on:
- Every due step holds instead of executing.
- An active enrollment self-heals: it re-enqueues itself on a short delay (about every 5 minutes) so it resumes automatically the moment you switch the kill-switch back off — no manual restart, nothing stranded.
- An awaiting-approval enrollment simply keeps waiting for its human approver; it isn't spun in a poll loop.
Use it as an emergency brake: a bad campaign, a billing surprise, a platform outage — flip it, investigate, flip it back. Your plans pick up exactly where they left off.
The dry-run window
When a plan is launched or approved, it's armed with a dry-run window based on your configured dry-run days safeguard. While the window is open, the plan's ad-launch steps record what they would do without spending:
- Each ad-launch writes an audit entry to the activity timeline — platform, operation, and the fact that it would spend — so you can watch exactly what the orchestrator intends to do before any real budget moves.
- Only ad launches are gated (they create and fund campaigns). Actions that don't move money — pausing an ad, syncing an audience — run normally during the dry run, so the plan still does useful non-spend work.
- When the window elapses, launches begin firing for real automatically.
This is the "watch it drive before you take your hands off the wheel" control — you get days of visibility into the orchestrator's decisions at zero spend risk.
| During dry-run | Behaviour |
|---|---|
| Ad launch (create/fund campaign) | Recorded, not fired — audited to the timeline |
| Ad pause | Runs normally |
| Audience sync | Runs normally |
The daily spend cap
A daily cap bounds how much orchestration can commit in a single day. When the day's automated spend would exceed the cap, further launches defer rather than proceed — so an aggressive plan (or a bug) can't blow through your budget in one run. Set it in onboarding alongside the other safeguards.
Re-authorization gates
Ad platform tokens expire. Rather than letting a step fail opaquely, orchestration gates execution on token validity: if a platform's connection has lapsed, the affected steps are blocked and you're prompted to reconnect — with a clear banner on the dashboard — so spend never rides on a broken connection.
Configuring safeguards
Safeguards are configured during onboarding and persisted to your account (they're not browser-local settings):
| Safeguard | Setting | Typical starting point |
|---|---|---|
| Kill-switch | Toggle on the dashboard | Off (armed for emergencies) |
| Dry-run window | Days a new/approved plan records before spending | 3–7 days |
| Daily spend cap | Max automated spend per day | Your comfortable daily ad budget |
| Re-auth gates | Automatic | Always on |
New and approved plans are armed with a dry-run window by default. Check the dry-run days and daily cap in Onboarding → Safeguards so a plan doesn't sit in record-only mode longer than you intend — or start spending sooner than you want.
Best practices
✅ Keep a dry-run window on new plans. A few days of recorded-not-fired launches is the cheapest insurance you'll ever buy on ad automation.
✅ Set the daily cap to a number you'd be comfortable seeing on your card, then raise it as you build trust.
✅ Treat the kill-switch as a real emergency brake — because it doesn't mutate plan state, using it costs you nothing but the paused time.
❌ Don't ignore a re-auth banner. A lapsed token means the affected platform is silently not executing; reconnect promptly.