Failure Modes & SLAs

Honest catalog of what can fail and how Layers handles it.

Ad account restricted / disabled

Symptoms: Spend stops; campaigns can't be edited; platform sends an email.

Layers' behavior:

• The next metrics sync detects the restriction.
• Optimizer is paused for that ad account.
• Notification fires to Project Admins.
• Account-health badge turns red on the layer page.

Your action:

• Resolve with the platform (typically a policy / billing / verification issue).
• Once restored, campaigns auto-resume.

SIFT outage

Impact: UGC discovery and metrics syncing pause.

Layers' behavior:

• Existing UGC continues running as ads.
• New posts aren't synced until SIFT recovers.
• No new UGC enters the eligible pool.
• Notification fires after 1h of consecutive failures.

Your action: Wait. SIFT outages are usually <2h. Layers retries automatically.

Model provider outage

Impact: Content generation pauses.

Layers' behavior:

• Generation queue holds (no failures).
• Once the provider returns, queue drains automatically.
• Scheduled posts that lacked generated content miss their slot gracefully (skipped, not failed).

Your action: None unless prolonged. If >30 min, manually approve any existing eligible content to fill upcoming slots.

Agent loop failure

Impact: App Machina or UGC creative generation stalls.

Layers' behavior:

• Stalled agents are killed and retried with backoff.
• After max retries, the workflow fails and a notification fires.

Your action: Look at the workflow runs panel. Re-run with adjusted inputs. Contact support if persistent.

Workflow processing outage

Impact: All scheduled commands pause.

Layers' behavior:

• Workflows are durable; they resume from the last completed step when processing recovers.
• No data loss; just delayed execution.

Your action: None. Layers handles this.

CAPI silent failure

Impact: Conversions reach sdk_events but don't reach Meta / TikTok. Platform-attributed conversions diverge from SDK truth.

Layers' behavior:

• The relay logs failures (meta-capi-error, meta-capi-failed, and capi-vault-decrypt-failed markers).
• CAPI relay silently no-ops if capi.enabled !== true, pixel_id is missing, or access_token_vault_id cannot be decrypted — so misconfiguration looks like "no CAPI traffic" rather than a loud error.

Your action:

• Check the layer's CAPI config: is enabled = true, pixel_id set, vault_id resolvable?
• Regenerate the CAPI token if expired and re-paste.
• See Pixel & CAPI troubleshooting.
• The SDK Event Ingestion & CAPI Forwarding doc describes how the relay resolves CAPI config per app_id.

Wallet runs out (agency mode)

Impact: Agency-mode ads can no longer be funded from the Layers wallet.

Your action: Top up the Org wallet. See Wallet & billing.

SLAs

Layers does not publish a standard uptime SLA. Availability targets, incident-notification windows, and RCA-delivery commitments are negotiated per agreement at contract time — email legal@layers.com to discuss. See also the DPA for contractual framing.

Real-time status and incidents: support@layers.com.

Reporting an incident

• Customer-impact: file via the in-app Support widget.
• Security incident: security@layers.com.
• Compliance / DPO: dpo@layers.com.