Auto-Apply

Auto-Apply evaluates incoming recommendations and decides whether to execute immediately, queue for manual review, or suppress based on your mode, limits, exclusions, and safety monitors.

Prerequisites

• Connected Amazon account
• Selected marketplace
• Contributor or Admin role
• Recommendation flow already active in Command Center

Expected outcome: You can configure and operate Auto-Apply in the correct account-marketplace context.

Quick Start

1. Open Settings > Auto-Apply.
2. Click Set Up Auto-Apply.
3. Complete wizard steps: Comfort Level, Safety Net, Activate.
4. Start in Shadow Mode for low-risk validation, or activate immediately.
5. Monitor Activity Log, Trust & Accuracy, and circuit breaker status.

Expected outcome: Auto-Apply is configured and operating with observable safety signals.

Detailed Workflows

Workflow: Complete the Setup Wizard (Step-by-Step)

1. Start wizard from Settings > Auto-Apply.
2. In Step 1: Comfort Level, select one mode (conservative, balanced, or aggressive).
3. Review simulation preview for that mode (auto-applied, manual queue, suppressed estimates).
4. Continue to Step 2: Safety Net.
5. In Safety Net, optionally exclude specific campaigns from auto-apply scope.
6. Review built-in safety notes (safety pause/circuit breaker and related protections).
7. Continue to Step 3: Activate.
8. Choose one launch option:
   1. Start Shadow Mode: decisions are evaluated but not executed.
   2. Activate Now: qualifying decisions execute immediately.
9. Finish setup and confirm settings persist on the Auto-Apply page.

Expected outcome: Wizard saves a complete baseline configuration including mode and exclusions.

Workflow: Choose Aggressiveness Level with Exact Preset Values

1. Choose Conservative when you need strict gating.
2. Choose Balanced for moderate automation with meaningful safeguards.
3. Choose Aggressive for high automation coverage.
4. Use Custom later if preset behavior is close but not exact for your risk model.

Expected outcome: Mode aligns with your operational risk tolerance and review bandwidth.

Workflow: Understand Preset Equivalence (What Each Mode Actually Means)

1. Use this mapping when selecting mode:

Mode	Confidence threshold	Max bid change	Daily spend limit	Hourly spend limit	Excluded recommendation types
Conservative	0.90	15%	$100	$20	16 types excluded
Balanced	0.85	25%	$250	$50	9 types excluded
Aggressive	0.75	40%	$500	$100	0 excluded

1. Treat excluded-type count as a major risk dial, not only confidence threshold.
2. Re-check mode after major account strategy changes (launches, restructures, budget policy changes).

Expected outcome: Team decisions about “aggressive vs conservative” are based on exact system behavior.

Workflow: Configure and Use Pause/Resume

1. Open Auto-Apply controls.
2. Pause with duration: 1h, 4h, 24h, or indefinite.
3. Verify paused status in the UI.
4. Resume when issue window closes.
5. Confirm decision flow restarts.

Expected outcome: Automation can be halted and restored without deleting configuration.

Workflow: Operate with Safety Monitors

1. Keep circuit breaker enabled.
2. Review current thresholds:
   1. ACOS threshold percent
   2. CVR decline percent
   3. Spend multiplier percent
   4. API failure count
   5. Revert rate percent
3. Review status banner for tripped or paused monitors.
4. If tripped, inspect failure patterns in Activity Log before resuming.
5. Adjust thresholds/exclusions only after identifying a repeatable cause.

Expected outcome: Safety events are handled as controlled incidents, not ad hoc toggles.

Workflow: Run Shadow Mode Validation Before Full Activation

1. Enable Auto-Apply via wizard using Start Shadow Mode.
2. Process recommendations manually for a validation period.
3. Compare manual decisions vs shadow decisions in trust/reporting views.
4. If alignment is acceptable, switch to live activation.
5. Re-evaluate mode after first live period.

Expected outcome: Activation is based on observed agreement, not assumption.

Workflow: Tune with Custom Mode

1. Start from the nearest preset.
2. Adjust confidence threshold and bid/spend limits.
3. Add exclusions for sensitive recommendation categories or campaigns.
4. Save and monitor revert rate plus outcome quality.
5. Iterate one control at a time.

Expected outcome: Automation profile converges to account-specific operating tolerances.

Common Errors

Error: “Auto-Apply enabled but nothing executes”

1. Confirm config enabled = true.
2. Check if system is paused or a circuit breaker is tripped.
3. Verify recommendation confidence meets current threshold.
4. Check excluded recommendation types and excluded campaigns.
5. Confirm decision outcome reason in history/activity views.

Expected outcome: You identify the specific gate preventing execution.

Error: “Too many unexpected auto-applied actions”

1. Reduce aggressiveness level (for example, aggressive to balanced).
2. Increase confidence threshold.
3. Reduce max bid change and spend limits.
4. Add exclusions for high-risk recommendation types.
5. Monitor revert rate after each change.

Expected outcome: Execution volume and risk profile return to acceptable bounds.

Error: “Frequent safety pauses”

1. Inspect which monitor is tripping most often.
2. Correlate with recent campaign or API instability.
3. Fix root cause before increasing thresholds.
4. Resume and monitor closely.

Expected outcome: Pause frequency declines because root causes are addressed.

Error: “Mode change did not behave as expected”

1. Confirm saved mode on config card.
2. Re-check expected preset values from the mapping table.
3. Ensure custom overrides are not being confused with preset behavior.

Expected outcome: Team has a shared, correct interpretation of active mode behavior.

FAQ

What are the wizard steps right now?

The current wizard flow is three steps: Comfort Level, Safety Net, and Activate.

What does “aggressive” concretely mean?

It sets lower confidence gating (0.75), wider bid movement (40%), higher spend limits ($500 daily / $100 hourly), and no default recommendation-type exclusions.

Is Balanced the default for Custom mode fallback values?

Yes. Custom-mode omitted fields are resolved from balanced-like defaults unless explicitly set.

What are baseline safety defaults when no saved config exists?

Circuit breaker enabled with threshold 3 in 60 minutes, plus default safety monitor values (ACOS 50, CVR decline 30, spend multiplier 150, API failures 3, revert rate 10).

Does enabling Auto-Apply remove manual control?

No. Manual queue, decision history, and revert workflows remain part of normal operations.

Last Updated

• Last updated: 2026-02-21
• Version assumptions: Wizard steps/components in AutoApplyOnboardingWizard, preset values in mode-presets.ts, and /settings/auto-apply config/pause/resume endpoints