Trust and Accuracy Dashboard

Trust & Accuracy tracks how recommendations perform against measured outcomes and helps tune automation confidence.

Prerequisites

• Connected Amazon account
• Selected marketplace
• Enough recommendation outcomes for meaningful statistics

Expected outcome: Trust metrics are interpretable and actionable.

Quick Start

1. Open Trust & Accuracy.
2. Review summary cards and measured outcomes count.
3. Check low-accuracy warning if present.
4. Review charts for accuracy by type, prediction scatter, confidence distribution, and data readiness.
5. Use the activity feed and shadow report to decide whether to adjust Auto-Apply or Smart Tuning settings.

Expected outcome: Accuracy signals are translated into concrete settings changes.

Detailed Workflows

Workflow: Accuracy Remediation Loop

1. Identify low-performing recommendation types.
2. Increase confidence threshold or tighten exclusions where needed.
3. Watch for the low-accuracy warning when measured outcomes are sufficient.
4. Monitor subsequent outcomes and re-assess after sufficient new samples.

Expected outcome: Accuracy trends improve with controlled policy changes.

Workflow: Validate Data Readiness for Calibration

1. Open readiness and report indicators.
2. Confirm measured outcomes and report counts.
3. Move to Smart Tuning activation when minimum criteria are met.

Expected outcome: Calibration activation is evidence-based.

Workflow: Evaluate Shadow Decision Alignment

1. Review shadow/simulation report section.
2. Compare predicted vs operator decisions.
3. Decide whether to move to stronger automation mode.

Expected outcome: Mode progression uses measured alignment data.

Workflow: Feed Insights Back into Operations

1. Use activity feed to identify repeated failure patterns.
2. Map issues to recommendation types.
3. Update SOP, Auto-Apply policy, or Smart Tuning expectations.

Expected outcome: Trust dashboard becomes a closed-loop quality control tool.

Workflow: Launch Auto-Apply from Trust & Accuracy

1. Open Trust & Accuracy.
2. If Auto-Apply onboarding has not been completed and the page has data, click Set Up Auto-Apply.
3. Complete the wizard: Comfort Level, Safety Net, Activate.
4. Return to the dashboard and monitor summary cards, circuit-breaker state, and shadow/live behavior.

Expected outcome: Operators can move directly from observability into controlled automation setup.

Common Errors

Error: Dashboard appears empty

1. Confirm recommendation outcomes exist.
2. Continue operations to accumulate measured results.
3. If Auto-Apply has not been configured yet, expect the welcome state and use it as the setup entry point.
4. Revisit after additional cycles.

Expected outcome: Empty state is understood as data-readiness stage.

Error: Low-accuracy warning persists

1. Verify sample size is meaningful.
2. Tighten high-risk automation settings.
3. Re-check after outcome lag window.

Expected outcome: Persistent warning triggers deliberate tuning, not blind toggling.

Error: Chart data intermittently fails to load

1. Refresh page.
2. Validate account/marketplace context.
3. Retry later if backend source is transiently unavailable.

Expected outcome: Operator distinguishes transient data fetch issues from persistent defects.

FAQ

When should I pay attention to the low-accuracy warning?

The warning appears when accuracy drops below 70% and there are at least 10 measured outcomes.

Is this the same thing as Auto-Apply settings?

No. Trust is observability; Auto-Apply is execution policy.

Should I use this page before turning automation up?

Yes. It helps validate recommendation quality before increasing automation scope.

Last Updated

• Last updated: 2026-03-18
• Version assumptions: current TrustDashboardPage warning logic, chart set, activity feed, shadow report, Smart Tuning summary, and Auto-Apply launch entry point