Quality Fortress

Quality control system — validates every statistic before it enters the database. Layers 1–8: pre-generation guardrails, stat verification, risk scoring, and kill switch.

Kill Switch: Inactive ⓘ The kill switch is a master circuit breaker for all content generation. When active, no new content is created or published regardless of queue state. Use this for emergencies — factual errors, brand crises, or platform policy changes. Activating does not delete queued content; it simply pauses the pipeline.

All content generation systems running normally

Mode A: Safe Authority ⓘ Mode A (Safe Authority): Content is framed as authoritative benchmarks — credible, data-backed, and brand-safe. Best for broad audiences and conservative channels.

Mode B (Contrarian): Content challenges conventional wisdom using contrarian data angles. Higher engagement potential but slightly more reputational risk. Switch only when the content strategy calls for it.

Content framed as authoritative benchmarks — safe, credible, data-backed

Critical Items (24h) ⓘ Number of content items assigned a critical risk level in the past 24 hours. If this reaches the threshold, the kill switch auto-activates to prevent brand-damaging content from publishing.
0 / 3 threshold
Unverified Approved/Published (24h) ⓘ Content items that were approved or published in the last 24 hours despite their core statistic not being verified. Any count above the threshold triggers an auto-kill. Zero is the ideal target.
0 / 5 threshold
Unverified Published (7d) ⓘ Content published in the last 7 days that contained an unverified statistic. The policy is zero tolerance — any value above 0 is a breach. This is a rolling compliance metric.
0 / 0 allowed

Risk Level Distribution

ⓘ Risk is assigned to each content item based on the statistics it contains. Low: well-sourced, verified, low sensitivity. Medium: sourced but not independently verified. High: weak source, outdated, or disputed. Critical: unverifiable, legally sensitive, or brand-dangerous.
0
Low Risk ⓘ Verified statistics from trusted, primary sources. Safe to publish.
0
Medium Risk ⓘ Sourced but not independently corroborated. Review before publishing.
0
High Risk ⓘ Weak or outdated source, or disputed data. Requires manual sign-off.
0
Critical Risk ⓘ Unverifiable, legally sensitive, or brand-damaging. Blocked by default.

Quality Gate Metrics

ⓘ Live counts from the database showing how content is flowing through the quality pipeline. All numbers reflect current state, not estimates.
0%
Verification Pass Rate ⓘ Percentage of content items whose core statistic was successfully verified against a primary source. Target is >95%. Calculated from all items with a stat_verified value in the queue.
0
Stats Verified ⓘ Total content items in the queue where the statistic has been confirmed against its source URL. These items passed the fact-check gate.
0
Auto-Approved ⓘ Items that passed all quality gates automatically and are cleared for publishing without human review. Auto-approval requires: verified stat, low/medium risk, no duplicate detection flag.
0
Pending Review ⓘ Items awaiting manual review — either because they failed a quality gate, have high/critical risk, or were flagged by duplicate detection. These sit in the queue until approved or rejected.

Correction Log

ⓘ A permanent audit trail of every correction made to published content. Corrections are logged when a statistic was found to be wrong, outdated, or needed updating after publication. This log is append-only and cannot be edited.
0 recent entries

All clear — no corrections logged yet

Corrections appear here when published content needs fixing. A clean log means quality gates are working.

+ Log a Correction

Use this form to manually record a correction to already-published content. This does not modify the content itself — it creates an audit record only.