What This Does

Quality Insights aggregates data from every piece of content you've graded. Instead of viewing quality article-by-article, you see patterns:

  • Which issues appear most frequently
  • Which fixes correlate with high-performing content
  • What the learning system has identified and is ready to act on
  • How quality scores trend over time

This dashboard is the control center for ILLIXIS's quality learning system. It shows what the system has learned about your content and lets you confirm or adjust automatic behaviors.

Finding Quality Insights

Navigate to Content Hub > Quality Insights from the main menu.


Dashboard Overview

The Quality Insights dashboard is organized into four sections:

Stats Cards (Top Row)

Four metric cards show your quality snapshot:

| Metric | What It Shows |
|--------|---------------|
| Issues (30d) | Total issues detected across all content in the last 30 days |
| Fix Rate | Percentage of detected issues you fixed vs. ignored |
| Active Preventions | Number of issue types being blocked in content generation |
| Suppressed | Number of issue types no longer detected |

What good looks like:

  • Fix rate: 60-80% (some issues should be ignored)
  • Active preventions: 5-12 after 2 months of use
  • Suppressed: 2-5 (issues genuinely irrelevant to your brand)

Ready to Learn Section

Shows issue types awaiting your confirmation. These are patterns the system has identified but won't act on without your approval.

Prevention Candidates: Issues you consistently fix. Enabling prevention adds an instruction to content generation prompts so future articles avoid this pattern.

  • Issue name and category
  • Times fixed (must be 10+)
  • Fix rate (must be 70%+)
  • "Enable Prevention" button

Suppression Candidates: Issues you consistently ignore. Enabling suppression stops detecting this issue type entirely.

  • Issue name
  • Times ignored (must be 10+)
  • Ignore rate (must be 80%+)
  • "Suppress Detection" button

Active Learnings Section

Shows currently enabled prevention and suppression rules. Each rule has a toggle switch.

Prevention rules (green): Issues being actively prevented in content generation. Toggle off to stop preventing.

Suppression rules (gray): Issues no longer being detected. Toggle off to resume detection.

Trend Charts

Two visualizations show quality over time:

Issues Over Time: Line chart of issues detected per day (30-day window). Should trend downward as learning improves.

Quality Grade Trends: Average content grade over time. Should trend upward (B- to B+ to A-).


Learning Controls

The learning system only acts with your explicit confirmation (except in extreme cases).

Confirming Prevention

When an issue reaches "Ready to Prevent" status:

  1. Review the issue type and description
  2. Ask: "Do I want to stop this from appearing in future content?"
  3. If yes, click Enable Prevention
  4. The system adds a prevention instruction to content generation prompts

Example: You've fixed "rhetorical questions" 15 times with a 90% fix rate. The system asks if you want to prevent rhetorical questions. Click "Enable Prevention" and future articles won't use them.

Confirming Suppression

When an issue reaches "Ready to Suppress" status:

  1. Review the issue type
  2. Ask: "Is this a false positive for my brand voice?"
  3. If yes, click Suppress Detection
  4. The system stops detecting this issue type

Example: You've ignored "passive voice" warnings 12 times with 100% ignore rate. Your academic tone intentionally uses passive voice. Click "Suppress Detection" to stop seeing these warnings.

Disabling Learnings

Toggle off any active learning to revert behavior:

  • Disable prevention: System stops adding the instruction to prompts. Future content may contain the issue again.
  • Disable suppression: System resumes detecting this issue type.

The system reverts to "collecting" status and waits for new patterns.


Batch Quality Analysis

Analyze multiple articles at once to quickly build learning data.

Running Batch Analysis

  1. Go to Content Hub > Quality Insights
  2. Click Run Batch Analysis
  3. Select content items to analyze (max 20 per batch)
  4. Click Analyze Selected

Each item is queued for quality grading. Analysis takes 2-3 seconds per article.

Use cases:

  • Analyze imported content after CMS connection
  • Re-grade content after changing brand voice settings
  • Build learning data faster by grading existing library

Monitoring Batch Progress

After starting batch analysis:

  • Progress bar shows completion percentage
  • Individual grades appear as they complete
  • Refresh the page to see updated statistics

GA4 Performance Correlation

Quality Insights correlates your fix decisions with content performance data from Google Analytics.

High-Impact Fix Types

The "Performance Correlation" section shows:

  • Issue types that, when fixed, correlate with higher pageviews
  • Average pageviews for content where each issue was fixed
  • Lift percentage vs. baseline average

Example:
| Issue Type | Avg Pageviews | Lift |
|------------|---------------|------|
| Weak Opening | 2,400 | +45% |
| Missing CTA | 1,900 | +22% |
| AI Clichés | 1,750 | +12% |

This tells you which fixes matter most for performance. Prioritize preventing issues with high lift.

How Correlation Works

Seven days after content publishes:

  1. System fetches GA4 pageviews for that content
  2. Calculates a performance weight (0.5 to 2.0)
  3. Updates resolution records with this weight
  4. Weighted statistics prioritize fixes on high-performing content

Result: The learning system trusts patterns from successful content more than underperforming content.


Quality Score Breakdown

Understanding how quality scores are calculated helps interpret the dashboard.

Score Categories

Each graded content receives scores in 7 categories:

| Category | Weight | What It Measures |
|----------|--------|------------------|
| Substance | 20% | Depth, accuracy, value |
| Voice | 15% | Brand consistency, tone |
| Structure | 15% | Organization, flow |
| Clarity | 15% | Readability, simplicity |
| Engagement | 15% | Hooks, retention |
| SEO | 10% | Keywords, meta data |
| Visual | 10% | Formatting, media |

Grade Distribution

Quality Insights shows grade distribution across your library:

| Grade | Score Range | Meaning |
|-------|-------------|---------|
| A+ | 95-100 | Exceptional |
| A | 90-94 | Excellent |
| A- | 85-89 | Very Good |
| B+ | 80-84 | Good |
| B | 75-79 | Above Average |
| B- | 70-74 | Average |
| C+ | 65-69 | Below Average |
| C | 60-64 | Needs Work |
| C- | Below 60 | Poor |


Trend Interpretation

Issues Declining

Good sign. Learning is working. Prevention rules are reducing issue occurrences.

Expected timeline:

  • Week 1-2: Baseline issue counts
  • Week 3-4: First preventions enabled
  • Month 2: 30-40% issue reduction
  • Month 3+: 50-60% issue reduction

Issues Flat or Increasing

Investigate. Either:

  • Learning not confirmed (check Ready to Learn section)
  • Prevention instructions not yet effective (try generating more content)
  • New content types introducing new issues

Grades Improving

Success indicator. Average grades trending from B- to B+ to A- shows learning is improving quality.

Grades Flat

Check fix rate. If you're ignoring most issues, the system can't learn what to prevent. Fix more issues or suppress false positives.


Common Questions

Q: How many articles before learning works? Minimum 10 resolutions per issue type. With regular content creation, expect first preventions after 2-3 weeks.

Q: What if I enabled prevention but issues still appear? Prevention takes effect on new content only. If issues persist after several new articles, the system may need more examples to learn from. Try fixing a few more instances to strengthen the pattern.

Q: Does learning work across all content types? Yes. Prevention instructions are injected into keyword briefs, trend briefs, custom content, and extensions.

Q: How do I reset learning? Toggle off all active learnings from the Quality Insights dashboard. The system reverts to "collecting" status. Historical data remains but stops influencing behavior.


Related Features

  • Content Quality Grading: How individual articles are scored
  • Learning Quality System: Detailed learning mechanics
  • Content Hub: Where graded content lives
  • Preference Learning: Separate system for content topic preferences

Ready to lose the stack?

One platform. You approve. ILLIXIS executes. Marketing that just happens.

Join the waitlistNo spam, everUnsubscribe anytime
First 20 founding members: 50% off any plan for your first year.

Marketing, Unstacked.

Quality Insights Dashboard | Help Center | ILLIXIS™ | ILLIXIS