Prediction Accuracy Tracking

What It Does

Tracks how accurately ILLIXIS predicts content performance before publication. Every brief generates a prediction of traffic and rankings. After publication, actual Google Search Console data validates these predictions at 30, 60, and 90 day checkpoints.

Purpose: Measure and improve the platform's ability to predict content success.

How Predictions Work

1. Prediction Creation (Automatic)

When a brief is generated, ILLIXIS automatically creates a prediction based on:

• Search volume - Monthly search volume for target keyword
• Keyword difficulty - How hard it is to rank (0-100)
• Opportunity score - Brief quality score (0-100)
• Competition density - SERP competition level
• SERP data - Top 10 results analysis

Predicted at 3 checkpoints:

• 30 days after publication
• 60 days after publication
• 90 days after publication

2. Prediction Storage

Each prediction includes:

• Estimated traffic - Expected clicks at 30/60/90 days
• Estimated position - Expected ranking position at 30/60/90 days
• Confidence score - How confident ILLIXIS is (0-100%)
• Target position - Goal ranking position (1-20)

3. Publication Tracking

When content is published to your CMS:

• Prediction status changes from "Pending Publication" to "Published"
• Publication date recorded
• Validation schedule set (30/60/90 days from publication)

4. Validation (Automatic)

Daily at 7 AM, ILLIXIS checks for predictions due for validation:

For each due prediction:

1. Fetch actual GSC data (clicks, impressions, position)
2. Compare predicted vs actual metrics
3. Calculate accuracy percentage
4. Update prediction status
5. Store result

Data quality assessment:

• Good - Sufficient data (100+ impressions)
• Partial - Limited data (< 100 impressions)
• Insufficient - No GSC data found

5. Accuracy Calculation

Traffic Accuracy: How close the predicted traffic was to actual traffic.

Position Accuracy: How close the predicted ranking position was to the actual position.

Overall Accuracy: The average of traffic accuracy and position accuracy.

Accuracy ratings:

• ≥70% = Accurate (Win)
• 50-69% = Close
• <50% = Missed

Dashboard Overview

Hero Narrative

Shows high-level accuracy summary:

• Headline - "ILLIXIS predicted X out of Y correctly"
• Sub-headline - Accuracy rate percentage
• Trend badge - Accuracy improving/declining vs previous weeks

Win/Loss Record Bar

Visual breakdown of all validated predictions:

• Green - Accurate predictions (≥70%)
• Yellow - Close predictions (50-69%)
• Red - Missed predictions (<50%)

Predicted vs Actual Cards

Shows up to 6 recent predictions with side-by-side comparison:

• Target keyword
• Predicted position → Actual position
• Predicted traffic → Actual traffic
• Accuracy percentage
• Verdict (Win/Close/Miss)
• Validation period (30/60/90 days)
• Confidence score

Accuracy Trend Chart

Line chart showing accuracy over last 12 weeks:

• X-axis: Week (W1, W2, W3...)
• Y-axis: Accuracy percentage (0-100%)
• Shows if predictions are getting better over time

Confidence Calibration Chart

Bar chart comparing confidence levels to actual outcomes:

• What it shows: Are high-confidence predictions actually accurate?
• Goal: Bars should match height (well-calibrated predictions)
• Example: If 80-90% confidence bucket has 85% actual accuracy → well-calibrated

Requires: At least 5 validated predictions

Upcoming Validations Timeline

Shows next 8 predictions due for validation:

• Days until validation
• Target keyword
• Validation period (30/60/90 days)
• Confidence score

Accuracy by Content Type

Horizontal bar chart showing accuracy by brief type:

• Keyword Brief
• Topic Brief
• Custom Brief
• Calendar Brief

Shows which types of content are most predictable.

Key Metrics Summary

Overall stats:

• Overall Accuracy - Average across all predictions
• Traffic Accuracy - How well traffic is predicted
• Position Accuracy - How well rankings are predicted
• Total Predictions - All predictions created
• Validated - Predictions with validation results

All Predictions Table

Sortable table of all predictions:

• Target keyword
• Content type
• Status (Pending/Published/Validated/Complete)
• Estimated 30-day traffic
• Confidence score
• Accuracy percentage (if validated)

Click any row to see full prediction details in modal.

Manual Operations

Run Validation (Button)

Forces immediate validation check for all predictions due today.

Use when:

• You just published content and want to see if predictions are due
• Daily validation hasn't run yet
• Testing prediction accuracy

Process:

1. Click "Run Validation" button
2. Background task processes the request
3. Checks 30/60/90 day due dates
4. Fetches GSC data for each due prediction
5. Calculates and stores accuracy results
6. Page auto-refreshes after 3 seconds

Prediction Detail Modal

Click any prediction row or comparison card to see full details:

Top section:

• Target keyword
• Content type
• Status
• Search volume
• Confidence score
• Keyword difficulty
• Publication date

Validation Results section:

For each validation period (30/60/90 days):

• Period label
• Predicted position → Actual position
• Predicted traffic → Actual traffic
• Accuracy percentage (color-coded)
• Data quality indicator

Empty state: "No validation results yet. This prediction is awaiting validation."

Prediction Lifecycle

1. Brief Created — Prediction record created (status: pending publication)
2. Content Published — Status changes to published, validation dates are set
3. 30 Days After Publication — First validation runs, 30-day results recorded
4. 60 Days After Publication — Second validation runs, 60-day results recorded
5. 90 Days After Publication — Final validation runs, prediction marked complete
6. Stats Updated — Weekly accuracy history and best/worst predictions recalculated

ML Learning Over Time

Accuracy Stats (Automatic)

After each validation, your accuracy stats are updated with:

Overall metrics:

• Average accuracy across all predictions
• Traffic accuracy average
• Position accuracy average
• Validated prediction count

Accuracy by type:

• Accuracy breakdown per content type
• Identifies which brief types are most predictable

Best predictions:

• Top 5 most accurate predictions
• Used to identify winning patterns

Learning opportunities:

• Bottom 5 least accurate predictions
• Used to improve future predictions

Weekly history:

• Last 12 weeks of accuracy scores
• Tracks improvement over time

Improvement Cycle

1. Brief creation - Platform makes prediction based on current model
2. Content published - Real-world test begins
3. Validation - Actual GSC data compared to prediction
4. Learning - Accuracy patterns analyzed
5. Model refinement - Future predictions adjusted based on learnings

Goal: Predictions get more accurate as the platform learns your niche, audience, and competitive landscape.

Understanding Confidence Scores

Confidence score (0-100%) indicates how certain ILLIXIS is about the prediction.

Higher confidence when:

• ✅ Search volume 100-10,000 (sweet spot)
• ✅ Keyword difficulty 20-50 (achievable)
• ✅ Opportunity score ≥70 (high quality brief)
• ✅ SERP data available (competitive analysis complete)

Lower confidence when:

• ❌ Very low search volume (<50)
• ❌ Very high difficulty (>70)
• ❌ Low opportunity score (<30)
• ❌ No SERP data

Example:

• 85% confidence = "Very likely to rank as predicted"
• 60% confidence = "Moderate uncertainty"
• 40% confidence = "High uncertainty, use prediction as rough guide"

Data Sources

Prediction Inputs (From Brief)

• Search volume from keyword research
• Keyword difficulty from SEO data analysis
• Opportunity score from brief analysis
• Competition density from SERP analysis
• SERP data from search engine intelligence

Validation Inputs (Actual Performance)

• Clicks, impressions, position, and CTR from Google Search Console

Date range: Publication date to validation date (30/60/90 days)

Automation Schedule

| Task | Schedule | What It Does |
|------|----------|--------------|
| Daily Validation | 7:00 AM UTC | Validates predictions due at 30/60/90 day checkpoints |
| Weekly Stats Update | Sunday 6:00 AM UTC | Recalculates accuracy statistics for all accounts |

Predictions are validated automatically—no manual action required. The system checks daily for predictions that have reached their 30, 60, or 90 day post-publication checkpoint.

Common Scenarios

Scenario: No Predictions Yet

Dashboard shows:

• "Your prediction engine is warming up"
• Empty state in All Predictions table
• Explanation: "Predictions are created automatically when briefs are generated"

To get predictions:

1. Generate keyword briefs (Strategy Hub → Approve opportunities)
2. Predictions created automatically with each brief
3. Dashboard populates immediately

Scenario: Predictions but No Validations

Dashboard shows:

• Total predictions count
• "Validated: 0"
• Upcoming Validations timeline

Why:

• Content not yet published, OR
• Content published less than 30 days ago

To get validations:

1. Publish content from briefs
2. Wait 30 days
3. Daily validation task runs automatically
4. Results appear on dashboard

Scenario: Insufficient Data

Result record shows:

• Data quality: "Insufficient"
• Note: "No GSC data found for this keyword"
• Accuracy: Not calculated

Causes:

• Content not indexed by Google yet
• Keyword not receiving impressions
• GSC not connected or syncing

Fix:

1. Verify GSC connection (Analytics → Data Sources)
2. Run GSC data sync manually
3. Wait for Google to index content
4. Re-run validation after content has traffic

Scenario: Low Accuracy (<50%)

Possible reasons:

• Keyword difficulty higher than expected - Took longer to rank
• Search volume fluctuated - Seasonal changes or trend died
• Content quality lower than predicted - Didn't match search intent
• SERP landscape changed - New competitors entered
• No backlinks - Content needs link building

Action:

• Check "Learning Opportunities" section
• Review common factors in low-accuracy predictions
• Adjust future content strategy

Scenario: High Accuracy (≥70%)

What it means:

• Platform understands your niche
• Brief quality is high
• Predictions are reliable

Action:

• Check "Best Predictions" section
• Identify winning patterns (keywords, topics, formats)
• Create more content following these patterns

FAQ

Q: When are predictions created? A: Automatically when any brief is generated (keyword, trend, custom, calendar).

Q: Can I create predictions manually? A: No. Predictions are tied to briefs and created during brief generation.

Q: Why is my prediction still "Pending Publication"? A: Content hasn't been published to your CMS yet. Publish the content to activate validation schedule.

Q: Can I delete a prediction? A: No. Predictions are permanent for accuracy tracking. They can be marked "Insufficient Data" if validation fails.

Q: Why don't I see accuracy for recent predictions? A: Predictions are validated 30 days after publication. Check "Upcoming Validations" to see when results will appear.

Q: What's a good accuracy percentage? A: 70%+ is excellent. 50-69% is acceptable. <50% indicates prediction model needs adjustment for your niche.

Q: Does accuracy improve over time? A: Yes. The platform learns from validated predictions and adjusts future estimates based on your actual performance patterns.

Q: What happens at 90 days? A: Final validation runs. Prediction status changes to "Complete". No further validations scheduled.

Q: Can I re-validate a prediction? A: Not currently. Each prediction validates once at 30, 60, and 90 days.

Q: Where do I see prediction details for a specific brief? A: Click any prediction in the All Predictions table, or view Brief Detail page.

Q: Why are some predictions missing validation results? A: Either validation not due yet, or GSC data insufficient (content not indexed/no traffic).

Next Steps After Reviewing Predictions

1. If accuracy is high (≥70%):

• Check best predictions for patterns
• Create more briefs targeting similar keywords
• Rely on predictions for content planning

2. If accuracy is low (<50%):

• Review learning opportunities section
• Check common factors (difficulty, volume)
• Adjust strategy (target easier keywords, longer content, etc.)

3. If insufficient data:

• Verify GSC connection
• Check content indexing status
• Ensure content is receiving traffic

4. For upcoming validations:

• Note which predictions validate soon
• Check actual rankings in GSC
• Compare predicted vs actual proactively

Related Help Guides:

• Strategy Hub - Where opportunities become briefs (which create predictions)
• Content Briefs - Understanding brief creation
• Google Search Console Integration - GSC data source for validations
• Analytics Dashboard - See actual content performance