🔮 Predictive Analytics
OmniSocial AI predicts future engagement and revenue before you publish, helping you make smarter decisions about what to post, when to post, and where to invest.
Engagement Predictions
Before publishing, the AI generates an EngagementPrediction for your post based on historical patterns and content analysis:
| Prediction Field | Description |
|---|---|
| Predicted Engagement | Expected total engagement (likes + comments + shares) |
| Predicted Reach | Estimated unique users who will see the post |
| Virality Score | Likelihood the post will be shared beyond your immediate audience (0–1 scale) |
| Confidence | How confident the model is in this prediction (0–1 scale) |
| Input Features | JSON of the signals used — content type, format, posting time, hashtag count, etc. |
Predictions are generated by analyzing your PostMetric history alongside PostClassification data. The model considers content type, format, topics, hook strength, caption length, hashtag usage, CTA presence, posting time, and audience activity patterns to produce a forecast.
Revenue Predictions
If you have an e-commerce store connected, the AI also generates RevenuePrediction data for product-related posts:
| Prediction Field | Description |
|---|---|
| Predicted Revenue | Expected revenue from this post |
| Predicted Conversions | Estimated number of purchases |
| Predicted ROAS | Expected return on ad spend (if the post will be boosted) |
| Confidence | Model confidence level (0–1 scale) |
This is especially powerful when combined with Auto-Boost rules — the system can predict which posts will generate revenue and automatically allocate ad budget to them.
Prediction Accuracy Tracking
Predictions are only useful if they're reliable. The PredictionAccuracy model tracks how well the system is performing:
| Metric | Description |
|---|---|
| Total Predictions | Number of predictions made in the tracking period |
| Avg Accuracy | Average accuracy score across all predictions |
| Median Error | Median deviation between predicted and actual values |
| P90 Error | 90th percentile error — worst-case accuracy boundary |
After a post is published, the system records actualEngagement and calculates an accuracyScore for each individual prediction. This feedback loop continuously improves the model.
Check Analytics > Predictions to see your prediction accuracy trends. A consistently high accuracy score (above 0.7) means you can trust the forecasts for decision-making. If accuracy is lower, the model may need more historical data — keep publishing and it will improve.
How to Use Predictions
1. Pre-Publish Optimization
Compare predicted performance across post variations before choosing which to publish. The post editor shows prediction scores for your draft content.
2. Identify High-Potential Content
Posts with high predicted engagement and virality scores are prime candidates for paid promotion. Don't wait for a post to go viral organically — amplify it proactively.
3. Content Calendar Planning
Plan your content calendar around predicted peak performance windows. The AI Advisor uses prediction data to suggest optimal posting schedules.
4. Budget Allocation
Use revenue predictions to decide where to allocate your ad budget. Posts with high predicted ROAS deserve more spend.
Prediction vs. Reality
Every prediction is paired with actual results after the fact:
EngagementPrediction {
predictedEngagement: 1,250
predictedReach: 15,000
viralityScore: 0.72
confidence: 0.85
actualEngagement: 1,180 // ← filled in after post goes live
accuracyScore: 0.94 // ← calculated automatically
}
This transparency lets you calibrate your trust in the system and understand where predictions are strongest (and weakest).
Related Pages
- Dashboard — See predictions alongside actual performance
- A/B Testing — Test predicted winners against alternatives
- AI Advisor — Recommendations informed by prediction data
- Auto-Boost & Ads — Automatically promote high-prediction posts