Skip to main content

​
Overview

Trendteller leverages OpenAI GPT-4 and Google Gemini AI to generate actionable insights, forecasts, and recommendations from your e-commerce data.

​
Types of AI Insights

Forecasts

Predictive analytics for future performance

Trend Detection

Emerging patterns and market shifts

Anomaly Detection

Unusual patterns requiring attention

Recommendations

Action-oriented business suggestions

​
Sales Forecasting

​
How It Works

1

Data Collection

AI analyzes 90+ days of historical sales data including:
  • Daily revenue and order counts
  • Seasonal patterns
  • Day-of-week effects
  • Marketing campaign impacts
2

Model Training

Multiple forecasting models are trained:
  • Time series analysis (ARIMA, Prophet)
  • Machine learning (Random Forest, XGBoost)
  • Deep learning (LSTM neural networks)
3

Ensemble Prediction

Models are combined for more accurate predictions:
  • Weighted average based on historical accuracy
  • Confidence intervals calculated
  • Best-case and worst-case scenarios
4

AI Narration

GPT-4 generates natural language explanation:
  • Key drivers of the forecast
  • Risk factors and opportunities
  • Recommended actions

​
Reading Forecasts

Forecast displays include: 
7-Day Revenue Forecast:
📈 $125,000 (±$15,000)

Confidence: 85%
Expected range: $110K - $140K

Key Factors:
✓ Strong weekend performance expected
✓ New product launch impact
⚠ Holiday period beginning (higher uncertainty)

Recommendation:
Consider increasing inventory for top 10 products
by 20% to meet anticipated demand.
Forecasts are regenerated daily with the latest data for maximum accuracy.

​
Trend Detection

AI identifies rising and declining products: AI detects shifts in customer behavior:
  • Purchase Frequency Changes: Customers buying more/less often
  • Basket Size Shifts: Average items per order changing
  • Category Preferences: Shifting product category interests
  • Price Sensitivity: Changes in discount response rates

​
Anomaly Detection

​
Unusual Patterns

AI automatically flags unusual patterns:
Detected anomalies:
  • Unexpected revenue spikes or drops (>2 standard deviations)
  • Order volume irregularities
  • Sudden AOV changes
  • Geographic concentration shifts
Alert Example: ”🚨 Revenue drop detected: Today’s revenue 8Kvs8K vs 15K expected (-47%). Potential causes: Payment gateway issues, website downtime, or inventory stock-outs.”
Stock issues:
  • Faster-than-expected depletion
  • Unusual stock accumulation
  • Inter-brand inventory imbalances
  • Supplier delay impacts
Alert Example: “⚠️ Product ABC stock depleting 3x faster than forecast. Current: 45 units. Days until stock-out: 3. Recommend emergency reorder.”
Behavioral changes:
  • Sudden churn increase
  • Unusual return rate spikes
  • Geographic demand shifts
  • Channel preference changes
Alert Example: ”📊 Return rate increased from 8% to 18% for Product XYZ. Common reason: ‘Size too small’. Recommend updating size guide.”

​
AI Recommendations

​
Types of Recommendations

Inventory

  • Restocking priorities
  • Quantity recommendations
  • Transfer between brands
  • Slow-moving item actions

Pricing

  • Competitive price adjustments
  • Promotional opportunities
  • Dynamic pricing suggestions
  • Margin optimization

Marketing

  • Target customer segments
  • Product bundling ideas
  • Campaign timing
  • Channel allocation

Product

  • New product opportunities
  • Product line extensions
  • SKU rationalization
  • Category expansion

​
Prioritized Action List

AI generates a prioritized list of recommended actions: 
Today's Top Recommendations:

1. [HIGH] Restock Product A (3 days until stock-out)
   Impact: $12K potential lost revenue
   Action: Order 200 units

2. [MEDIUM] Price adjustment for Product B
   Impact: +15% margin improvement
   Action: Increase price by $5 (still competitive)

3. [LOW] Marketing campaign for Category C
   Impact: +8% category revenue
   Action: Run 3-day promotional campaign

​
Customizing AI Insights

​
Insight Preferences

Configure what insights you want to see:
  • Frequency: Daily, weekly, or real-time
  • Channels: Dashboard, email, Slack, API
  • Thresholds: Anomaly sensitivity levels
  • Focus Areas: Prioritize certain metrics or categories

​
AI Training Feedback

Improve AI accuracy by providing feedback:
  • ✅ Mark helpful insights
  • ❌ Flag incorrect predictions
  • 📝 Add context for better understanding
  • 🎯 Rate recommendation effectiveness
Your feedback helps train the AI models specifically for your business patterns, improving accuracy over time.

​
Next Steps

Dashboard Guide

View AI insights in the dashboard

AI Workflows

Learn how AI insights are generated

API Reference

Access forecasts programmatically

Data Exports

Export AI insights and forecasts