Skip to main content

Overview

Demand Prediction uses machine learning to analyze your historical sales data and forecast future vehicle demand by segment. This powerful feature helps you make data-driven decisions about inventory management, pricing, and purchasing strategies.

How It Works

SGIVU’s demand prediction system analyzes:
  • Historical sales contracts and trends
  • Vehicle characteristics (type, make, model, year, price range)
  • Seasonal patterns and market conditions
  • Customer preferences and buying patterns
The machine learning model processes this data to predict expected demand for different vehicle segments, helping you optimize your inventory composition.

Getting Predictions

Requesting a Prediction

1

Navigate to Demand Prediction

Access the Demand Prediction section from your dashboard or analytics menu.
2

Select Vehicle Segment

Choose the vehicle characteristics you want to predict demand for:
  • Vehicle type (car or motorcycle)
  • Make and model
  • Year range
  • Price range
  • Other relevant features
3

Set Time Period

Specify the time period for your prediction (next month, next quarter, next year).
4

Generate Prediction

Click to generate the prediction. The system sends your parameters to the ML service and returns the forecasted demand.
5

Review Results

Analyze the prediction results, including:
  • Expected demand volume
  • Confidence level
  • Historical trends
  • Recommendations

Understanding Prediction Results

Demand Estimate

The system provides an estimated number of units expected to sell in your specified segment and timeframe. This is based on historical patterns and current trends.

Confidence Level

Each prediction includes a confidence score indicating how reliable the forecast is based on data quality and historical consistency.

Historical Context

View historical sales data for the same segment to understand trends and validate predictions against past performance.
For more detailed analysis, use the “Prediction with History” feature:
1

Request Detailed Prediction

Select the “Prediction with History” option when generating forecasts.
2

View Historical Data

The system retrieves historical sales data for your selected segment, showing past performance over time.
3

Compare Trends

Visualize the historical trend alongside the future prediction to identify patterns and validate the forecast.
4

Identify Opportunities

Use the combined view to spot market opportunities, seasonal patterns, and optimal inventory levels.

Use Cases

Inventory Optimization

Stock Planning

Predict which vehicle segments will be in high demand to guide your purchasing decisions.

Avoid Overstocking

Identify slow-moving segments to avoid tying up capital in low-demand inventory.

Seasonal Preparation

Anticipate seasonal demand changes and adjust inventory accordingly.

Budget Allocation

Allocate purchasing budget to high-demand segments for better ROI.

Pricing Strategy

Dynamic Pricing

Use demand predictions to inform pricing decisions. High predicted demand may support premium pricing, while low demand segments may need competitive pricing.

Sales Planning

Sales Targets

Set realistic sales targets based on predicted demand for different vehicle segments.

Marketing Focus

Target marketing efforts toward high-demand segments predicted by the system.

Model Training

The prediction system continuously improves through model retraining:

Automatic Training

Continuous Learning

The ML model is periodically retrained with your latest sales data, ensuring predictions remain accurate as your business evolves.

Manual Retraining

1

Request Retraining

Access the model management section and request a manual retrain if you have significant new data or market changes.
2

Specify Data Range

Optionally specify a date range for training data. By default, the system uses all available historical data.
3

Training Process

The system retrains the model using your latest contract data. This process may take several minutes depending on data volume.
4

Model Validation

The new model is validated against test data to ensure improved accuracy before being deployed.
5

Deployment

Once validated, the new model automatically becomes active and is used for all future predictions.

Model Information

Viewing Model Details

Model Metadata

Access information about the current prediction model, including:
  • Training date and version
  • Data volume used for training
  • Model performance metrics
  • Feature importance

Best Practices

Generate predictions regularly (monthly or quarterly) to stay ahead of market trends and adjust inventory proactively.
Track prediction accuracy by comparing forecasts with actual sales. This helps you understand model reliability and refine your strategies.
Generate predictions for various vehicle segments to get a comprehensive view of your entire inventory needs.
Remember that predictions are based on historical data. Consider external factors like economic changes, new regulations, or market disruptions.
Use demand predictions alongside traditional market research and industry insights for the most informed decisions.

Data Requirements

Minimum Data

Historical Sales Data

The prediction system requires sufficient historical sales contracts to generate accurate forecasts. More data generally leads to more reliable predictions.

Data Quality

Complete Contracts

Ensure contracts have complete vehicle and sale information for better predictions.

Accurate Dates

Correct transaction dates are crucial for identifying trends and patterns.

Consistent Categories

Use consistent vehicle categorization (type, make, model) for meaningful segment analysis.

Regular Updates

Keep contract data current by regularly entering new transactions.

Integration with Other Features

Purchase & Sales Contracts

Training Data Source

The prediction system uses your contract history as training data. More complete contract records lead to better predictions.

Vehicle Management

Inventory Planning

Use predictions to inform vehicle purchasing decisions and inventory composition in Vehicle Management.

Technical Details

SGIVU uses advanced machine learning algorithms including scikit-learn and optionally XGBoost for demand forecasting. The models are trained on your specific data for customized predictions.
The system automatically extracts relevant features from your contract data including temporal patterns, vehicle characteristics, and transaction trends.
Each trained model is versioned and stored, allowing you to track model improvements over time and roll back if needed.
The prediction service exposes REST APIs that can be integrated with other systems or custom applications.

Troubleshooting

This usually means no model has been trained yet. Ensure you have sufficient historical sales data and trigger a model training.
Low confidence may indicate insufficient historical data for the specific segment. Try broader segment definitions or accumulate more sales history.
Compare predictions with actual results. If consistently inaccurate, retrain the model with more recent data or review your contract data quality.
Verify you have sufficient and valid historical data. Check system logs or contact your administrator if training errors persist.
Very complex predictions may take time to process. Simplify your segment criteria or try again during off-peak hours.

Future Enhancements

The demand prediction system is continuously evolving. Future enhancements may include:
  • External market data integration
  • Competitor pricing analysis
  • Economic indicator incorporation
  • Multi-location demand forecasting
  • Automated inventory recommendations
  • Real-time prediction updates

Build docs developers (and LLMs) love