Ikigai Logoikigai

Model Performance Studio
Experiment Mode

Evaluate and backtest the active forecasting model against challengers

Evaluation Period: Last 12 Months
Live

Select a model to benchmark against

80% / 20%
Training DataTest Data

Active Model

Base + Promo/Pricing

0WMAPE

Performing within tolerance

Comparison

Base

0WMAPE
Active is worse by 0.0 pts

Sensei Insight

The Active Model demonstrates superior stability on out-of-sample data. Comparison model shows high variance during promo periods.

Backtesting Results
Comparing actuals vs predicted values over the evaluation period
Test Region
  • Actuals
  • Active: Base + Promo/Pricing
  • Comparison: Base
Error Distribution (Out-of-Sample)
Model Bias Trend

Bias analysis requires granular segment selection