Better Data

Plan

Forecast Accuracy

Edit this page

Forecast Accuracy

Forecast accuracy measures how well your forecasts match actual demand. Understanding and improving accuracy is critical for effective planning.

Planner (Forecasting + Replenishment)

Start here:

  1. Accuracy Metrics - Review MAPE, WAPE, and bias metrics
  2. Forecast vs Actuals - Compare forecasts to actual demand
  3. Deviation Analysis - Identify products/channels with poor accuracy
  4. Improvement Actions - Take actions to improve accuracy

Daily workflow:

  • Review forecast accuracy metrics
  • Compare forecasts to actuals
  • Identify accuracy issues
  • Adjust forecast models and events
  • Track accuracy improvements

Finance (COGS, Holding Cost, Cost-to-Serve)

Start here:

  1. Accuracy Reports - Generate forecast accuracy reports
  2. Cost Impact - Analyze forecast accuracy impact on costs
  3. Bias Analysis - Review forecast bias trends
  4. Accuracy Trends - Monitor accuracy over time

Daily workflow:

  • Review accuracy metrics and trends
  • Analyze cost impact of forecast errors
  • Monitor bias trends (over/under forecasting)
  • Generate accuracy reports for analysis

Executive (Summary Dashboards)

Start here:

  1. Accuracy Dashboard - High-level accuracy overview
  2. Accuracy KPIs - Key accuracy performance indicators
  3. Accuracy Trends - Long-term accuracy trends

Daily workflow:

  • Review accuracy dashboard summaries
  • Monitor accuracy KPIs
  • Track accuracy trends over time
  • Review accuracy improvement initiatives

Accuracy Metrics Explained

MAPE (Mean Absolute Percentage Error)

Average percentage error across all forecasts:

1MAPE = (1/n) × Σ|Actual - Forecast| / Actual × 100

Interpretation:

  • < 10%: Excellent accuracy
  • 10-20%: Good accuracy
  • 20-30%: Fair accuracy
  • > 30%: Poor accuracy (needs improvement)

Use cases:

  • Overall forecast accuracy assessment
  • Comparing different forecast models
  • Tracking accuracy improvements

WAPE (Weighted Absolute Percentage Error)

Weighted average percentage error, giving more weight to high-volume items:

1WAPE = Σ|Actual - Forecast| / ΣActual × 100

Interpretation:

  • Similar to MAPE but weighted by volume
  • Better reflects impact on high-volume items
  • More relevant for inventory planning

Use cases:

  • Accuracy assessment for high-volume items
  • Inventory planning accuracy
  • Cost impact analysis

Bias

Measures whether forecasts are consistently over or under:

1Bias = (Σ(Forecast - Actual) / ΣActual) × 100

Interpretation:

  • Positive Bias: Over-forecasting (forecast > actual)
  • Negative Bias: Under-forecasting (forecast < actual)
  • Zero Bias: Balanced forecasting (ideal)

Use cases:

  • Identifying systematic over/under forecasting
  • Adjusting forecast models
  • Balancing inventory levels

Hit Rate

Percentage of forecasts within acceptable tolerance:

  • Within 10%: Forecasts within 10% of actual
  • Within 20%: Forecasts within 20% of actual
  • Within 50%: Forecasts within 50% of actual

Use cases:

  • Service level assessment
  • Planning confidence levels
  • Risk assessment

Step-by-Step: Measuring Forecast Accuracy

1. View Accuracy Metrics

  1. Navigate to Plan → Forecast Accuracy
  2. Select:
    • Forecast Scenario: Scenario to analyze
    • Time Period: Date range for analysis
    • Products: Products to analyze (or all)
    • Channels: Channels to analyze (or all)
    • Locations: Locations to analyze (or all)

2. Review Overall Metrics

View overall accuracy:

  • MAPE: Mean Absolute Percentage Error
  • WAPE: Weighted Absolute Percentage Error
  • Bias: Forecast bias
  • Hit Rate: Percentage within tolerance

3. Analyze by Dimension

Analyze accuracy by:

  • Product: Accuracy per product (identify problem products)
  • Channel: Accuracy per channel (identify problem channels)
  • Location: Accuracy per location (identify problem locations)
  • Time Period: Accuracy over time (identify trends)

4. Review Forecast vs Actuals

  1. Navigate to Analyze → Forecast vs Actuals
  2. View side-by-side comparison:
    • Forecast values
    • Actual values
    • Deviation
    • Percentage error

5. Identify Problem Areas

Identify:

  • Top Under-Forecast: Products/channels consistently under-forecast
  • Top Over-Forecast: Products/channels consistently over-forecast
  • High MAPE Items: Items with poor accuracy
  • Trending Issues: Accuracy getting worse over time

Forecast vs Actuals Analysis

Viewing Forecast vs Actuals

  1. Navigate to Analyze → Forecast vs Actuals
  2. Configure:
    • Forecast Scenario: Scenario to compare
    • Time Period: Date range
    • Products: Products to compare
    • Channels: Channels to compare
  3. View comparison:
    • Forecast: Forecasted demand
    • Actual: Actual demand
    • Deviation: Difference (Actual - Forecast)
    • Percentage Error: Percentage deviation

Understanding Deviations

Positive Deviation (Actual > Forecast):

  • Under-forecasted demand
  • Risk of stockout
  • May indicate missed events or trends

Negative Deviation (Actual < Forecast):

  • Over-forecasted demand
  • Risk of overstock
  • May indicate events didn't have expected impact

Deviation Alerts

System can alert when deviations exceed thresholds:

  • Alert Threshold: Configurable (e.g., ±20%)
  • Alert Frequency: Configurable (e.g., daily, weekly)
  • Alert Delivery: Email, webhook, Slack

Note: Push alerts may be work in progress. Check alert configuration for current status.

Improving Forecast Accuracy

1. Apply Events Properly

Action: Ensure all relevant events are applied:

  • Promotions
  • Seasonality
  • Marketing campaigns
  • External factors

Impact: Events can significantly improve accuracy when properly applied.

2. Adjust Forecast Models

Action: Try different forecast models:

  • Time series models
  • Seasonal models
  • Trend models
  • Custom models

Impact: Different models work better for different products/channels.

3. Review Historical Patterns

Action: Analyze historical demand patterns:

  • Identify trends
  • Recognize seasonality
  • Understand cycles
  • Account for anomalies

Impact: Better understanding of patterns improves forecasts.

4. Address Bias

Action: If bias is consistently positive or negative:

  • Positive Bias (Over-forecasting): Reduce forecast values
  • Negative Bias (Under-forecasting): Increase forecast values
  • Adjust forecast model parameters
  • Review event applications

Impact: Reducing bias improves overall accuracy.

5. Focus on Problem Areas

Action: Prioritize improving accuracy for:

  • High-volume items (biggest impact)
  • High-value items (cost impact)
  • Problem products/channels (biggest improvement opportunity)

Impact: Focused improvements yield better results.

What Success Looks Like

Planner Success

  • ✅ MAPE below 20% for most products
  • ✅ Bias near zero (balanced forecasting)
  • ✅ Accuracy improving over time
  • ✅ Problem areas identified and addressed
  • ✅ Forecasts trusted for planning

Finance Success

  • ✅ Accuracy metrics tracked regularly
  • ✅ Cost impact of forecast errors understood
  • ✅ Bias trends monitored
  • ✅ Accuracy reports generated for analysis

Executive Success

  • ✅ Accuracy meets targets
  • ✅ Accuracy trends positive
  • ✅ Accuracy improvements demonstrated
  • ✅ Forecasts drive confident decisions

Common Pitfalls

1. Not Measuring Accuracy

Problem: Creating forecasts but not measuring how accurate they are.

Solution: Regularly measure forecast accuracy:

  • Set up accuracy reviews
  • Track accuracy metrics
  • Compare forecasts to actuals
  • Monitor accuracy trends

How to avoid: Make accuracy measurement part of forecast workflow.

2. Ignoring Bias

Problem: Focusing only on MAPE without addressing bias.

Solution: Monitor and address bias:

  • Review bias metrics regularly
  • Identify systematic over/under forecasting
  • Adjust forecast models to reduce bias
  • Track bias trends

How to avoid: Include bias in accuracy reviews.

3. Not Acting on Accuracy Insights

Problem: Measuring accuracy but not taking action to improve.

Solution: Take action based on accuracy insights:

  • Identify problem products/channels
  • Adjust forecast models
  • Apply events properly
  • Focus improvement efforts

How to avoid: Make accuracy improvement part of planning process.

4. Not Understanding Channel Policy Effects

Problem: Not understanding how channel policies affect forecast interpretation.

Solution: Understand channel policy effects:

  • Review ATP policies
  • Understand replenishment rules
  • Consider alert thresholds
  • Account for channel allocations

How to avoid: Familiarize yourself with channel policy configuration.

Troubleshooting

Accuracy Metrics Not Available

Symptoms: Forecast accuracy metrics not showing or unavailable.

Possible causes:

  1. Forecasts not generated
  2. Actuals data not available
  3. Time period mismatch
  4. System calculation issue

Steps to resolve:

  1. Verify forecasts are generated
  2. Check actuals data availability
  3. Verify time period alignment
  4. Review calculation settings
  5. Contact support if issue persists

Accuracy Consistently Poor

Symptoms: Forecast accuracy (MAPE) consistently high across products/channels.

Possible causes:

  1. Events not applied
  2. Forecast models not appropriate
  3. Insufficient historical data
  4. External factors not accounted for
  5. Data quality issues

Steps to resolve:

  1. Review and apply relevant events
  2. Try different forecast models
  3. Verify sufficient historical data
  4. Account for external factors
  5. Check data quality
  6. Adjust forecast parameters
  7. Monitor accuracy improvements

Bias Not Improving

Symptoms: Forecast bias remains consistently positive or negative.

Possible causes:

  1. Forecast model parameters not adjusted
  2. Events not properly applied
  3. Systematic data issues
  4. Model not appropriate for data

Steps to resolve:

  1. Adjust forecast model parameters
  2. Review event applications
  3. Check for data quality issues
  4. Try different forecast models
  5. Monitor bias trends

Deviation Alerts Not Working

Symptoms: Forecasts deviate significantly but alerts not triggering.

Possible causes:

  1. Alert thresholds not configured
  2. Alert policies not enabled
  3. Deviation calculation issue
  4. Alert delivery issue
  5. Push alerts work in progress

Steps to resolve:

  1. Check alert threshold configuration
  2. Verify alert policies are enabled
  3. Review deviation calculations
  4. Check alert delivery settings
  5. Verify push alert status (may be WIP)
  6. Test alert functionality

Permissions & Roles

Viewing forecast accuracy requires standard user permissions. All accuracy metrics are scoped to your organization.