AI summary by TwinMind
Unit 2: the Forecast is always wrong, Video 1: Cost Estimates
MIT OpenCourseWare
9m 54s
Introduction to Forecasting Problems
- Topic: "The forecast is always wrong" (in quotes - acknowledges forecasts can match reality but rarely do)
- Objective: Provide realistic assessment of forecasting ability accuracy
- Evidence shows forecasts for system performance 5-15 years (or even 6 months-1 year for fast technology) differ substantially from actual outcomes
Runway Resurfacing Case Study - Simplest Technology Example
- Analysis of ratio: real cost vs estimated cost
- Graph shows percent of observations (0-15%) vs cost ratio
- Perfect forecast would show ratio of 1.0
- Results:
- Range: 0.5 to 2.5 ratio (factor of 5 spread)
- Median value: 1.25 (most frequent)
- Ability to predict actual cost off by factor of 2 (half or double)
- Process description:
- Simple task: layer asphalt over runway after aircraft damage from repeated landings
- Technology: drive truck, dump asphalt, roll down
- Volume calculation straightforward: width × length × depth
Cost Variation Drivers
- Energy prices major factor:
- Oil prices varied from $15-$150 per barrel over 20 years
- Factor of 8-10 variation
- Affects asphalt, cement, steel production (energy-intensive processes)
- Electricity price changes similarly
- Labor cost changes:
- High economic activity periods: overtime pay required
- Recessions/bad economic times: low bids received
- Conclusion: Largely driven by factors beyond control
General Project Cost Analysis
- Survey of various project types (highways, ordinary activities)
- Results:
- Average cost vs predicted cost ratio up to 2.5
- Standard deviation: ~0.5 (±50% variation)
- Common experience: projected costs consistently under actual costs
- Large variations both up and down
NASA High-Tech Project Analysis
- National Aeronautics and Space Administration projects
- High-tech, innovative, less usual activities
- Results:
- Cost variation much higher than ordinary projects
- Many projects reached ~2× estimated cost
- Standard deviation: >60%
- More variation than standard projects
Key Conclusions
- Forecasts consistently wrong across all project types
- Not due to expert incompetence but world changes/external factors
- Need to account for forecasting limitations in project estimation and management
- Higher technology/innovation correlates with greater forecast uncertainty
