Best-Case Projection Bias

Issue: Sales forecasting model uses deal probability from salesperson input, which is inherently optimistic; projects deals that salesperson hopes will close rather than actual likelihood

Frequency: Common

Symptoms

• Pipeline shows $5M in “likely” deals next quarter
• Actual closed deals: $2M (40% variance)
• Sales team consistently over-optimistic by 40-60%
• Quarterly revenue targets miss due to forecast inaccuracy

Root Cause Salespeople are incentivized to be optimistic; they estimate deal probability. Models trust input probability estimates without adjusting for optimism bias. Salespeople genuinely believe in their deals but lack statistical calibration—they think 70% probability when reality is 30%.

Example

Sales pipeline forecast:
- Deal A: $1M, salesperson says "90% close probability"
- Deal B: $500k, salesperson says "85%"
- Deal C: $400k, salesperson says "80%"
Total forecast: $1.9M at 85% avg probability = $1.6M expected value
Actual outcome:
- Deal A: LOST (salesperson misunderstood buyer intent)
- Deal B: CLOSED ($500k)
- Deal C: LOST (budget cut by buyer)
Actual revenue: $500k vs forecasted $1.6M
Impact: Revenue miss; stock price impact; forecast unreliable

Key Statistics

Mitigation Strategies

1. Probability calibration: Adjust salesperson estimates downward (e.g., multiply by 0.6)
2. Historical accuracy tracking: Compare actual close rates to predicted; adjust model
3. Independent scoring: Have second person (manager/data analyst) estimate independent of salesperson

Production Signals

Alerts

References

• Sales Forecast Accuracy - Research on bias
• Probability Calibration in Forecasting - Calibration methods