Description
Reduce optimistic completion-time estimates by making people tie past completion times to the current task, or by asking uninvolved observers to forecast.
Planners take an inside view built on success scenarios and discount their own history of lateness. In Buehler, Griffin and Ross's studies even 99 per cent confidence estimates were too early, and students finished theses 22 days later than predicted and seven days later than their own worst case. Observers given the plan and past records predicted later and more accurately. Simply asking for aggregate frequencies or pessimistic scenarios did not help. The recall-relevance procedure, in which people state when they would finish if they finished as typically as on similar past tasks and write a scenario leading to that time, eliminated the optimistic bias in one study. It is a cheap debiasing protocol for research and design project estimates.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Unmigrated
- Stage
- Strategy
- Data
- Analytical / expert
- Effort
- Hours
Steps
- Ask the estimator to recall how far before or after deadlines similar past tasks were actually finished.
- Ask when this task would be finished if it went as those past tasks typically did.
- Ask for a plausible scenario, grounded in that experience, leading to the typical finishing time.
- Only then elicit the estimate.
- Optionally obtain an independent observer's forecast from the plan and past record, and compare the two.
Sources
- Heuristics and Biases (Gilovich et al., 2002)
Also known as
- Planning-fallacy debiasing by recall-relevance and observer forecasts
- Recall-relevance procedure
- Outside-view time prediction
- Observer forecasts
Categories
- Research Operations
- Research Design & Rigor