Outside-View Forecasting
Forecast from comparable outcomes before trusting your personal plan
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 97%
Outside-View Forecasting counters the planning fallacy by starting with what happened in comparable cases rather than with an elaborate story about why the current case will be different. First define the prediction, then assemble an appropriate reference class and establish its typical result. Record the intuitive, internal forecast separately so its optimism remains visible. Use specific evidence to adjust the reference-class baseline, but do not discard that baseline merely because the current team feels unusually prepared. The output is a forecast anchored in observed frequencies rather than confidence. Repeatedly comparing predictions with actual outcomes creates a feedback loop that improves reference-class selection and exposes recurring sources of underestimation.
Origin
Giovinazzo illustrated the method through his University of Chicago PhD, which he expected to finish in three years but completed in the normal five. Extracted from Masters in Business.
Core principles
- 01Personal confidence is not statistical evidence
- 02Comparable outcomes provide a stronger baseline
- 03Special circumstances should adjust a baseline, not replace it
- 04Forecasts improve when internal and external views are separated
How to run it
- 1
Define the forecast
State precisely what must be predicted, including the outcome, measurement unit, and time horizon.
Pro tip Use a range as well as a single-point estimate.
Watch out A vague prediction cannot be calibrated later.
- 2
Choose a reference class
Find completed cases that resemble the present situation in the factors most likely to affect the result.
Pro tip Prefer actual outcomes over stated plans.
Watch out Selecting only unusually successful comparables recreates the original optimism.
- 3
Establish the baseline
Calculate the typical duration, cost, or success rate among the comparable cases.
Pro tip Inspect the full distribution, including delays and failures.
Watch out The average alone can hide a long tail of overruns.
- 4
Record the internal view
Estimate the outcome from the specific plan and document why it differs from the baseline.
Pro tip Write this estimate before reviewing it with others.
Watch out Do not let confidence substitute for evidence.
- 5
Reconcile the forecasts
Start from the external baseline and adjust only for concrete, measurable differences in the present case.
Pro tip Make each adjustment explicit and independently challengeable.
Watch out Treating every feature as exceptional will nullify the outside view.
- 6
Calibrate with results
Compare the forecast with the eventual outcome and update how you select reference classes or adjustments.
Pro tip Keep a prediction log across similar projects.
Watch out Without reviewing misses, the same bias will persist.
In the wild
Giovinazzo believed prior coursework, a focused research area, and two years of experience would let him complete his PhD in three years. The external evidence showed that nobody finished in three years and almost nobody finished in four. He ultimately completed it in five years, matching the reference-class norm.
→ The external baseline proved more accurate than the personalized forecast.
A team estimates six weeks from its detailed task list, while ten comparable migrations took a median of eleven weeks. It adopts eleven weeks as the baseline, then reduces the estimate to ten because one verified dependency has already been removed.
→ The schedule reflects observed delivery rates without ignoring genuine project-specific advantages.
Common mistakes
Declaring the current case exceptional
People discount comparable outcomes because they know more details about their own situation. Familiarity with those details does not prove that the result will differ.
Using a flattering reference class
Choosing only elite performers or unusually easy projects creates a biased baseline rather than a true outside view.
Ignoring forecast feedback
Failing to compare estimates with actual outcomes prevents calibration and allows systematic optimism to continue.
Is it for you?
Best for
It is best for estimating project durations, costs, completion rates, and other uncertain outcomes with historical comparables.
Not ideal for
It is less useful for genuinely unprecedented situations with no meaningful reference class.
From the transcript
“people use an internal model the forecast. They think about their own personal situation, and they downweight the external model of forecasting, which would be…”
“It turns out nobody does it in three years, and almost nobody does it in four years. And so I was the norm. I did…”
From the episode
Dr. Raife Giovinazzo Discusses Research on Investors' Mistakes
Dr. Raife Giovinazzo