Chapter 23: The Outside View

Type: Chapter 23

Category: Psychology

Difficulty: Intermediate

Reading time: 6 min

Core idea

In 1976, Kahneman was part of a team designing a curriculum for a new high school course. The team had been working for a year and asked each member to estimate independently how long completion would take. Estimates ranged from 18 to 30 months. Kahneman then asked the team’s curriculum expert — who had led many similar projects — what fraction of similar projects had been completed, and how long they took. The answer: none had been completed on time; most had taken 7–10 years; some had been abandoned.

The team discussed this reference class information and then continued predicting completion in 18–30 months. The actual completion time was eight years.

The team was engaged in the planning fallacy: generating predictions based on the specific features and optimistic scenarios of their current project while ignoring the historical distribution of how similar projects actually turn out — the outside view. The inside view (what could this project achieve?) produces optimistic predictions; the outside view (what do similar projects actually achieve?) produces calibrated ones.

Why it matters

Inside view vs. outside view

The inside view is System 1’s natural mode: build a detailed scenario of the specific project — its unique strengths, the team’s competence, the optimal path to completion. The scenario is vivid and coherent. It generates an estimate based on the path you can imagine, not the distribution of outcomes from similar paths.

The outside view requires System 2 deliberation: identify the reference class of similar past projects; determine the distribution of outcomes in that class; anchor on that distribution; then adjust for specific features of the current project. This is effortful and resisted, because the specific project always seems special.

The inside view is almost always more optimistic than the outside view. This is partly motivated — people want the project to succeed and generate scenarios accordingly. But it is also cognitive: scenario-based thinking does not naturally incorporate base rates.

Reference class forecasting

Bent Flyvbjerg’s research on large public projects documented the planning fallacy on a massive scale. Major infrastructure projects — railways, roads, bridges, tunnels, stadiums — exceeded their initial cost estimates by an average of 45% globally. The projects were not poorly planned; they were planned using inside views that ignored the reference class. The standard tool — which Flyvbjerg formalized and which Kahneman endorses — is reference class forecasting:

1. Identify the reference class of past projects similar to the current one.
2. Get the distribution of cost, time, and success rates from that class.
3. Use that distribution as your prior.
4. Adjust for specific features of the current project by the actual predictive validity of those features.

The optimism cascade

The planning fallacy is amplified by competition between projects for funding. If a project’s proponents honestly represent the historical reference class failure rate and cost overrun distribution, the project will not be funded. Competing projects whose proponents present optimistic inside-view estimates will win. The result: a systematic tendency toward optimistic project selection, which produces a portfolio of projects that collectively exceed budget, schedule, and fail at higher rates than stated.

Author’s argument: Optimism is a competitive advantage in persuasion but a systematic source of error in prediction. The solution is not to eliminate optimism from planning but to insist that outside-view base rates accompany inside-view scenarios.

Key takeaways

Key takeaways

• Planning fallacy: intuitive project plans are generated from optimistic inside-view scenarios that ignore the reference class distribution of similar past projects.
• Inside view: focuses on the specific project's features, team, and optimal path. Outside view: anchors on the historical distribution of outcomes from similar projects.
• Kahneman's curriculum example: the team predicted 18-30 months despite knowing similar projects took 7-10 years and some were abandoned. Actual completion: 8 years.
• Reference class forecasting: the prescriptive correction — identify the reference class, obtain its distribution, anchor there, then adjust by the validated predictive validity of specific features.
• Optimism cascade: projects compete for funding on inside-view optimistic estimates; the system selects for the most optimistic rather than the most accurate forecasts.
• The adjustment from outside to inside view is justified only by evidence that the specific features are genuinely diagnostic of better-than-average outcomes.

Mental model

click to zoom

Inside view vs. outside view in project planning — and the reference class forecasting correction.

Read it as: The inside view generates vivid scenarios of what could go right for this specific project and produces systematically optimistic estimates. The outside view forces the question: what actually happened to similar projects? Starting from that distribution and adjusting only for validated specific features produces a less optimistic and more accurate forecast. The adjustment from outside to inside view must be earned with evidence — it is not a free upgrade.

Practical application

Pre-mortem + reference class

Combine the reference class approach with the pre-mortem (from Chapter 19): first, establish what the reference class says will happen. Then ask the team to imagine the project has failed and write why. The reference class anchors the probability; the pre-mortem generates the specific failure modes. Together, they produce a better forecast than either approach alone.

Implementation for project planning:

• Require reference class estimates before internal estimates: before any project sponsor presents their cost and time estimate, require them to present the reference class distribution first.
• Track base rates internally: build organizational memory of what similar past projects actually cost and took. This is the raw material of the outside view for the next project.
• Treat “this project is different” claims as hypothesis, not conclusion: teams always have reasons why their project is exceptional. Test those reasons against the reference class — what specific features of the project actually predict better-than-average outcomes?

Example

A product team estimates six months to build a new feature. The project lead has read reference class data: similar-scope features at this company have averaged 11 months, with 70% running over the initial estimate. She presents the outside view estimate: “The reference class says 11 months and 30% chance of delivering on time.” The team’s inside view — assuming the specific team strength, the clear requirements, the modern tooling — predicts six months.

She suggests anchoring at nine months (adjusting modestly from the reference class mean toward the inside estimate, adjusting by the partial correlation between those specific features and delivery speed). The product roadmap is revised accordingly. The feature delivers in ten months — close to the outside view anchor, not the inside view optimism.

Related lessons

Related

Chapter 18: Taming Intuitive Predictions The statistical correction for intuitive predictions — anchoring on base rates and adjusting by correlation, the methodology underlying reference class forecasting. Chapter 24: The Engine of Capitalism Optimism bias as a feature of entrepreneurial motivation — the positive side of the same bias that produces the planning fallacy. Cognitive Bias The broader taxonomy — the planning fallacy as one of the most consequential biases in organizational decision-making.