Tool/Mental Model/No. 0704
Outside View vs. Inside View
The outside view forecasts from outcomes of similar cases; the inside view forecasts from a case’s specific details and plans. The contrast, articulated by Daniel Kahneman and Dan Lovallo, links to reference-class forecasting: comparable results set a baseline, adjusted for evidence about the case.
- Evidence
- Useful, modest evidence
- Read
- 6 min
- Links
- 13 connections
- Useful when
- Deciding under uncertainty · Estimating time and cost · Forecasting · Running projects
01You've seen this when…
- in life
You map out a three-month career change. People who made the same switch describe a year of classes, applications, and false starts.
- at work
A sales manager forecasts next quarter by adding up the deals each rep expects to close. Last year’s deals at those same stages produced far less revenue.
- out in the world
A city presents a new tram schedule built from detailed engineering tasks. Its last two tram extensions opened years after their announced dates.
02The idea
A plan can look convincing from close up. Capable people have laid out tasks that fit together and backed each estimate with a reason. Yet projects with equally convincing plans often finish late.
The inside view builds a forecast from the particulars: your team, resources, intentions, and sequence of steps. The outside view starts with outcomes: what happened when other people attempted comparable things?
A well-grounded forecast needs both views. The inside view explains how you intend to succeed. It helps you design the work and identify advantages. The outside view checks whether your forecast fits the track record of similar attempts, including the problems their planners didn’t foresee.
The useful order is to let comparable outcomes establish a starting forecast, then adjust it using evidence about your case. A shorter schedule needs more than confidence in the team. It needs a difference that plausibly changes the result.
The outside view adds a focus on groups of relevant cases and their spread of outcomes to reasoning by analogy. One memorable success can be a misleading comparison. Reference-class forecasting turns that principle into a structured forecasting method.
03How to use it
-
Define exactly what you’re predicting. Specify the outcome and the clock. Time from approving a project to opening it is different from time spent on construction. Include what counts as completion, failure, or abandonment.
-
Choose comparisons before inspecting their results. Look for cases with similar scale, starting conditions, constraints, and completion criteria. Choose features that affect the outcome to separate useful comparisons from superficial similarities. If several groups look reasonable, check whether your conclusion holds across them.
-
Collect outcomes from the full range of attempts. Use records where possible: actual costs, elapsed time, conversion rates, or completion rates. Include stalled and failed attempts. If the data omit them, say so; selection bias can make the starting forecast too favorable.
-
Establish a baseline and a range. For a typical completion time, a median can be useful. Also examine slower cases and the chance of not finishing. For an expected financial cost, the mean may matter more because unusually expensive cases still consume money. Don’t turn a small sample into a precise probability.
-
Adjust only for differences with support. A repeatable design might save time. A team’s claim of unusual motivation provides weak evidence for an adjustment. Check whether comparable cases already had your supposed advantage. Where possible, compare cases with and without the relevant feature. This is the spirit of Bayesian updating: start with prior evidence, then revise it using new information.
-
Separate the forecast from the commitment. Your preferred deadline sets a target. The forecast estimates when you can expect to finish. Use the forecast to choose staffing, reserves, milestones, and what you can safely promise. If the inside and outside views still disagree sharply, investigate the disagreement. Splitting the difference leaves its causes unexplained.
04A worked example
Consider this illustrative case. A museum plans to open an interactive exhibit four months after design approval. Its schedule includes fabrication, installation, software testing, and staff training. Each department signs off on its portion.
What it looks like A grounded forecast. The experienced team has explicit tasks and a supplier booked. Four months seems defensible because no individual step looks unreasonable.
What’s actually going on Records for 16 comparably sized exhibit builds show a median completion time of eight months, with most taking six to eleven. The four-month plan leaves little room for problems between departments. Hardware may arrive before the software is ready, and changes may be required by accessibility testing or a supplier who needs revised drawings. Each problem is a possibility, and comparable builds regularly encounter some of them.
What made it work The museum uses the eight-month baseline to revise its four-month plan. Its modular fabrication system could make this build faster. Earlier projects using that same system support a one-month reduction. With seven months as its working estimate, the team retains a wider planning range and keeps a temporary exhibit available. The inside view identifies a supported adjustment and shapes the backup plan around the outside view’s revised forecast.
05When to reach for it
06When it misleads
-
The comparison group is wrong. Small website updates won’t tell you much about replacing an entire platform. An outside forecast can be confidently wrong if the chosen cases miss the constraints that drive results.
-
The records contain only survivors. Successful businesses, published research, and finished projects are easier to find than abandoned attempts. Their outcomes describe the survivors’ experience, so the missing attempts limit what you can infer about everyone’s chances. Investigate what is missing before relying on the apparent track record.
-
The underlying conditions have changed. New regulations, technology, or supply constraints can make historical outcomes less relevant. This is distribution shift. Look for evidence that the change affects your forecast, and use it to test assumptions that history will repeat or that it has become irrelevant.
-
The typical outcome becomes a promise. A median deadline leaves roughly half of comparable cases taking longer. A high-confidence delivery date needs more room for slower cases. Likewise, typical cost and expected cost can differ substantially when overruns are large.
-
Adjustments quietly restore the original story. Starting with eight months and subtracting four for enthusiasm defeats the exercise. Record each adjustment and its basis. If you lack reliable comparisons, admit that limitation and use other checks, such as a pre-mortem. Inventing a baseline would create false reassurance.
07Roots
Daniel Kahneman was helping a team in Israel develop a high-school curriculum on judgment and decision making. As he later recalled, the team had made progress and expected to finish without a long delay. Then he asked curriculum specialist Seymour Fox to consider comparable efforts. Fox remembered a much worse track record: long development times and projects that never finished.
The striking detail was that Fox had already supplied an optimistic estimate for their own project. His estimate left out the history he already knew when he considered the case in front of him. Kahneman’s account offers an anecdotal illustration of the problem: detailed involvement can crowd out knowledge you already possess.
Kahneman and Amos Tversky developed comparison-based corrective procedures in their earlier work on intuitive prediction. In 1993, Kahneman and Dan Lovallo gave the inside–outside contrast a prominent role in explaining bold project forecasts alongside cautious attitudes toward explicit risks. The approach later traveled into practical project appraisal. Bent Flyvbjerg helped develop reference-class forecasting for infrastructure, making records of actual projects a check on the persuasive plans for the next one.
08How solid is this?
The forecasting errors it targets are well documented, and comparison-based forecasting has substantial applied support. Its value depends on relevant, reasonably complete data and disciplined adjustments; an outside view is not automatically the better forecast.
09Connections
- Helps counter Anchoring Bias, Focalism, Impact Bias, Planning Fallacy, Optimism Bias, Selection Bias, Distribution Shift, Overconfidence Effect
- Part ofAnalogical Reasoning
- IncludesReference-Class Forecasting
- See also Decision Journal, Bayesian Updating, Pre-Mortem
+ 3 more in the list
10Origin and sources
Developed through Daniel Kahneman and Amos Tversky’s work on intuitive prediction, including their 1979 account of corrective procedures. Daniel Kahneman and Dan Lovallo articulated the inside–outside contrast in 1993.
- [1]Kahneman, D., & Tversky, A. (1979). Intuitive prediction: Biases and corrective procedures. TIMS Studies in Management Science, 12, 313–327.
- [2]Kahneman, D., & Lovallo, D. (1993). Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking. Management Science, 39(1), 17–31.
- [3]Kahneman, D. (2011). Thinking, Fast and Slow, chapter 23. Farrar, Straus and Giroux.
- [4]Flyvbjerg, B. (2006). From Nobel Prize to Project Management: Getting Risks Right. Project Management Journal, 37(3), 5–15.
Suggest an edit· Updated 2026-10-02