Within Decision Routines
What happened when others tried this?
Looking at how similar projects actually performed can stop a persuasive inside story from becoming the only evidence in the room.
On this page
- Inside view versus outside view
- Finding a useful comparison class
- Using base rates without ignoring context
Page outline Jump by section
Introduction
When a forecast matters, the most persuasive story is often the least reliable evidence. Teams naturally focus on the distinctive features of their own project, investment or strategy—the inside view. Reference classes provide an outside view: instead of asking only whether this plan looks convincing, they ask what actually happened to a sufficiently similar set of completed cases. The resulting base rates do not eliminate judgement, but they make it much harder to ignore recurring patterns of delay, cost escalation, failure or unexpected success. Research by Daniel Kahneman, Amos Tversky and later Bent Flyvbjerg has shown that this approach can substantially reduce optimism bias, particularly in major projects where conventional forecasts repeatedly prove too favourable. [Project Management Institute+2IDEAS/RePEc]pmi.orgProject Management InstituteFrom Nobel Prize to project managementReference class forecasting promises more accuracy in forecasts by taki…
Inside view versus outside view
The inside view builds a forecast from the details of the current case: the team’s expertise, the planned milestones, identified risks and expected advantages. This approach is valuable because it incorporates information that outsiders may not know. Its weakness is that people tend to overweight evidence supporting the plan while underestimating unforeseen setbacks.
The outside view starts elsewhere. It asks:
- What similar projects have already been completed?
- How long did they actually take?
- How often did they exceed their budgets?
- How variable were the outcomes?
Rather than treating the present project as unique, the outside view begins with historical distributions and only then considers whether there is convincing evidence that the current case genuinely differs.
Kahneman and Tversky proposed this logic as a way to counter the planning fallacy—the tendency to underestimate costs, completion times and risks even when similar projects have repeatedly overrun. Flyvbjerg later applied the idea systematically to infrastructure forecasting, where decades of evidence showed recurring optimism in cost and schedule estimates. [Project Management Institute+2IDEAS/RePEc]pmi.orgProject Management InstituteFrom Nobel Prize to project managementReference class forecasting promises more accuracy in forecasts by taki…
What makes a useful reference class?
Choosing a comparison group is the critical judgement. A reference class that is too broad ignores important differences; one that is too narrow becomes statistically weak because there are too few comparable cases.
Useful comparison classes usually share several important characteristics:
- similar objectives rather than superficial appearance
- comparable technical complexity
- similar regulatory or organisational environments
- similar project scale
- comparable delivery methods and incentives
For example, forecasting the construction time for a new urban rail line should rely primarily on completed urban rail projects with similar procurement arrangements, not on all transport infrastructure or all construction work.
The challenge is that every project belongs to many possible categories. Statisticians call this the reference class problem: different comparison groups may produce different base rates, so selecting the most informative class requires careful reasoning rather than mechanical matching. [Wikipedia]WikipediaReference class problemReference class problem
What happened when others tried this?
The value of reference classes comes from repeated empirical patterns rather than isolated anecdotes.
Large transport infrastructure projects provided some of the earliest practical tests. Flyvbjerg’s work found systematic underestimation of costs and schedules across many completed projects, leading to the development of reference class forecasting methods that adjust forecasts using observed historical distributions instead of relying solely on project-specific estimates. These methods have since been incorporated into project governance guidance in several jurisdictions. [IDEAS/RePEc+2arXiv]ideas.repec.orgv16y2006i1p3 21First, the paper documents that inaccurate…Read more…
One notable application occurred in Hong Kong, where researchers developed reference classes for major road projects by comparing local outcomes with hundreds of international projects. Rather than producing a single “best guess”, the method estimated probabilities of different cost and schedule outcomes, allowing decision-makers to choose contingency levels that matched their tolerance for risk. [arXiv]arxiv.orgBent Flyvbjerg, Chi-keung Hon, and Wing Huen Fok, 2016by B Flyvbjerg · 2017 · Cited by 77 — Reference Class Forecasting (RCF) is a m…
The broader lesson extends beyond construction. Whether evaluating product launches, technology deployments or organisational change programmes, historical outcome distributions often reveal recurring patterns that individual planning teams overlook.
Using base rates without ignoring context
Reference classes are not intended to replace expert judgement. They provide a disciplined starting point.
A practical sequence is:
- Produce the normal project estimate.
- Identify an appropriate comparison class.
- Examine the actual distribution of past outcomes.
- Position the current case within that distribution.
- Adjust only when there is strong evidence that the current case differs materially from previous ones.
This approach preserves useful local knowledge while making optimistic assumptions earn their place.
Suppose similar software migrations historically exceeded schedules by around 30% on average. A team forecasting completion exactly on schedule should explain specifically why their project escapes the historical pattern. General confidence, enthusiasm or claims of exceptional competence are weak evidence; demonstrable structural differences, such as radically simplified scope or proven automation, are stronger.
Common mistakes when using reference classes
Reference class forecasting(#endnote-2 “Endnote 2”) is powerful, but it is not automatic. [Wikipedia]WikipediaReference class problemReference class problem
Frequent errors include:
- Selecting only successful examples. Excluding failures produces misleadingly optimistic base rates.
- Assuming uniqueness too quickly. Most projects contain novel elements, yet still share important statistical similarities with earlier work.
- Using outdated comparisons. Changes in technology, regulation or markets may require updating the comparison group.
- Ignoring the spread of outcomes. Average performance alone hides the probability of unusually good or bad results.
- Treating the outside view as deterministic. Base rates describe probabilities, not certainties.
Recent reviews of the method emphasise that defining similarity remains one of its hardest practical problems. Researchers increasingly argue for more transparent and evidence-based selection of comparison classes rather than relying on intuition alone. [Taylor & Francis Online]tandfonline.comTaylor & Francis OnlineReference class forecasting: promises, problems, and a…by CC Cantarelli · 2025 · Cited by 7 — Kahneman and Tver…
When should the outside view override the inside view?
The outside view deserves greater weight when:
- historical outcomes are abundant and consistent;
- forecasts repeatedly display optimism in the same direction;
- projects involve substantial uncertainty;
- organisational incentives encourage favourable forecasts.
The inside view deserves relatively more weight when there is credible evidence that the current case genuinely differs from the historical population—for example through fundamentally different technology, regulation or delivery methods—and those differences have measurable effects rather than merely persuasive narratives.
The strongest decisions combine both perspectives. The inside view explains why this case may differ; the outside view asks whether similar claims have proved true before. When they disagree sharply, decision-makers should investigate the disagreement rather than automatically trusting either one.
Why reference classes improve high-stakes decisions
Reference classes introduce disciplined scepticism into forecasting. They challenge the natural tendency to believe that careful planning, talented teams or exceptional circumstances guarantee exceptional outcomes. By anchoring forecasts in observed experience before adjusting for genuine differences, they reduce the influence of optimism bias and encourage decisions that are calibrated to how comparable situations have actually unfolded.
In high-stakes decisions, that shift from persuasive stories to historical base rates often matters more than making the forecast marginally more sophisticated. It changes the central question from “Why will our plan succeed?” to “What usually happens when people attempt something like this, and what evidence justifies expecting a different result?” [Project Management Institute+2IDEAS/RePEc]pmi.orgProject Management InstituteFrom Nobel Prize to project managementReference class forecasting promises more accuracy in forecasts by taki…
Amazon book picks
Further Reading
Books and field guides related to What happened when others tried this?. Use these as the next step if you want deeper reading beyond the article.
Superforecasting
Directly teaches base rates, outside-view thinking, updating beliefs, and disciplined comparison.
How Big Things Get Done
Centres on reference-class forecasting, planning fallacy, cost overruns, and learning from comparable projects.
Thinking, Fast and Slow
Introduces the inside view, outside view, optimism bias, and judgement errors in forecasting.
The Signal and the Noise
Explores why forecasts fail, how evidence should be weighed, and how uncertainty should be represented.
eBay marketplace picks
Marketplace Samples
Topic-anchored marketplace searches for visual, collectible, or second-hand items related to this page.
Endnotes
-
Source: ideas.repec.org
Title: v16y2006i1p3 21
Link: https://ideas.repec.org/a/taf/eurpls/v16y2006i1p3-21.htmlSource snippet
First, the paper documents that inaccurate...Read more...
-
Source: Wikipedia
Title: Reference class problem
Link: https://en.wikipedia.org/wiki/Reference_class_problem -
Source: arxiv.org
Link: https://arxiv.org/pdf/1302.3642Source snippet
From Nobel Prize to Project Management: Getting Risks...by B Flyvbjerg · 2013 · Cited by 826 — Reference class forecasting was originall...
-
Source: arxiv.org
Link: https://arxiv.org/pdf/1710.09419Source snippet
Bent Flyvbjerg, Chi-keung Hon, and Wing Huen Fok, 2016by B Flyvbjerg · 2017 · Cited by 77 — Reference Class Forecasting (RCF) is a m...
-
Source: arxiv.org
Title: arXiv Reference Class Forecasting for Hong Kong’s Major Roadworks Projects
Link: https://arxiv.org/abs/1710.09419 -
Source: arxiv.org
Title: arXiv Uniqueness Bias: Why It Matters, How to Curb It
Link: https://arxiv.org/abs/2408.07710 -
Source: Wikipedia
Title: Reference class forecasting
Link: https://en.wikipedia.org/wiki/Reference_class_forecastingSource snippet
Reference class forecastingReference class forecasting or comparison class forecasting is a method of predicting the future by looking...
-
Source: youtube.com
Title: What is Reference Class Forecasting
Link: https://www.youtube.com/watch?v=Kyh8Juyl_pQSource snippet
Reference Class Forecasting: Mastering Project Budgeting...
-
Source: youtube.com
Title: Reference Class Forecasting: Mastering Project Budgeting
Link: https://www.youtube.com/watch?v=feV4OPj2wvQSource snippet
The Planning Fallacy: Why You Always Underestimate How Long Everything Takes...
-
Source: pmi.org
Link: https://www.pmi.org/learning/library/nobel-project-management-reference-class-forecasting-8068Source snippet
Project Management InstituteFrom Nobel Prize to project managementReference class forecasting promises more accuracy in forecasts by taki...
-
Source: tandfonline.com
Link: https://www.tandfonline.com/doi/full/10.1080/09537287.2025.2578708Source snippet
Taylor & Francis OnlineReference class forecasting: promises, problems, and a...by CC Cantarelli · 2025 · Cited by 7 — Kahneman and Tver...
-
Source: projectmanagement.com
Title: reference class forecasting depends on how you define similar
Link: https://www.projectmanagement.com/blog-post/79682/reference-class-forecasting-depends-on-how-you-define–similar-Source snippet
Reference Class Forecasting Depends on How You Define...6 May 2026 — Reference Class Forecasting (RCF), popularized by Bent Flyvbjerg, i...
Published: May 2026
-
Source: cenex.au
Title: Reference Class Forecasting & Optimism Bias
Link: https://cenex.au/risk/reference-class-forecasting.htmlSource snippet
Reference class forecasting answers optimism bias with the “outside view”: instead of trusting the plan in front of you, you anchor...
-
Source: apm.org.uk
Title: Bent Flyvbjerg, as optimism bias
Link: https://www.apm.org.uk/news/reference-class-forecasting-useful-method-or-random-number-generator-webinar/Source snippet
Reference Class Forecasting - useful method, or random...11 Oct 2023 — Reference Class Forecasting (RCF) as a method to predict uncertai...
Additional References
-
Source: citethisforme.com
Link: https://www.citethisforme.com/Source snippet
Cite This For Me: Harvard, APA, MLA Reference GeneratorAutomatic works cited and bibliography formatting for MLA, APA and Chicago/Turabia...
-
Source: scribbr.com
Link: https://www.scribbr.com/citation/generator/Source snippet
Free Citation Generator | APA, MLA, ChicagoGenerate citations in APA, MLA, Chicago, and Harvard style with Scribbr's free Citation Genera...
-
Source: mybib.com
Link: https://www.mybib.com/Source snippet
A New FREE APA, Harvard, & MLA Citation GeneratorMyBib creates accurate citations automatically for books, journals, websites, an...
-
Source: linkedin.com
Link: https://www.linkedin.com/posts/flyvbjerg_reducing-risks-in-megaprojects-the-potential-activity-7132215991085756416-VeprSource snippet
Prof. Bent Flyvbjerg's PostThe first full literature review of reference class forecasting is now available: "Reducing risks in megaproje...
-
Source: corporate.jasoncollins.blog
Link: https://corporate.jasoncollins.blog/outside-viewSource snippet
jasoncollins.blog23 The outside view – Course notes...The planning fallacy is the tendency of people to underestimate the completion tim...
-
Source: committees.parliament.uk
Link: https://committees.parliament.uk/writtenevidence/100937/html/Source snippet
2008. 'Curbing optimism bias and strategic misrepresentation in planning: Reference class forecasting in practice', European Planning Stu...
-
Source: researchgate.net
Link: https://www.researchgate.net/publication/233258056_Curbing_Optimism_Bias_and_Strategic_Misrepresentation_in_Planning_Reference_Class_Forecasting_in_PracticeSource snippet
lass forecasting in planning practice.Read more...
-
Source: youtube.com
Title: How Big Things Get Done with Prof Bent Flyvbjerg
Link: https://www.youtube.com/watch?v=yUEycqrCuIYSource snippet
What is Reference Class Forecasting | Explained in 2 min - YouTube What is Reference Class Forecasting | Explained in 2 min - YouTube...
-
Source: dspace.lib.cranfield.ac.uk
Link: https://dspace.lib.cranfield.ac.uk/server/api/core/bitstreams/076d9c29-659d-4bfd-9b1b-e0019ea9ad20/contentSource snippet
class forecasting: promises, problems, and a...by CC Cantarelli · Cited by 7 — Reference Class Forecasting (RCF) has emerged as a promin...
-
Source: youtube.com
Title: The Planning Fallacy: Why You Always Underestimate How Long Everything Takes
Link: https://www.youtube.com/watch?v=6bvtMRHlJN8Source snippet
3 Reasons Your Schedule Always Slips...
Topic Tree