Outside view

What it is

The outside view forecasts a decision from what happened to comparable attempts, rather than from a story about this one. You place the case in a reference class, a set of past attempts that resemble it in the ways that drive the outcome, read the spread of results it produced, and start your estimate there. What you know about the present plan enters afterwards, as an adjustment you have to justify.

Where it comes from

The theory is Daniel Kahneman and Amos Tversky's. Their paper "Intuitive prediction: Biases and corrective procedures" circulated as a 1977 technical report and was reprinted in Judgment Under Uncertainty in 1982. Ignoring distributional information, meaning the record of similar past cases, is perhaps the largest single source of forecasting error, they argued, and analysts should frame the problem so that such information gets used.

The named contrast between inside view and outside view belongs to a later paper: Daniel Kahneman and Dan Lovallo, "Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking," Management Science 39(1), 1993. The inside view treats each forecast as unique and works from the specifics. The outside view works from statistics on cases judged similar in key respects. Lovallo and Kahneman restated the pair for managers in Harvard Business Review in July 2003, under the planning fallacy.

The working procedure is younger. Reference class forecasting was built by Bent Flyvbjerg with the consultancy COWI for the UK Department for Transport, and published as official guidance in August 2004.

What it corrects

You know more about this project than anyone else, and a competent person uses that knowledge: they forecast from the plan, its steps, its team, the obstacles they can see. What comes out is systematically low on cost and time and high on benefits, because a plan describes the path where nothing unlisted happens.

Ordinary care makes this worse rather than better. More diligence adds detail to the same inside story, and the missing risks are the ones no internal detail contains: the comparable teams that gave up, the technology that arrived late, the demand that never appeared. Kahneman and Tversky also found that knowing about the illusion does not dissolve it. Organisations add a second layer: a project competing for funds gives its sponsors a reason to be optimistic, and a documented base rate is harder to talk past than a scenario.

How it works

1. Name the quantity you are forecasting and the date it will be settled.
2. Choose the reference class: past attempts sharing the mechanism that drives the outcome, broad enough for data, narrow enough to compare.
3. Collect the outcomes that class produced, including the attempts abandoned.
4. Take the class's typical value as your opening estimate, its spread as your range.
5. Adjust only where you hold evidence that this case differs on something the distribution measures, and write down the size of the adjustment.
6. Record what would send you back to the base rate.

Worked example

Edinburgh Tram Line 2, October 2004. Ove Arup and Partners Scotland, appointed by the Scottish Parliament's Edinburgh Tram Bill Committee, reviewed the business case prepared by Transport Initiatives Edinburgh. The inside view there was careful work: a base cost of £255 million built from a cost database, compared with other UK light rail schemes and reconciled with earlier estimates, plus £64 million, some 25 per cent, for contingency and optimism bias. Total, about £320 million.

Arup then applied the Department for Transport's uplifts, drawn from recorded overruns on completed rail projects. For an 80 per cent chance of staying within budget the uplift was 57 per cent: £255 million multiplied by 1.57 gives £400 million. For a 50 per cent chance, 40 per cent, or £357 million. Arup noted that even these figures were probably low: the uplifts are meant to be applied at the decision to build, and Line 2 had not reached its outline business case.

On the same project, on the same day, the careful internal estimate and the base rate stood £80 million apart at the 80 per cent level. Line 2 was never built. The shortened line that opened in May 2014 was reported at £776 million, and the 2023 public inquiry put the real figure near £836 million.

In a Business Case Weekly case

In the SoftBank case, take the fork where a Vision Fund limited partner in 2022 is asked to back a larger, newer thesis. The inside view is the thesis and the fund's account of what changed after WeWork. The outside view asks which class this commitment belongs to: late-stage technology funds of that vintage, funds of that size, funds where one person decides. It then asks what those funds returned, counting the ones quietly wound down, and which fact about this fund justifies a number away from that. The method supplies the figure the memo has to argue against, not the answer.

In your answer

• "The reference class here is …, and across it the usual outcome is …"
• "Before reading this company's own projection, the base rate for … is …"
• "This case differs from the class in one measurable way, …, which moves my estimate from … to …"
• "If that difference does not hold, I return to the class figure of …"

Common misuse

The version that looks like the method is a reference class chosen after the answer. Pick comparables by industry label, or by which ones flatter the plan, and any number you wanted returns wearing the authority of a base rate. The test is order, plus honesty about failure: name the class and its outcome range before you read the case's own forecast, and be able to say which members of the class failed. A class with no failures in it is a marketing list.

The opposite misuse treats the base rate as a verdict. It is where the estimate starts; decisive local evidence can move it far, but the move has to be argued and sized.

References

• Daniel Kahneman and Dan Lovallo, "Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking," Management Science 39(1), 1993, 17–31. PDF
• Bent Flyvbjerg, "From Nobel Prize to Project Management: Getting Risks Right", Project Management Journal 37(3), 2006, 5–15. The Edinburgh figures and the whole procedure; one evening's reading.
• HM Treasury, Green Book supplementary guidance: optimism bias. The uplift tables a UK public project has to use.
• Daniel Kahneman and Amos Tversky, "Intuitive prediction: Biases and corrective procedures," 1977; reprinted in Judgment Under Uncertainty: Heuristics and Biases, Cambridge University Press, 1982.