Outside view

What it is

A friend is sure her new restaurant will profit in its first year. Before you judge her plan, ask how most new restaurants do. That question is the outside view.

You forecast from a reference class: past attempts whose outcome depended on the same things as yours. Your estimate starts from the range of its results. What you know about your own plan comes later, as changes you must argue for.

Where it comes from

The idea comes from the psychologists Daniel Kahneman and Amos Tversky. In a 1977 paper they argued that forecasters often ignore the record of similar past cases. They called this record distributional information. Ignoring it may be the biggest cause of forecasting error, they said. So analysts should set up each forecast to use that record.

The names "inside view" and "outside view" came in a 1993 paper by Kahneman and the strategy researcher Dan Lovallo. With the inside view, you treat each case as one of a kind and work from its details. With the outside view, you work from statistics on cases alike in what matters. In July 2003 the two explained the idea to managers in Harvard Business Review. They wrote about the planning fallacy, the habit of underestimating time and cost.

Bent Flyvbjerg, a Danish expert on large projects, turned the idea into a procedure, reference class forecasting. He built it with the consultancy COWI for the UK Department for Transport. It became official guidance in August 2004.

What it corrects

You know your own project better than anyone, so you forecast from the plan: its steps, its team, the obstacles you can see. These forecasts go wrong the same way, again and again. Cost and time come out too low and benefits too high, because a plan assumes nothing unexpected will happen.

Working harder makes this worse. More effort adds detail to the same plan. The missing risks are not in it. They are what happened to others: similar teams that gave up, technology that came late, demand that never appeared. Kahneman and Tversky also found that even people who fully understand these errors keep making them.

Organisations add a second problem. When projects compete for money, their backers have a reason to be hopeful. They can easily argue with a scenario. A record of how similar projects turned out, the base rate, is harder to argue with.

How it works

1. Name what you are forecasting and when you will know the answer.
2. Choose the reference class: past attempts like yours in what matters, wide enough to have data and narrow enough to compare.
3. Collect the results of that class, including abandoned attempts.
4. Take the usual result as your first estimate and the spread as your range.
5. Adjust only where evidence shows your case differs from the class in something its data measures. Write down the size of the change.
6. Write down what would send you back to the base rate.

Worked example

Edinburgh was planning tram Line 2. In October 2004 the engineering firm Ove Arup and Partners Scotland reviewed the line's business case for the Scottish Parliament's Edinburgh Tram Bill Committee. Transport Initiatives Edinburgh, the project company, had written it. This inside view was careful work. The base cost, £255 million, came from a cost database. It was compared with other UK light rail schemes and checked against earlier estimates. On top came £64 million, about 25 per cent, for unexpected costs and optimism bias. The total was about £320 million.

Arup then applied the Department for Transport's uplifts: extra percentages based on overspending on finished rail projects. For an 80 per cent chance of staying within budget, the uplift was 57 per cent: £255 million times 1.57 gives £400 million. For a 50 per cent chance it was 40 per cent, or £357 million. Arup said even these were probably too low. The uplifts are meant for the decision to build. Line 2 was not even at the outline business case stage. The earlier the stage, the more is unknown, and the larger the uplift should be.

For one project on one day, the careful estimate and the base rate differed by £80 million at the 80 per cent level. Line 2 was never built. Trams did run on a shortened Line 1, from the airport to York Place. It opened in May 2014 and reportedly cost £776 million, £231 million over the budget for the full line to Newhaven. The 2023 public inquiry put the real cost near £836 million. The rail uplifts were based on overspending of this size.

In a Business Case Weekly case

In the SoftBank case, take the 2022 fork. As one of the Vision Fund's limited partners, its outside investors, you are asked to back a newer, bigger idea. The inside view is that idea and the fund's account of what changed after WeWork.

From the outside, first ask which group this bet belongs to. Late-stage tech funds of the same years? Funds of this size? Funds where one person decides? Then ask what those funds earned, counting those quietly closed. Last, ask which fact about this fund justifies a different number. You get no answer this way, only the figure your memo must argue against.

In your answer

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

Common misuse

The fake version is a reference class chosen after the answer. Say you pick cases by industry label, or because they flatter the plan. Then any number you wanted comes back, looking like a base rate. To avoid this, name the class and its range of results before reading the case's own forecast. And be able to say which members of the class failed. If none failed, you probably picked only the success stories.

The opposite mistake is to treat the base rate as a verdict. It is only where the estimate starts. Strong evidence about this case can move it far. But you must argue for the move and say how big it is.

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.
• HM Treasury, Green Book supplementary guidance: optimism bias. The Treasury's general uplift tables for UK public projects, for use when there is no better data. The Department for Transport built its rail uplifts, used at Edinburgh, in response to the Green Book.
• Daniel Kahneman and Amos Tversky, "Intuitive prediction: Biases and corrective procedures," 1977; reprinted in Judgment Under Uncertainty: Heuristics and Biases, Cambridge University Press, 1982.