The Inside View and the Outside View

The idea in plain words

Ask a team how long their project will take, and they will think about the project: its tasks, its people, its plan. They will imagine it unfolding, perhaps allowing for a few problems. That is the inside view. Now ask how long similar projects have taken, at this company and elsewhere. That is the outside view, and it almost always gives a longer, more expensive and more accurate answer.

The systematic tendency to underestimate time and cost, even when we know that past projects ran over, is called the planning fallacy.

Where it comes from

Daniel Kahneman told the story of how he found the idea. In the 1970s he was part of a team in Israel writing a curriculum on judgement and decision-making. Asked to estimate how long the project would take, team members guessed around two years. Kahneman then asked the curriculum expert on the team how long similar projects had taken. After some thought, the expert said that about forty per cent never finished, and none he knew of had finished in less than seven years. He also admitted that this team was somewhat below average. The team continued anyway. The project took eight years, and by the time it was done the ministry had lost interest.

In 1979 Kahneman and Amos Tversky named the planning fallacy and proposed a corrective: start from the distribution of outcomes in a reference class of similar cases, then adjust for what is genuinely special about this case. In 1993 Kahneman and Dan Lovallo applied it to management, showing how executives combine “bold forecasts”, which are inside-view optimism about their own projects, with “timid choices”, meaning excessive caution when they consider risks one at a time.

Roger Buehler, Dale Griffin and Michael Ross showed in 1994 that students predicting when they would finish their theses were badly optimistic, even when asked to predict a “worst case”. Bent Flyvbjerg then gathered data on thousands of large infrastructure and IT projects and found that cost overruns were the norm, and that for IT projects the distribution had a long, heavy tail: a minority of projects overran catastrophically. His method of reference class forecasting has since been adopted by several governments.

What it does to the mind

The inside view is a story, and stories are built from the case at hand. When we imagine our project, we imagine it going roughly to plan, because the plan is what we have. Things that could go wrong are many, separate and individually unlikely, so each is discounted, but together they make it almost certain that something will.

The inside view also feels more respectful of the work. Telling a team that “projects like this usually take twice as long” sounds like ignoring what makes them good. People resist base rates because they want to believe that this time, with these people, things will be different. Sometimes they are right, but less often than they believe.

Where it shows up

Large technology bets are especially exposed. Iridium took over a decade from conception to launch, during which the world it was planned for changed. Itanium shipped years late into a market that had moved on. The Google Wave team’s ambition was to rethink communication, and it is worth asking what the base rate is for products that require users to change how they communicate. The opposite lesson appears in the small first pattern: technologies that start small are less exposed to catastrophic overruns.

What it asks of someone deciding

Before approving a plan, find the reference class. How long did comparable launches, migrations or integrations take, inside and outside the company?

Start from the base rate, then adjust. Make the team argue for why this case should differ from the median, instead of arguing from their plan outward.

Build in stages. Projects broken into small, independently valuable pieces have thinner tails. The worst overruns come from large, indivisible commitments that deliver nothing until they deliver everything.