neur.ro

Planning fallacy

principle · origin: study · evidence: supported

In short

The planning fallacy is the tendency to believe that your own project will go according to plan, even when you know that most similar projects ran late. The term was proposed by Daniel Kahneman and Amos Tversky (1979). The big examples are well known: the Sydney Opera House was supposed to be finished in 1963 for 7 million dollars, and it opened in 1973, in a scaled-down version, for 102 million. But the effect also shows up on a small scale: in assignments, renovations, a report to finish.

What it says

The core study. Buehler, Griffin and Ross (1994) asked 37 final-year students who were close to finishing their honors thesis when they thought they would hand it in. They asked for three estimates: the most realistic one, one for the case where everything goes perfectly, and one for the case where everything goes badly. They then recorded when each thesis was actually submitted.

Students estimated they would hand in their thesis in 27.4 days if everything went perfectly, in 33.9 days in their most realistic estimate, and in 48.6 days if everything went badly. In reality it took them 55.5 days. What they estimated and what it took days until the honors thesis was submitted if everything goes perfectly 27.4 most realistic estimate 33.9 if everything goes badly 48.6 what it actually took 55.5 Finished on time: 29.7% Even against the pessimistic case: only 48.7%
Data from Buehler, Griffin and Ross (1994), Study 1. The means are based on the 33 students who submitted their thesis.

On average, the students estimated 33.9 days, and the thesis took them 55.5. Only 29.7% finished by their estimated date. What’s more, even the pessimistic case wasn’t pessimistic enough: only 48.7% finished by the date they had given for the case where “everything went as poorly as it possibly could”. The estimates were not worthless, though: those who estimated longer really did take longer. The error was a shift of everyone toward optimism.

The same pattern appeared in the article’s other studies, with school assignments and household tasks. For example, students who had a deadline for an assignment estimated on average that they would finish in 5.9 days, and reported finishing in 11.0. Participants said they were 70–84% sure of their estimates.

Why it happens. When the authors asked participants to think aloud while estimating, they found that almost all their thoughts were about the future plan: how they would work, what they would do first. Only 7% of thoughts were about past experiences, and very few were about possible obstacles. When they thought about past delays, participants explained them by causes that were transient and specific to that situation.

Kahneman and Lovallo (1993) call this the inside view: you treat the project as a unique case, build scenarios about how it will go, and ignore the statistics of similar projects. The outside view, as Flyvbjerg (2006) describes it following Kahneman, works the other way round: how long did projects like this usually take?

What worked. In the fourth study of Buehler et al. (1994), simply remembering how long previous assignments had taken changed nothing. Estimates became realistic only when participants were asked to explicitly connect their past experiences to the current assignment: to imagine how their usual pattern might repeat itself. The share who finished on time rose from 29% to 60%.

In large projects. Flyvbjerg (2006) describes applying the outside view to project management, under the name reference class forecasting, in three steps:

  1. choose a class of similar past projects, broad enough to be statistically meaningful and narrow enough to be comparable;
  2. establish how outcomes were distributed in that class, for example cost or time overruns;
  3. place your project within that distribution, instead of estimating it from the plan.

The method was used in the UK for the costs of transport projects. Flyvbjerg describes the application, but the article is not an experiment that measures its effect.

Example

Flyvbjerg (2006) relates a story told by Kahneman. A team writing a new school curriculum estimated, each member separately, between 18 and 30 months to finish. Then one of them, an expert in such curricula, was asked how long similar teams had taken. The answer: about 40% had given up, and the rest had finished in 7 to 10 years. The team carried on anyway. It finished after eight years, and the curriculum was rarely used.

How to apply it

The steps below are a practical approach we propose, based on the studies cited.

  1. Start from similar cases, not from the plan. First ask how long your last projects of this kind took you, or how long they took others. Only then adjust for what is different now.
  2. Explicitly connect the past to the current project. In Buehler et al. (1994), remembering alone didn’t help. What helped was picturing concretely how the usual delays could repeat in the current project.
  3. Don’t rely on the pessimistic case. Even the “if everything goes badly” estimate wasn’t long enough for half of the students.
  4. Keep a record. Write down your estimates and the actual durations. Roy et al. (2005) show that people also remember past durations as shorter than they were, so memory alone is not a reliable base.
  5. Let someone else estimate. In Buehler et al. (1994), observers estimating for someone else gave longer timelines. Careful, though: they erred in the opposite direction, so they weren’t more accurate either.

Limits and nuances

Sources

See also: Circle of competence, Implementation intentions