Your friends are late because of the planning fallacy. People picture a trip going smoothly and leave out the delays they hit last time. In a 1994 study, thesis students predicted 33.9 days and took 55.5. Fewer than half finished even by their worst-case date. What helps is asking how long the trip took last time, and counting each step of it.
What 37 thesis students got wrong
In 1994, three psychologists phoned 37 final-year psychology students at the University of Waterloo. Each one was writing an honours thesis. The researchers asked when they would hand it in. They also asked for a worst case: when would they finish "if everything went as poorly as it possibly could"?
The psychologists were Roger Buehler, Dale Griffin and Michael Ross. Their paper reports five studies. It is one of the main tests of what Daniel Kahneman and Amos Tversky named the planning fallacy. People expect their own plans to take less time than they do. They keep expecting it after being wrong in the same way many times before.
| Group in the paper | Guess vs. reality | Finished by the guessed date |
|---|---|---|
| Thesis students, best guess | 33.9 days vs. 55.5 | 29.7% |
| Thesis students, worst case | 48.6 days vs. 55.5 | 48.7% |
| Computer assignment, no prompt | 1.3 days too soon, on average | 29.3% |
| Computer assignment, told to link past to plan | 0.1 days too soon, on average | 60.0% |
| Observers guessing for a student | 1.7 days too late, on average | 66% to 71% |
Even the worst case was too hopeful. More than half the students missed a date they had picked by imagining everything going wrong.
Each delay feels like a one-off
In the computer assignment study, students wrote down what they were thinking while they made their guess. Almost all of them (93.5%) described their plans: what they would do and when. Only 8.9% mentioned how long similar assignments had taken before.
The students did remember being late. Asked about earlier plans that had run over, they blamed causes outside themselves. Those causes seemed temporary and specific to that one project. A delay that feels like a one-off gives no reason to plan for it next time. So each new plan starts from zero.
A trip to dinner follows the same pattern. Your friend pictures the route. The drive is twenty minutes, so they leave twenty minutes before. Finding the keys, waiting for the lift, a slow bus and the walk from the car park are not in the picture. All of those happened last week too. Last week, each one had its own excuse.
Travel time in general is a different story. In a Dutch peak-hour driving experiment, Peer, Knockaert, Koster and Verhoef asked commuters how long their morning drive took. On average, they overstated it by about 1.5 times. People know that travel is slow. They forget it when they plan one particular trip.
The person waiting guesses long
In the last study, observers read notes from a student in the assignment study. Some saw the student's plans, some saw how long their past assignments had taken, and some saw both. Each observer then guessed when that student would finish.
The observers guessed 8.5 days on average. The students themselves had guessed 5.5, and the real figure was 6.8. The observers also used the past far more: 48% of them mentioned the student's history, against 2.4% of the students.
The observers were no more accurate, though. They missed by a similar amount in the other direction. At dinner, the friend who says "7:00" and the person at the table who says "7:20" are both guessing. In the study, the real answer fell between the two on average.
Linking last time to this time
One instruction worked. In the assignment study, some students were asked to recall similar past tasks and to say how those experiences applied to the new one. Their guesses stopped running early. Students who only recalled the past, without connecting it to the plan, kept underestimating.
There is a catch. The linked guesses were less biased but not more accurate. They were as likely to be too long as too short. That still helps when the question is whether everyone arrives before the table is given away.
Breaking the task into steps helps too. In three experiments, Forsyth and Burt (2008) found that people allowed more time for a task when they estimated each step and added them up. For a trip, the steps are getting ready, getting to the car, the drive, parking and the walk in.
Before the next dinner, the group can ask the late friend three questions:
- How long did it take last time, door to door?
- What made it slow, and will that happen again?
- How long is each step at both ends, not just the drive?
When the table can see the map
We built OnTime so a group does not have to guess. One person starts a Radar with an end time, such as an hour after the booking. The group joins it, and everyone sees each other's live location until it ends. The people at the table watch the late friend move toward them. If someone adds the restaurant as a stop, the friend's Live Activity shows the distance left.
The map cannot make anyone leave on time. It only helps once they are on the way, and it shows their last known position if their phone loses signal.