Blog / The Planning Fallacy: Use Comparable History Alongside the Story of the New Task

The Planning Fallacy: Use Comparable History Alongside the Story of the New Task

A plan identifies steps; comparable completion records test how long similar work has actually taken.

October 9, 2026 · 6 min read · By Cepingku Editorial Team

A time estimate often begins with a smooth future story: materials arrive on Monday, a draft is written Tuesday, someone confirms it Wednesday, and delivery follows Friday. That story can reveal steps, but it can also treat each step occurring as hoped as the default. In their 1994 study, Roger Buehler, Dale Griffin, and Michael Ross found that people predicting their own task-completion times focused mainly on future scenarios rather than relevant past experience, and their predictions were optimistic for several kinds of task.

That finding does not mean every optimistic estimate is wrong, or that a previous duration dictates the next one. It does mean that a useful estimate needs two sources of information: the actual differences in the new task and records of completion for genuinely comparable prior tasks. A plan that contains only the first source can be coherent while remaining untested against usual duration.

Define comparable before looking at three past tasks

Consider a fully fictional work record, not site data, a performance commitment, or a research sample. The new task is a quarterly briefing for an existing client using a fixed structure. A prior task enters the comparison set only if it used the same delivery template, information from the same kind of system, one primary writer, one review of the same level, and no new regulatory research or data migration. Those conditions are an editorial definition of comparability for this example.

Under that definition, three fictional completed tasks took 6, 9, and 12 working days. Ordered from smallest to largest, the middle value is 9, so the median is 9 working days. The initial estimate for the new work is 5 working days. The number nine is not a promised completion date. It is a reference from three limited, comparable records placed beside the future story. The 6-to-12-day range also shows that duration varied even inside the defined set.

Record or estimateWorking days and meaning
Prior task A6; met the stated comparable-task conditions.
Prior task B9; met the stated comparable-task conditions.
Prior task C12; met the stated comparable-task conditions.
Initial estimate for the new task5; a future plan, not a completed historical record.

Adjust for differences instead of applying one median everywhere

Next, compare the new task with the history set item by item. Suppose the new client materials are more complete, which may shorten organisation work, but one extra cross-team confirmation is required, which may add waiting and rework. Do not select only the detail that supports a five-day schedule, or only the detail that supports delay. Record the difference, its possible direction, and what remains uncertain so the discussion identifies which part of the duration may change.

For this fictional example, a team might retain 6–12 working days as the historical range, record the complete materials and extra confirmation, and agree to revise the remaining estimate after the draft. A careful update would say: ‘Comparable tasks took 6–12 working days, with a median of 9. This task has two known differences; 5 days remains an initial plan to revisit after confirmation.’ That is a practical editorial method, not a formula guaranteed by the study.

In Study 4, under particular experimental conditions, the researchers instructed participants to connect relevant past experience with their predictions and the optimistic bias in that sample was eliminated. This supports deliberately bringing relevant history into an estimate. It does not show that calculating any median will always prevent underestimation or ensure that every team finishes on time.

Do not force unlike history into one number

‘We have made a briefing before’ is not enough for comparability. A different template, team size, review level, source-data quality, or one-off outage can change duration substantially. Comparability does not require every condition to match, but it requires naming conditions likely to matter before retaining a case, separating it, or using it only as a risk signal. Mixing unlike cases into one number creates false precision.

Keep the time unit visible as well. The 6, 9, and 12 here are working days, not calendar days and not person-hours. A record involving two editors working in parallel cannot be directly pooled with one primary writer’s elapsed days. Recording start and end definitions, waiting time, rework, and staffing makes a later ‘relevant past experience’ inspectable. The Management Development assessment can prompt reflection on planning, coordination, and task follow-through. It does not predict a project date or establish that these fictional records or the study necessarily apply to a particular task or team outcome.

Want to see where you land?

Take the Management Development and get your results in minutes.

Take the Management Development →

← Back to the blog