Productivity advice usually starts after the plan is made, but decades of research suggest many missed deadlines begin earlier, with an optimistic estimate of how long the work will take

There is a familiar sequence after a deadline slips. The calendar is tightened, the task list is rebuilt and the person doing the work starts wondering whether they lacked discipline.

All of that happens after a quieter decision has already shaped the result: someone decided how long the work ought to take.

That estimate determines what else gets promised, how many tasks fit into the week and whether an ordinary interruption becomes a manageable delay or a small crisis. Yet productivity advice tends to treat the estimate as neutral. It concentrates on execution, as if the plan arrived from somewhere outside human judgement.

I think the research on time prediction makes that picture harder to defend. It does not say every late project began with a bad forecast, or that effort never matters. It suggests that some apparent failures of execution are better understood as failures of prediction.

The thesis study that defined the pattern

In 1994, Roger Buehler, Dale Griffin and Michael Ross published a series of studies in the Journal of Personality and Social Psychology under the title “Exploring the planning fallacy”. The best-known part followed 37 psychology students completing their honours theses.

The students’ average best estimate was 33.9 days. Their average completion time was 55.5 days, and only about 30 per cent finished by the date they had predicted. Even the average worst-case forecast, 48.6 days, came in roughly a week short of the eventual result.

Thirty-seven students are not a workforce, and an honours thesis is not a product launch. The paper mattered because the same broad pattern appeared across several experiments involving both academic and everyday tasks. Participants’ forecasts for their own completion were generally too optimistic, while their estimates for other people were less biased.

The researchers called this the planning fallacy: people can know that similar work ran late before and still believe the current project will be different.

A plan is a story about an uninterrupted future

When the 1994 researchers asked participants to think aloud, most of the reasoning concerned future steps. People pictured how they would proceed through the current task. Far fewer referred to how earlier tasks had actually unfolded.

That makes intuitive sense. A current project arrives with specific requirements, and a planner has to imagine doing them. The problem is that a mental sequence usually contains the work itself. It may not contain the meeting that runs long, the response that takes two days to arrive, the first approach that fails or the smaller obligation that becomes urgent halfway through.

Each interruption can be genuinely unusual on its own. Taken together, however, interruptions are a normal property of work.

This helps explain why experience does not automatically settle the matter. A delayed project can be remembered as an exception caused by a difficult client, a sick colleague or an unexpected technical fault. Those explanations may be accurate. They also make it easy to treat the next clean plan as the realistic one and the untidy past as irrelevant.

The wider evidence is less tidy than the slogan

“People always underestimate tasks” is a stronger claim than the literature supports. Torleif Halkjelsvik and Magne Jørgensen reviewed research ranging from origami exercises to software projects for a 2012 paper in Psychological Bulletin. Their review found underestimation more often than overestimation in engineering and management studies, but not in the psychology literature as a whole.

The distinction between work hours and calendar completion also matters. A task might require eight focused hours but still take four working days to finish because the hours are fragmented or because progress depends on other people. Asking “How much effort will this take?” is not the same as asking “On what date will it be done?” Many studies, the reviewers noted, do not keep those questions separate.

Small tasks can be overestimated. Familiar and repetitive work is easier to forecast than novel work. Some apparent patterns can also emerge from random error rather than a stable bias. The review is useful precisely because it resists turning the planning fallacy into another universal productivity rule.

The narrower reading holds up better: judgement-based estimates are sensitive to task type, framing, incentives and the information available when the estimate is made. In project settings, optimistic errors appear often enough to deserve attention, but not so consistently that one fixed multiplier can repair every schedule.

Workplace forecasts are social decisions

A time estimate is rarely just a private calculation. It can be a promise to a manager, a price in a client proposal or an answer given while other people wait for a date. That social setting can reward a confident number before anyone knows whether it is realistic.

Mario Weick and Ana Guinote examined one part of this dynamic in four studies published in the Journal of Experimental Social Psychology in 2010. Across several tasks and different ways of measuring or inducing power, participants in the higher-power condition gave more optimistic and less accurate time predictions. The authors linked the result to narrower attention on the envisaged goal, rather than to differences in mood or self-confidence.

Laboratory studies of power do not recreate a board meeting or a software sprint. They do show that forecasts can move with the position of the forecaster, even when the task itself has not changed.

Initial numbers can exert their own pull. In a 2019 laboratory experiment reported in the Journal of Economic Behavior & Organization, Matej Lorko, Maroš Servátka and Le Zhang asked 93 participants to estimate and then complete a repeated task. The researchers found strong effects from high and low numerical anchors. Those effects persisted across rounds, and a participant’s own earlier estimate could become an anchor too.

This is an uncomfortable feature of workplace planning. A date suggested casually at the start of a discussion can remain in the room long after everyone has gathered better information.

What a more defensible estimate contains

The early planning-fallacy experiments offer one useful distinction. Participants tended to improve when they were instructed to connect relevant past experience directly to the current prediction, rather than merely recalling that earlier projects had run late. The past had to become part of the estimate.

That approach is often described as taking an outside view. Instead of asking only how the current project should unfold, the forecaster also asks how long a reference class of similar projects took. The comparison will never be perfect, but it forces ordinary delay back into the calculation.

The 2012 review gives reason to be cautious about neat remedies. Breaking a task into components sometimes changes an estimate, but the direction can depend on whether the newly visible components are unusually long or short. Incentives, group estimation, the level of detail and the wording of the question can all matter. There is no research-backed rule that says every estimate should simply be doubled.

A defensible forecast therefore carries some of its uncertainty in public. It distinguishes uninterrupted effort from elapsed calendar time, identifies dependencies and shows the past projects used as comparisons. That will not make the future predictable. It makes the reasoning behind the date inspectable.

Productivity begins before execution

Once an optimistic date becomes a commitment, every later event is interpreted against it. Normal variation looks like lost focus. Waiting for information looks like poor momentum. A task that was never likely to fit into the available time becomes evidence that someone works too slowly.

None of this removes responsibility from the person doing the work. Estimates can be careless, and execution can be poor. Scope can also change after a date is agreed, priorities can conflict and organisations can knowingly set deadlines for reasons other than prediction.

But it changes the first question worth asking when work arrives late. Before rebuilding the productivity system, look at what the original estimate assumed and what evidence supported it.

Sometimes the work did not fail to follow the plan. The plan failed to describe the work.