Cone of Uncertainty
The Cone of Uncertainty is a model of estimation variance over a software project's lifecycle, articulated by Barry Boehm in 1981. Estimates at project inception have ~4× variance (a 100-staff-week budget may take 25-400); the cone narrows to 1× (the actual outcome) as the project progresses through requirements, design, coding, and acceptance.
The cone is descriptive, what estimation accuracy looks like in practice across a wide range of projects, not prescriptive. It does not say 'estimates must improve as the project proceeds' but rather 'across observed projects, they do, because uncertainty falls as decisions get locked in.' Boehm revisited the cone in 2002 with co-author Richard Turner, broadly replicating the shape on more recent projects. Modern Agile pushback: the cone assumes a single project trajectory with progressive lock-in; sprints introduce repeated cones at smaller scale. Either way, the cone's most useful implication is that early estimates are wide bands, not point predictions. Committing to a point inside the wide band is the planning anti-pattern that the planning fallacy explains.