Decompose tasks and sum the pieces

Estimate each sub-task independently, then add them up — the sum is closer to truth than a top-down estimate.

Why it works

Top-down estimates are anchored on best-case scenarios. Decomposition forces you to confront each step concretely, which surfaces tasks you would otherwise ignore entirely. Research on the "unpacking effect" shows that explicitly enumerating components increases perceived likelihood or time allocation for the whole — because each component is a reminder of what has to go right. Summing component estimates also catches the planning fallacy’s tendency to skip "unknown unknowns" that only surface when you think about execution in detail.

How to do it

1. Break the project into its smallest actionable steps — not phases, but individual tasks.
2. Estimate each step’s time independently, without looking at the running total.
3. Sum them up; add 20–50% for integration and hand-off friction.
4. Compare the sum to your original top-down estimate; investigate large gaps.

Evidence

The unpacking effect in probability judgment (Tversky & Koehler, 1994) shows enumerated components raise aggregate estimates. Software estimation research (Jorgensen, 2004) finds bottom-up estimation is more accurate than top-down for task work, though overruns still occur. (observational)

Decomposition helps but does not eliminate overruns; it mainly prevents the biggest systematic misses caused by unacknowledged steps.

Sources

• Tversky & Koehler (1994), Support theory: A nonextensional representation of subjective probability, Psychological Review

Common mistake

Decomposing phases ("planning", "execution", "review") instead of tasks — coarse decomposition preserves the optimism bias on each phase.