Why estimates drift

Teams often underestimate because they silently assume favourable conditions and treat a multi-step item as one simple task. The video recommends two complementary corrections.

Technique 1 — question assumptions

Before estimating, ask:

• What must already be true for this task to start?
• Which dependency, approval, dataset, account or environment could be unavailable?
• What quality, security, accessibility or testing work is hidden in “done”?
• Is the estimate based on the person most familiar with the work rather than whoever may actually do it?
• What could invalidate the proposed approach?

Make the assumptions explicit in the ticket. Unknowns that materially affect effort should become a spike, dependency or risk—not invisible optimism.

Technique 2 — unpack the work

Break the item into observable steps before estimating:

• understand the requirement and acceptance criteria;
• design or prototype;
• implement;
• integrate data/services;
• test and fix;
• document, deploy and verify.

Unpacking reduces the planning fallacy because the team estimates the actual sequence instead of an abstract label.

Warning signs of chronic underestimation

• work repeatedly rolls into the next sprint;
• “almost done” tasks hide testing or integration;
• estimates are reused without revisiting assumptions;
• one person estimates for the whole team;
• work items are too large to expose uncertainty.

FIT5120 estimation checklist

• Acceptance criteria and Definition of Done are visible.
• Dependencies and assumptions are named.
• The task is small enough to complete and verify inside the iteration.
• Design, testing, accessibility, security, deployment and documentation are included.
• The team—not one specialist—has discussed the estimate.
• New evidence triggers re-estimation or backlog refinement.

Source and connection

• Two Ways to Create Better Sprint Task Estimates — Mountain Goat Software
• Reuse the Week 7 pages on Story Points & Relative Estimation and Desk Checks: relative sizing supports planning; unpacking and desk checks expose the evidence needed to calibrate it.