The Ninety-Ninety Rule: Why the Last 10% Moves Your Release Date
The ninety-ninety rule says the last 10% takes another 90% of the schedule. Here’s why “almost done” hides erosion of a committed release date.
The ninety-ninety rule says the last 10% takes another 90% of the schedule. Here’s why “almost done” hides erosion of a committed release date.
A hardening sprint is schedule buffer the confidence vote already priced. Here is what quietly spends it before the last iteration starts, and how to read what is left in days.
Escalation of commitment is why a feature nobody believes in still ships in the release. What makes agile teams especially vulnerable, and the four mechanics that make cutting scope a normal decision instead of an admission of failure.
A velocity forecast gives you one date. Three-point estimation gives you a range and the width of your own uncertainty — which is the number stakeholders actually need before they commit to a release.
The LinkedIn debate over whether to re-estimate carryover stories has two camps and one missing question: what is the number for? Here is how to decide, and how to keep carryover from silently eating next sprint’s capacity.
SPI 0.87 says you’re behind plan — not when you’ll land. Translate earned value and schedule variance into a committed release date on a live 50/85/95% confidence band in Jira.
SAFe fixes the train’s departure date — so post-commitment erosion decides what’s on board instead. Keep the PI confidence vote live against a 50/85/95% band in Jira, and choose your scope in days, before the boundary chooses for you.
A premortem surfaces how your release could fail — scope, dependencies, capacity. Give that judgment a pulse: track it live against the committed date.
A change control board approves scope one small change at a time. Bring the live Monte Carlo confidence band into the meeting and price each approval in days — before the vote.
The planning fallacy isn’t proof your team can’t estimate — it’s a missing outside view. Forecast the committed release date from your team’s real Jira throughput, on a live 50/85/95% confidence band.
© 2026 — Divim, Inc. Digital Visual Management Systems and Consulting, Inc.