Every agile team has seen the drawing. Two curved lines flare out from today and converge on a delivery date somewhere in the future: the cone of uncertainty. Early in a release, the honest range around “when will this ship?” is wide — the classic software-estimation research put it as wide as 4x in either direction at project start. As the work progresses, the range tightens until, on release day, it collapses to a single point.
It’s a genuinely useful picture, and it has earned its place on the wall of every PI planning room. It’s also quietly misread in almost every one of those rooms — because the comforting version of the story says the cone narrows with time. Six weeks from now, we’ll simply know more. The range will tighten on its own. Just keep working.
It won’t. The cone narrows with evidence — delivered work, measured throughput, surfaced risk — and only when someone actually feeds that evidence back into the forecast. A release plan nobody has re-forecast since commitment day isn’t further down the cone. It’s still standing at the wide end, holding a day-one snapshot with a confident band drawn around it.
Your vote was right — at the widest point of the cone
When your team holds its confidence vote at PI planning, that vote is expert engineering judgment applied at exactly the moment uncertainty peaks. The team weighs the throughput history it trusts, decides whose capacity is real, argues about what’s genuinely in scope — and commits to a date with its eyes open. That is the SAFe confidence vote working as designed, on evidence the team validates. (We’ve written about what that day-one vote gets right in Your Fist of Five Was Right on Day One.)
There is nothing wrong with that judgment. The problem is what happens to it afterward: usually, nothing. The vote gets averaged, recorded, and left on the wall — a reading taken at the cone’s widest point, treated as if it were still current in week four, week eight, week twelve.
What actually re-widens the cone
Meanwhile, three forces go to work on the committed date — the three things that erode a committed release date: scope creep, dependencies surfaced too late, and capacity changes. Each one moves the real delivery range while the recorded snapshot stays green. On those weeks the cone isn’t narrowing at all. It’s being pushed back open — invisibly, because nobody is re-measuring it.
That’s the trap in the classic diagram: it shows uncertainty decreasing smoothly and monotonically, as if learning only ever flows one way. Real releases don’t behave like that. A cross-team dependency that surfaces in week six can throw the range wider than it was on commitment day. Two “small” scope additions and a departed senior engineer can do the same. The team that believes the diagram keeps reporting the day-one number anyway, because it’s the only number anyone wrote down.
Narrowing the cone on purpose
None of this requires abandoning the vote or the diagram. It requires doing what the diagram assumed someone was doing all along: re-forecasting continuously as evidence arrives. That’s arithmetic — and it’s arithmetic no human should be re-running by hand every week for every release.
Advanced Release Planning for Jira re-runs a Monte Carlo forecast against your team’s live Jira throughput and expresses the result the way the cone actually works: a committed date with a 50/85/95% confidence band around it. Every sprint of delivered work is new evidence, and the band tightens because the forecast is current — not because the calendar moved. Your team still owns every assumption: which throughput window to trust, whose capacity counts, what’s in scope. The app does the arithmetic at Monte Carlo scale and keeps the reading live.
And when the erosion forces push back, you see it in days, attributed: the scope added this week cost four days; the dependency that surfaced Tuesday cost three; the capacity change costs two more. The band widens honestly on screen instead of silently in reality — which means you can renegotiate scope or dates before the committed date slips, not explain the slip after. A committed range that narrows as risk clears is also the more honest way to promise in the first place — that’s the argument in Stop Promising a Release Date. Promise a Release Window.
Keep the vote live
The cone of uncertainty was never a promise that time heals estimates. It’s a description of what disciplined re-forecasting looks like from above. Your team’s confidence vote deserves better than to age on a wall — it deserves to stay as current as the release it describes.
See how live confidence bands work in the docs, or try Advanced Release Planning free for 30 days — and watch your cone narrow on evidence, not optimism.
Leave a Reply
Your email is safe with us.