Sprint velocity predictability: what the confidence band actually tells you

Velocity with Confidence Band plots completed work per sprint with a shaded ±1 standard-deviation band around it. Ignore the line for a second — sprint velocity predictability lives in the band's width. Narrow means you can plan on this team. Wide means you're guessing.

deckgauge · Board · Intelligence

Completed points per sprint with the ±1σ band.

What it measures, and where the data comes from

One point of completed work per closed sprint, connected into a line, plus a shaded band showing ±1 standard deviation of that number across your recent sprint history. The line is the average; the band is the spread around it.

  • Jira — completed story points (or issue count, if your team doesn't point) summed for each closed sprint.
  • Azure DevOps — the same, summed over closed iterations.
Why not GitHub or GitLabVelocity needs a sprint boundary to mean anything. GitHub Issues and GitLab Issues don't have one, so this widget only lights up for boards synced from Jira or ADO with sprints configured. No sprint data, no widget.

How to read it

Look at the band first, the line second. The line tells you how much got done. The band tells you whether you can trust that number to repeat. A team that ships 40 points one sprint and 15 the next has an average of 27.5 that means almost nothing — and the band will show it, wide and ugly.

A tight band around a modest average is a better position than a wide band around a high one. You can plan against the first. The second is a coin flip with extra steps.

Be honest about what's underneath it, too. Velocity — raw or banded — is trivially gameable. Inflate estimates, split tickets to pad the count, or quietly stop taking on the hard, unpredictable work, and the number goes up without anything actually changing. The band doesn't fix that, but it does make one kind of gaming harder to hide: a team manufacturing a stable-looking average by re-negotiating story points every sprint usually leaves a jagged trail in the underlying spread before the average settles down. If the band looks suspiciously perfect the same sprint someone got nervous about a review, look closer.

What it tells you over time

Watch the band's width sprint over sprint, not any single reading. A band that's been wide for months isn't bad luck — it's the team's actual operating condition, usually driven by unstable scope, constant interrupts, or a roster that hasn't settled. A band that narrows over several consecutive sprints is the more interesting story: it usually means the team found a rhythm — steadier scope, fewer surprise fires, a stable roster — even if the average velocity itself barely moved.

Don't over-read one clean sprint. One narrow reading after months of chop is noise. Three or four in a row, with the average holding, is a trend worth acting on.

Example situations

1. A wide band that never closes

deckgauge · Velocity with Confidence Band

Wide band, jagged completed-work line, sprint after sprint.

What you're seeing: the shaded band covers most of the chart's height, and the completed-work line whipsaws inside it — a big sprint, then a bad one, then a big one again. There's no rhythm here. Whatever the average says, this team's actual output on any given sprint is close to unknowable in advance.

How to react: stop planning to the average — it's a mirage built from two very different kinds of sprints. Before you touch process, find out why: mid-sprint scope changes, a rotating on-call load, unplanned production fires, or estimates that were never calibrated in the first place. Pair this with Work in Progress and Investment Allocation; a spiky WIP count or a KTLO share that swings wildly sprint to sprint usually explains a wide band better than "the team is inconsistent."

Managerial playPlan commitments to the lower edge of the band, not the average — that's the number you can actually defend to a stakeholder. Put a hard rule on mid-sprint scope changes: anything new that lands after planning bumps something else out, it doesn't just stack on top. Give the team one full quarter of protecting sprint scope before you even think about asking for more throughput — you can't push speed on a foundation that isn't stable yet.

2. The band tightening after a rocky stretch

deckgauge · Velocity with Confidence Band

Band narrowing sprint over sprint; average holding flat.

What you're seeing: the band has been closing in for several sprints running, and the average line barely moved. That combination — flat average, shrinking spread — is a team that's found its footing. Whatever caused the chop before (a reorg, ramp-up, churny scope) has settled.

How to react: don't dismiss this because "the number didn't go up." A tightening band is arguably the better news of the two — it means you can finally rely on this team's word. Confirm it's real, not a one-sprint fluke, by checking Investment Allocation for a steadier feature/KTLO mix over the same window.

Managerial playOnce the band's held tight for three or four sprints, start using its floor as the number you commit externally — not the average, and definitely not the best sprint on record. This is also the moment to make real promises: a roadmap date, a cross-team dependency commitment, something you'd have hedged on six months ago. Say out loud in the retro what changed to get here, so the team knows which habits to protect when the next disruption shows up.

Frequently asked

What is sprint velocity predictability?
How reliably a team delivers a similar amount each sprint. The widget plots completed work per sprint with a shaded band of one standard deviation either side of the line.
Should I read the line or the band?
The band. Predictability lives in its width, not in the height of the line. A narrow band means you can plan around this team; a wide one means any forecast is a guess regardless of the average.
How do I narrow a wide confidence band?
Reduce variance in what is committed rather than trying to raise the average. Consistent slice sizes and fewer mid-sprint additions tighten the band faster than any push on output.

Related widgets

Last updated