Short answer: right-size backlog work by estimating epics coarsely, splitting them into stories small enough that no single story consumes more than about a third of one sprint’s adjusted capacity, and using subtasks only to make assignment and days-off visible. Then commit sprints against capacity that already subtracts holidays, leave, and training, not against raw velocity.
Picture a state parks agency with a hiring freeze. Its six-person digital team inherited an epic called “Make reservation checkout WCAG 2.1 AA compliant,” estimated once, two years ago, at “about a quarter.” Leadership has promised the fix before the next booking season. Nobody on the team can say whether that promise is realistic, because the epic has never been broken down into anything a sprint can actually hold.
That is the right-sizing problem, and it is common in public-sector backlogs: work arrives as large, mandate-shaped epics, and the gap between the epic and the sprint is where predictability gets lost.
Why is right-sizing backlog work harder in government?
Commercial teams can often absorb a mis-sized epic by adding people or moving a launch date. Agencies usually can’t. Headcount is fixed or shrinking, budgets are set by appropriation, and deadlines are tied to fiscal years, statutory dates, or a season the public notices. Fewer people with the same mission means every oversized story that rolls over is a visible miss, and it lands on a status report that travels up the chain.
There is also a records angle. When a program manager is asked why a commitment slipped, “the epic was bigger than we thought” is a weak answer. A backlog that shows how the epic was decomposed, what each piece was estimated at, and what capacity the team had when it committed is a much stronger one.
How do you break an epic down into sprint-sized work?
Use a simple, repeatable five-step framework. It works whether your team estimates in story points or hours.
- Estimate the epic in ranges, not points. At the epic level, a T-shirt size or a range (for example, 60–100 points) is honest. False precision at this level is what creates the “about a quarter” promises that later fail.
- Split by outcome, not by function. Slice the epic into stories that each deliver something testable to a user or reviewer, such as “keyboard-only users can complete date selection,” rather than “front-end work” and “back-end work.” Outcome slices can be accepted, demonstrated, and reported independently.
- Set a size ceiling from capacity. Calculate the team’s adjusted sprint capacity (Step 4), then cap any single story at roughly one third of it. Anything larger gets split again before it can enter a sprint.
- Calculate adjusted capacity before you commit. Start from average velocity over the last three sprints, then scale it by the share of person-days actually available after federal holidays, approved leave, mandatory training, and details to other offices.
- Use subtasks for visibility, not estimation. Subtasks are where you assign named people and see who is overloaded. Estimating at the subtask level as well as the story level double-counts effort; keep the commitment math at the story level.
What size should a user story be in a two-week sprint?
A practical rule: a story should be finishable by one or two people within the sprint, with room for review. Tying the ceiling to adjusted capacity, instead of a fixed number, means the rule tightens automatically in short or holiday-heavy sprints, which is exactly when oversized stories do the most damage.
- Too big: the story needs more than a third of adjusted capacity, or its acceptance criteria span several user outcomes.
- About right: one clear outcome, acceptance criteria a reviewer can check, and an estimate well under the ceiling.
- Too small to track separately: it is really a checklist item; make it a subtask.
Worked example: the reservation checkout epic
The parks team has six people and two-week sprints of 10 working days, so 60 person-days per sprint. Velocity for the last three sprints was 34, 32, and 30 points, an average of 32.
The upcoming sprint includes a federal holiday (6 person-days), approved leave for one developer (3 days), and a mandatory security-awareness training block (3 person-days). That removes 12 person-days, leaving 48 of 60, or 80% availability. Adjusted capacity is 32 × 0.8 = 25.6, so the team plans to 25 points. The story ceiling is about a third of that: 8 points.
Refinement turns the epic into eight stories totaling 71 points. One of them, “screen-reader support across the full checkout flow,” is estimated at 21 points, well over the ceiling, so it is split into three stories of 5, 8, and 5. The team then commits 24 points (8 + 5 + 5 + 3 + 3), leaving one point of buffer, and the forecast for the whole epic is 71 ÷ 25 ≈ 3 sprints. Leadership now gets a dated, defensible answer instead of “about a quarter.”
Making right-sizing repeatable inside Jira
The method above breaks down when the pieces live in different places: estimates in Jira, leave in a shared calendar, training dates in an email, and the capacity math in someone’s spreadsheet. As agencies modernize onto Atlassian Cloud, including Atlassian Government Cloud (AGC), it makes sense to bring that math into the backlog where the work already lives.
Sprint planning with Capacity Planning for Jira puts sprint and team capacity on one screen in the Jira backlog, using past velocity, days-off, and allocation, so you can see whether a sprint fits before you commit. That way an oversized commitment or an over-allocated person shows up during planning, not at the sprint review. Because the plan is built in Jira, the team keeps a record of who was allocated, which days-off were subtracted, and why it committed to a given number of points. That record supports the documentation and decision trail your leadership and oversight reviewers expect, though the app itself is a planning tool, not a compliance product. The app is Cloud Fortified and runs on Atlassian’s SOC 2 Type II / ISO 27001 platform.
FAQ
Should we estimate epics in story points?
Use ranges or T-shirt sizes at the epic level, and convert to story points only once the epic is split into stories. Point estimates on unrefined epics tend to become promises nobody agreed to.
Do subtasks need their own estimates?
Not for commitment math. Estimate stories, and use subtasks to assign work to named people and spot over-allocation. If your team tracks hours on subtasks for reporting, keep the sprint commitment at the story level so effort isn’t counted twice.
How small is too small for a story?
If a story has no independently testable outcome, or would take less than a few hours, it is usually a subtask of a larger story rather than a backlog item of its own.
Try it on your next refinement session
A practical starting point: before your next sprint planning, run a quick right-sizing readiness check. Pick your largest open epic, compute adjusted capacity using the worked example above, set the one-third ceiling, and count how many stories exceed it. That number tells you how much refinement you need before you can make a commitment you can defend.
Try it: Install Sprint planning with Capacity Planning for Jira free · Help docs




Leave a Reply
Your email is safe with us.