How to run a Scrum or Kanban team: sprints, WIP limits and cycle time explained
The short answer
To run a Scrum or Kanban team, agree one goal and one Product Owner, put all the work on one board, limit work in progress, and finish before you start. Plan short sprints from real capacity, review working results with users, and improve one or two things each time. Measure cycle time and throughput, and forecast with ranges.
Want the tools, not just the method? The The Agile and Scrum Team Kit kit turns this guide into ready-to-use templates, spreadsheets and printables, with an instant download.
- The 2020 Scrum Guide is still the current edition; an Expansion Pack is a companion, not a replacement.
- The Kanban Guide's four flow measures are WIP, throughput, work item age and cycle time.
- Little's Law links them: average WIP equals throughput times average cycle time in a stable system.
- Daily stand-ups go wrong when they become status reports to the facilitator.
- Forecast with a range from your own throughput, never a single promised date.
What are the steps to run a Scrum or Kanban team?
The steps are the same whether you call it Scrum, Kanban or simply 'agile', because both rest on the same short list of values: individuals and interactions, working results, customer collaboration and responding to change, with the closing reminder that plans and documents still matter, just less than those four. The difference between the frameworks is how you put the values into practice, and the steps below cover both.
Do them in this order. Measuring first stops you mistaking a quiet month for success. Naming one Product Owner before you build a backlog stops three managers each setting priorities. And limiting work in progress before you plan your first sprint stops a very common failure: a full board, a busy team and very little finished.
- 1. Capture a baseline: export eight weeks of work items and measure cycle time, throughput and work in progress.
- 2. Choose Scrum, Kanban or Scrum with flow metrics by looking at how your work really arrives.
- 3. Name one Product Owner, a Scrum Master (or flow manager) and the people who do the work.
- 4. Build one ordered backlog of user stories with testable acceptance criteria.
- 5. Design the board: start and finish points, columns and a WIP limit for each in-progress column.
- 6. Plan a short sprint from real capacity with one Sprint Goal, or set a weekly replenishment rhythm for Kanban.
- 7. Meet daily at the board about the goal; review working results with users; run a retrospective with one or two owned actions.
- 8. Track four numbers and forecast with ranges; re-score your practices every quarter.
Should you choose Scrum or Kanban?
Scrum is a lightweight framework built on Sprints, fixed timeboxes of one month or less, with three accountabilities (Product Owner, Scrum Master and Developers) and four events: Sprint Planning, the Daily Scrum, the Sprint Review and the Sprint Retrospective. Each of its three artifacts has a commitment: the Product Goal, the Sprint Goal and the Definition of Done. It suits work that serves one product goal, where stakeholders want to see and steer results every couple of weeks.
Kanban is a strategy for optimising the flow of value. The Kanban Guide asks for three practices: define and visualise the workflow, actively manage the items in it, and improve it. It has no required roles or sprints. It suits work that arrives continuously, in different types and sizes, with urgent items that cannot wait for the next sprint: support, HR, content and operations teams often fit here.
Many teams blend the two. A product team runs Scrum's events and adds Kanban's WIP limits and flow measures to the sprint board. A support team runs Kanban but borrows a monthly review with stakeholders. Ask four questions: is there one product goal behind most of the work; must much of it start the same day; do stakeholders need to see and steer results; and is work in progress already high? The last question is the most important, because high work in progress hurts every framework equally.
How do you write a user story the team can actually build?
Use the three-part pattern: as a [real user], I want [one capability], so that [their benefit]. The 'so that' is what lets a Product Owner order the backlog and lets the team suggest a cheaper way to reach the same outcome. If 'I want' contains an 'and', you may have two stories.
Then add two to five acceptance criteria a tester could check. Given/When/Then works well: given a starting state, when an action happens, then an observable result. 'Given a photo over 15 MB, when I upload it, then I see the size limit and a tip to compress it' is testable. 'The upload is fast' is not. Avoid vague words such as fast, easy, user-friendly and appropriate unless you replace them with a number.
Keep stories small: a good rule is that any story bigger than half a sprint gets split, by workflow step, by business rule, happy path first or by platform. Never split by technical layer, because neither half is valuable to a user on its own.
How do you set a WIP limit and why does it work?
A work-in-progress (WIP) limit caps how many items can be in a column at once. When the column is full, nobody starts new work; people help finish, review or unblock what is already there. The Kanban Guide does not prescribe a number: it asks teams to control WIP explicitly and write down how.
Little's Law explains why it works. In a stable system, the average number of items in progress equals throughput multiplied by the average time each item spends in progress. Rearranged, average cycle time equals average WIP divided by throughput. If throughput stays about the same and you halve the work in progress, items finish in roughly half the time.
Start with a limit clearly below today's actual WIP, so it bites without leaving people idle on day one. For a team of seven, five in progress is a reasonable first choice because it nudges people to pair. Review the limit after two sprints with your cycle-time numbers in front of you, and make the rule explicit in the team's working agreement: when the column is full, we swarm.
How do you calculate cycle time, throughput and a forecast? (worked example)
Cycle time is the number of days from when an item started to when it finished. Count both days, so an item started and finished on the same day has a cycle time of one. Throughput is the number of items finished per week. Average WIP is the average, over every day in the period, of items started and not yet finished. Report the median and the 85th percentile of cycle time, not the average, because a few slow items distort an average.
The table shows a sample team's numbers from an eight-week ticket-queue baseline and the six two-week sprints that followed. In the baseline, Little's Law predicted about 5.6 items in progress from throughput and cycle time, but the actual average was 12.1. The gap was eight stale tickets that had been started and never finished, quietly holding the team's attention. Closing them was the first flow fix.
For a forecast, use ranges. A burn-up divides the remaining work by your slowest and fastest recent sprints. A Monte Carlo forecast replays your real daily throughput thousands of times: for 12 more items, the sample team had a 50% chance of finishing within 17 days and an 85% chance within 21 days.
| Measure | Baseline (8 weeks) | Six sprints (12 weeks) | Sprint 6 |
|---|---|---|---|
| Items finished | 18 | 59 | 12 |
| Throughput per week | 2.25 | 4.92 | 6.0 |
| Median cycle time (days) | 15 | 6 | 4.5 |
| 85th percentile (days) | 28.8 | 11 | 8.3 |
| Average WIP | 12.1 | 5.1 | 6.9 |
Reading helps; measuring tells you what to work on. These AI-graded assessments on AssessAll pair with this topic:
- Project & Delivery Leadership (AssessAll)
- Product Management Judgment (AssessAll)
- Task Prioritisation & Planning (AssessAll)
How should the daily scrum and retrospective run?
The daily scrum is a 15-minute meeting for the people doing the work to inspect progress toward the Sprint Goal and adapt the plan. It is not a status report. Researchers who observed 102 daily stand-up meetings in 15 teams across five countries found that the 'what did I do yesterday' question tended to turn into status reporting, and that meetings addressed to the facilitator became status reports. They suggested dropping that question, rotating the facilitator, holding the meeting just before lunch and adapting how often you meet.
A practical alternative is to walk the board from right to left: what is closest to done, and what is blocking it? Keep problem-solving for after the meeting, with only the people who need to be there.
In the retrospective, start by checking last sprint's actions, then gather data, find causes and agree one or two actions with an owner, a due date and a way to know they worked. A case study of retrospectives in a New Zealand company found that the one team that agreed no action points never reached the deepest level of reflection, and the authors call following up actions critical. Change the format every few sprints: start-stop-continue, the 4 Ls, sailboat and timeline all work.
Why do estimates go wrong, and should you use story points?
People are optimistic about their own tasks. In a classic study, students predicted they would finish their theses in about 34 days on average; the real average was about 55, and only around 30% finished by their own best guess. The researchers found people plan from the scenario in front of them rather than from how long similar work took before.
Story points are a relative size: an item is a 5 because it feels like other 5s, not because it takes five hours. Planning poker, where everyone reveals a card at the same time, stops anchoring and surfaces risk. Points are useful for a team's own planning, but they mean nothing across teams. Never convert them to hours, never compare velocity between teams and never make velocity a target.
If your items are small and similar, you can skip sizing and forecast from how many items finish each week. Many teams use both: points for the planning conversation, item counts for the forecast.
What are the most common mistakes?
Most failures come from copying the meetings without the ideas behind them. These are the ones that show up again and again in new teams, with the fix for each.
- Installing stand-ups and sprints while keeping top-down task assignment: fix the ownership first.
- A committee as Product Owner: one person, with time and authority, orders the backlog.
- No Definition of Done: write 6 to 10 lines the team can meet today.
- Overcommitting the first sprint: plan from capacity and mark the items to drop first.
- WIP limits that nobody honours: check the limit first thing at the daily meeting.
- Retrospective actions that vanish: one or two, owned, dated and checked next time.
- Velocity league tables: report cycle time, throughput, goals met and escaped defects instead.
- Configuring the tool before agreeing the workflow: design on a wall, then configure.
Which tools work: Jira, Asana or Trello?
All three can run Scrum or Kanban; decide your workflow first. Jira's CSV importer needs a Summary on every row and can import epics with their stories using Work item ID and Parent columns; company-managed boards can show a maximum number of items per column and turn the header red when it is exceeded. Asana imports CSV files in a documented column order with month/day/year dates and parent tasks above their subtasks. Trello says it has no generic import tool, but pasting several lines into a card creates one card per line.
Menus, plan limits and field names change often, so check each vendor's official help page before you rely on a detail, and keep personal data out of ticket titles and attachments unless your tool is approved for it.

Run the whole method with working files
The Agile and Scrum Team Kit gives you the workbook, cycle-time calculator, backlog import files, templates and six-sprint plan used in this guide.
Sources
- Ken Schwaber and Jeff Sutherland (scrumguides.org): The Scrum Guide: The Definitive Guide to Scrum, the Rules of the Game
- Beck, K. et al. (agilemanifesto.org): Manifesto for Agile Software Development
- Coleman, J., Vacanti, D. et al. (kanbanguides.org): The Kanban Guide (v2025.5)
- Operations Research (INFORMS): Little, J. D. C. A Proof for the Queuing Formula: L = lambda W. Operations Research 9(3): 383-387
- IEEE Software (preprint on arXiv): Stray, V., Moe, N. B. and Sjoberg, D. I. K. Daily Stand-Up Meetings: Start Breaking the Rules
- Journal of Personality and Social Psychology 67(3): 366-381: Buehler, R., Griffin, D. and Ross, M. Exploring the 'planning fallacy': Why people underestimate their task completion times
- XP 2017, Lecture Notes in Business Information Processing 283 (Springer): Andriyani, Y., Hoda, R. and Amor, R. Reflection in Agile Retrospectives
- Information and Software Technology 50(9-10): Dyba, T. and Dingsoyr, T. Empirical studies of agile software development: A systematic review
- Atlassian Support (Jira Cloud): Prepare a CSV file for import
- Atlassian Support (Trello): Import data into Trello
More from the Bodhih family
Questions people ask next
Is the 2020 Scrum Guide still the current version?
Yes, as of November 2026 the November 2020 Scrum Guide is the current edition. A Scrum Guide Expansion Pack by Jeff Sutherland and colleagues presents itself as a companion to the 2020 Guide, not a replacement.
How long should a sprint be?
One month or less. Two weeks is a common starting point. Keep the length fixed for at least three sprints so your numbers mean something.
What is a good WIP limit for a team of seven?
There is no universal number. Start clearly below your actual work in progress; five in progress for seven people is a reasonable first choice. Adjust after two sprints using cycle time.
What is the difference between cycle time and lead time?
Cycle time runs from when work starts to when it finishes. Lead time usually runs from when the request was made, so it includes waiting in the backlog.
Can a non-software team use Scrum?
Yes, when the work serves one goal and stakeholders want regular reviews, such as a product launch or a new process. Continuous request work usually fits Kanban better.
Who should be the Product Owner?
One person with the time and authority to order the backlog and say 'not now'. The Scrum Guide is explicit that the Product Owner is not a committee.
Should managers attend the retrospective?
Only if the team invites them. Psychological safety, a shared belief that it is safe to take interpersonal risks, was linked to team learning in a study of 51 work teams.