Scrum vs Kanban: timeboxes vs WIP-limited flow
Scrum fixes the time and aims it at one goal; Kanban fixes how much is started and starts the next item when one finishes.
Predict first
Before you read: predict
On this lesson's flow model, in exam season ExamHall's support team works proctor escalations from one board. This morning it reads: Ready 8 · In progress 4 · Waiting for the proctor's reply 6. For three weeks the team has finished 2 escalations a day, and the board has kept this shape.
The team lead says: "Only 4 are in progress, so one we start today will be done in about 2 days." About how long will an escalation started today take to reach Done?
The 30-second version
Scrum works in Sprints of a month or less, each aimed at one Sprint Goal, with events to inspect and adapt; scope bends, the goal holds. Kanban defines a workflow, counts everything started and not finished as WIP, and starts new work only when there is room. At a steady finish rate, start-to-Done time is WIP ÷ throughput.
Scrum fixes the time and aims it at one goal; Kanban fixes how much is started and starts the next item when one finishes.
How it works
Step 1 of 15
In the kitchen
- a heavy check booked into a four-week slot; the next aircraft's slot starts the day it ends → a Sprint (step 1)
Analogy: An airline's maintenance hangar (aircraft maintenance)
A hangar does two kinds of work. Heavy checks are booked into fixed slots; quick repairs flow through four line bays.
| In the hangar | In software |
|---|---|
| a heavy check booked into a four-week slot; the next aircraft's slot starts the day it ends | a Sprint (step 1) |
| the slot's one aim: the aircraft back in service with its new cabin | the Sprint Goal (step 2) |
| engineers and the planner adjust the task list as panels come off; the aim stays | scope renegotiated, goal kept (step 4) |
| four bays on the line | the WIP control (steps 8, 9) |
| an aircraft waiting for a part still holds its bay | WIP in any state (step 8) |
| the next aircraft is towed in only when one rolls out | pull (step 9) |
| each bay's board: days since its aircraft came in | Work Item Age (step 10) |
| aircraft parked on the apron, waiting for a bay | Ready (step 14) |
Breaks down when: an aircraft is one job that cannot be split, while software work can be cut into smaller items and released one by one.
In real terms: count what is started, read the finish rate off the board, and divide, before adding people or starting more.
Examples, run for real
4.1 ExamHall EXH-4700 → MR !5000: escalations in exam season
In exam season proctors report 2 escalations a day, and the support team finishes 2 a day. Every report was started at once, so 12 were open and everything was late. MR !5000 changed the board's policy: at most 4 started, severity 1 pulled first, then the oldest.
EXH-4700 · proctor escalations in exam season · 2 reported a day · 2 finished a day
every report started at once started 12 · Ready 0 · start-to-Done 6 days · ordinary report to Done 6 days · severity-1 report to Done 6 days
MR !5000 · at most 4 started · severity 1 pulled first, then the oldest
WIP control 4 started 4 · Ready 8 · start-to-Done 2 days · ordinary report to Done 6 days (4 in Ready) · severity-1 report to Done 2 days
finished a day 2 before · 2 afterThe team finishes 2 a day under both policies. Each started escalation now takes 2 days instead of 6 (step 13). An ordinary report still waits 6 days in all, 4 of them in Ready (step 14): the limit alone did not shorten it. What changed is that Ready became a visible queue the team orders, so the severity-1 report went first.
Try it
Answer first, then check. Nothing is saved.
Question 1 of 7
On this lesson's flow model, a payments-ledger team's SLE reads: "85% of tickets finish within 8 days of starting." On Wednesday morning its board shows three open tickets. Age is the days since a ticket was started; work left is days of work for the one developer on it, who works on it every day from today until it is done:
- Refund rounding: age 7 days, 1 day of work left
- FX rates import: age 3 days, 6 days of work left
- Ledger export: age 1 day, 7 days of work left
Which ticket will reach Done more than 8 days after it started, unless someone helps?
Question 2 of 7
A payments-ledger team is on day 4 of a two-week Sprint whose Sprint Goal is "Refunds post to the ledger in one step." That morning the bank partner reports that yesterday's settlement file is missing a column. The fix is two days' work, touches nothing in refunds, and the team has no spare capacity this Sprint.
Which move does the Scrum Guide support?
Question 3 of 7
A payments-ledger team runs Kanban. On Tuesday its board shows tickets piling up in Review: the column's WIP control is 4, and tickets sit there for days. The team agrees that the control should be 2.
Which move fits the Kanban Guide?
Question 4 of 7
A payments-ledger team's two-week Sprint runs from day 1 to day 10, with the Sprint Review on day 10. Its Definition of Done for a ledger change has three checks: the code is reviewed, the nightly reconciliation run passes on staging, and the bank's sandbox accepts a test payout file. A payout-rounding fix is merged and reviewed on day 3, and that night's reconciliation run passes. The bank's sandbox accepts its test file on day 6. The bank wants the fix as soon as possible.
When may the team release it, by the Scrum Guide?
Question 5 of 7
An interviewer asks: "Your team has 12 tickets in progress and everything is late. Write your answer in at most 150 words.
What do you change?
Question 6 of 7
On Monday morning, on this lesson's flow model, a payments-ledger team of two developers has 10 small reconciliation tickets, each one day of one developer's work, and nothing else to do. Under plan X, each developer starts five tickets on Monday and splits every day evenly across them. Under plan Y, each developer works on one ticket at a time and starts the next only when it is Done.
Under plan Y, on average how many working days after Monday morning does a ticket reach Done?
Question 7 of 7
A payments platform team runs two-week Sprints and tracks each one on a board with To do, Doing and Done columns and no limit on any column. Every day two or three urgent tickets arrive from other teams (failed payouts, ledger mismatches), and each Sprint's plan breaks by its second week.
The lead says: "Scrum or Kanban: we have to pick one." Which reply do the Scrum Guide and the Kanban Guide support?
Go deeper
Pro tip
- Propose a WIP control before more people. When a team is late on everything, count what is started and the finish rate first: at the same throughput, halving what is started halves the time from start to Done; then order Ready so the urgent work goes first. When: the first answer to "everything is late". Status: SELF (the row's pro tip, narrowed to start-to-Done time; not run).
Source
- S1 · The 2020 Scrum Guide (Ken Schwaber and Jeff Sutherland), https://scrumguides.org/scrum-guide.html, the site's HTML port of the November 2020 PDF (
scrum-guide-2020.txt). "They are fixed length events of one month or less to create consistency." (step 1); "A new Sprint starts immediately after the conclusion of the previous Sprint." (step 1); "The Sprint Goal is the single objective for the Sprint." (step 2); "The Daily Scrum is a 15-minute event for the Developers of the Scrum Team." (step 3); "Only the Product Owner has the authority to cancel the Sprint." (step 5); "When a Sprint’s horizon is too long the Sprint Goal may become invalid, complexity may rise, and risk may increase." - S2 · The Kanban Guide, May 2025 (v2025.5) (John Coleman and Daniel Vacanti), https://kanbanguides.org/the-kanban-guide/2025.5/, reached from the row's kanbanguides.org through the site's English link (
kanban-guide.txt). "Any work items between a started point and a finished point are considered work in progress (WIP)." (step 8); "Kanban system members should start work on an item (pull or select) only when there is a clear signal that there is capacity to do so." (step 9); "Kanban system members should refrain from selecting more than the number of work items into a given part of the workflow beyond the WIP control." (step 9); "Throughput: The number of work items finished per unit of time." (step 10); "Work Item Age: The elapsed time between when a work item started and the current date." (step 10); "Cycle Time: The elapsed time between when a work item started and when a work item finished." (step 10); "The SLE itself has two parts: a period of elapsed time and a probability associated with that period" (step 11); "Ensuring work items do not age unnecessarily, using the SLE as a reference." (step 11).
Steps 13 and 14 are this lesson's model and the link to scale-littles-law-capacity; neither guide states them, and no cited key rests on them.
Related lessons
Checked against the tools and docs as of 2026-10-17.