Both Scrum and Kanban come out of the Agile family, and both try to stop teams drowning in half-finished work. But they start from different questions. Scrum asks: what can we finish in the next two weeks? Kanban asks: what is stuck right now, and are we taking on too much? My argument in this essay is that most teams pick between them out of habit rather than fit, and that the fit depends mostly on one property of the work.

What Scrum prescribes

Scrum, as described in the Scrum Guide, is a small framework built on time-boxed iterations called sprints, usually one to four weeks long. It names three accountabilities: the Product Owner, who decides what is worth building and in what order; the Developers, who do the work; and the Scrum Master, who helps the team follow the process and removes obstacles.

The rhythm has fixed events. Sprint Planning chooses what to attempt and how. Short daily meetings (the Daily Scrum) let the team adjust. A Sprint Review shows the working increment to stakeholders. A Retrospective examines how the team worked and what to change. Work enters the sprint at planning, and the team protects it from new requests until the sprint ends. That protection is the main thing Scrum offers: a predictable cadence, and room to think without interruption.

What Kanban prescribes

Kanban has no sprints, no required roles and no required meetings. Its core is a visible board with columns for each stage of your workflow, say Backlog, Analysis, Development, Testing, Done, plus a limit on how many items can sit in each stage at once. These are work-in-progress (WIP) limits.

Work flows continuously. When an item leaves Testing, a slot opens, and someone pulls the next item in. If Testing is full at its limit of three, developers cannot push a fourth item there; they must help clear the bottleneck first. This is the lesson Kanban teaches well: starting more work does not finish more work. The measures that matter are cycle time (how long one item takes from start to finish) and throughput (how many finish per week).

A small numerical picture

A useful rule, Little’s Law, says that on average, work in progress equals throughput multiplied by cycle time. If a team finishes 5 items a week and typically has 20 items in progress, the average item takes 4 weeks. Cut the work in progress to 10 without changing the finishing rate, and items take about 2 weeks. Nothing got faster in effort; things simply stopped waiting in queues. That is the whole logic behind WIP limits.

Where each one fits

Consider two imaginary teams. The first builds a new insurance-claims portal over eight months. The work is planned in chunks, stakeholders want regular demos, and the team benefits from agreeing a goal every fortnight. Scrum suits it.

The second is a four-person support-engineering team that handles production bugs and small requests for a bank’s internal systems. Items arrive unpredictably, some need to be done the same day, and a two-week sprint plan would be broken by Tuesday. Forcing Scrum on them means either ignoring the sprint boundary constantly or rejecting urgent work, and both options are painful. Kanban fits, with perhaps an expedite lane for genuine emergencies.

So the property that decides is how interruptible and unpredictable the work is. Stable, plannable, product-style work leans Scrum. Interrupt-driven, service-style work leans Kanban.

Where each goes wrong

Scrum in practice often decays into mini-waterfall with ceremonies. Sprint commitments become targets that management quietly reads as promises, and velocity (story points per sprint) gets reported upward as a productivity score, which makes teams inflate estimates. Velocity was meant as a planning aid for the team itself, not a ranking tool. A retrospective that produces the same complaints every month and no change is a sign the practice has become theatre.

Kanban fails differently. Without a time-box there is no natural moment to reflect, so boards can become permanent dumping grounds with WIP limits that everyone ignores. A board whose limits are never enforced is just a to-do list with columns. Teams also sometimes skip the improvement habit, because nothing forces a review.

My side

If I had to give one piece of advice, it would be to start with the board and WIP limits, whichever framework you pick. The limit on in-progress work is the single most effective idea in either approach, and Scrum teams benefit from it too, since a sprint with eight stories all “in progress” on the last day is the classic failure. Then add Scrum’s cadence if your work has a plannable shape and your stakeholders need a regular rhythm. Many teams end up with a hybrid, often called Scrumban, which is sensible rather than impure.

What I would not do is adopt a framework because a client’s contract demands “Agile”, or because the whole company is doing Scrum. Look at your last two months of work. If most of it could have been planned two weeks ahead, use sprints. If most of it arrived on the day, do not.

Figures and names in the examples are illustrative.

Leave a Reply

Trending