Corey Ladas has written an interesting paper titled Scrum-ban in which he describes how a Scrum team might introduce the lean practice of kanban. He goes on to describe an evolutionary process, which if taken far enough, replaces most of Scrum. Even for those who don't want to scrap Scrum and go lean, the paper provides a useful view into what kanban is and how it can augment Scrum.
Kanban, as used in the context of Lean and Just-in-Time production, is a card-based signaling mechanism used to control inventory levels, and work-in-progress. As applied to software development, the cards are often index cards or sticky notes on a corkboard or whiteboard. The cards represent units of work, such as features or stories, and they move across the board to indicate their progress toward completion.
As the cards move across the board, they enter and leave various states. A card can only be in one state at any given time, and can only move to the next state when it meets the completion criteria for the current state, and when there is an open slot in the next state. Sometimes waiting states, or queues, are inserted between the 'active' states. If work starts piling up in one of these waiting states, it is an indication that a bottleneck exists. A hallmark of lean, just-in-time, systems is continuously identifying and eliminating such bottlenecks, toward the goal of maximizing the throughput of the system.
The paper draws an analogy between a kanban (card) and a currency note. The currency note is part of the money supply in an economy. As such, it represents a portion of the production capacity of the economy, and can be used to acquire actual goods. Purchasers have a limited supply of money, and must make decisions about which goods to buy and when. Similarly, a kanban (card) represents some portion of the production capacity of an internal economy. The business, or perhaps more specifically to Scrum, the product owner, has limited kanban to spend in any given time, and thus must make priority decisions about how to 'spend' them.
Corey's paper goes on to describe how Lean practices could modify or even replace many of the traditional Scrum practices, such as the daily standup and burn-down charts. Cory asserts that estimation is a form of waste and could largely be eliminated. He proposes a system by which the story authors are responsible for creating appropriately sized stories, but does not go on to suggest how they might do this.
Regardless of whether one agrees or takes issue with the conclusions of the paper, it presents Lean, and especially kanban, in a way that can help a Scrum team evaluate its usefulness. The article also outlines a practical way that a Scrum team might adopt the use of kanban to augment their existing processes.