liveBook · Manning

Wip limit

As explained in Build High Value Software Systems

A rule that caps the number of work items that may be in progress at once.

Starting another task can feel productive when a pull request waits for review or an integration is blocked. But unfinished features do not deliver value in production, and more open work gives queues and blockers room to grow. A work-in-progress (WIP) limit caps the number of items active at once. When the limit is reached, the next move is to finish or unblock work—not pull in another item. section 6.3

Two active tasks, then stop

Suppose an authentication refactor is blocked, so you start a CSV export. With a personal limit of two, you finish and merge the export before taking anything else. If the refactor remains blocked, you pursue the security review rather than start a search bug. Once the refactor is merged, you can take the bug with full attention. The limit turns a blocker into something to resolve instead of something to work around indefinitely. section 6.3

Apply the ceiling where work accumulates:

  • For a team, count active work across developers. If an urgent request fills the remaining capacity, stop pulling new work and help clear existing items. section 6.3
  • For a review queue, cap waiting pull requests. When the queue is full, help with reviews or slow intake; producing more small PRs does not remove the inventory. section 6.3 section 6.2
  • For yourself, count more than tickets: promised reviews, mentoring, learning, and design work also compete for attention. Put them on a visible board or list and plan ticket capacity around them. section 14.4 section 14.4.1

A one-ticket personal limit is stricter than the two-task example: actively pursue a blocker, or return the ticket to the backlog before pulling another. A genuinely independent task that can be finished quickly is another option. These are choices for managing attention, not a universal number. Make non-ticket commitments visible to the team so planned ticket capacity reflects the time actually available. section 14.4.1 section 6.3

Table 14.3 WIP impact on weekly throughput
Metric
Developer A: No WIP limit
Developer B: Ticket limit only
Developer C: Holistic WIP
Tickets in progress
4
1
1
Non-ticket commitments
Untracked, scattered
Untracked, scattered
Visible, planned into the week
Planned deep work hours
40 (assumes full week)
40 (assumes full week)
24 (accounts for 16h of meetings, reviews, mentoring)
Actual deep work hours
~15 (rest lost to switching)
~25 (less switching, but meetings still fragment)
~22 (fewer hours planned, more of them protected)
Context switches/week
12-15
4-6
0-2
Tickets completed
0-1
2-3
3-4
Non-ticket quality
Shallow, rushed between ticket switches
Attended but unplanned, displaces deep work unpredictably
Prepared, scheduled, and given real attention

The weekly comparison contrasts no limit, a ticket-only limit, and a holistic limit using planned deep-work hours, context switches, completed tickets, and non-ticket quality. section 14.4.1

Set the limit at a real constraint, such as a backed-up review queue, then watch whether it prompts completion and shorter waits. A WIP limit works as an intake rule: capacity to finish work determines when new work starts. section 6.1.2 section 6.3

How Build High Value Software Systems covers it

How deep it goes
Works through individual, team, and review-queue decisions without prescribing one universal numeric limit.
Examples
Uses a blocked authentication refactor, a CSV export, and a weekly comparison of ticket-only versus holistic limits.
What it assumes
Assumes familiarity with tickets, pull requests, and sprint boards while showing how to count less-visible commitments.
What you can do after
Apply it via a Kanban board to limit active work, resolve blockers, and plan around non-ticket commitments.
How it connects
Links Kanban pull systems and review queues with context switching and visible personal commitments.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build High Value Software Systems. The section links point to the parts of the book it draws on.