AWS Builder Center

Stop Starting More Work: Limit WIP to Shorten Cycle Time

Adding another item to the board does not make a team deliver faster; it makes the work wait. This article argues that the defensible reason to limit work in progress is queueing logic, not a vague claim that multitasking destroys hours, and that once the cognitive evidence is read honestly, the strongest case for WIP limits is that they shorten and steady cycle time without anyone working faster.

Series: developer-tools (8 articles)

  1. …
  2. 7
    Stop Starting More Work: Limit WIP to Shorten Cycle Time This article
Decision point
Rule
Why
Limit per stage
Set a cap on items in each workflow stage, not just a total
Blocked handoffs hide inside a single total number
Start versus finish
New work starts only when capacity opens downstream
Finishing items is what lowers cycle time
Blocked items
Flag and unblock before pulling more work
A blocked item is inventory that still occupies the queue
Reviewer capacity
Count review and test as constrained stages
A coding stage that outruns review builds a hidden pile
Measurement
Track cycle time, throughput, and WIP together over weeks
Little's Law only holds for stable averages
Adjustment
Move limits based on observed queues, not discomfort
Tightening a limit is a change, test it against the metrics
  1. Running it without the mythology
Put the board where the queue is visible and write the limit on each stage, then enforce it by stopping new pulls rather than by asking people to hurry. Expect the first effect to be visible idle or blocked capacity at the upstream stage; that is the signal that work was previously being pushed into a system that could not absorb it, and it is where the team should attack the real constraint, often review or a specialist handoff. Keep blocked work explicit and time-boxed so it does not rot quietly. Measure cycle time and throughput for several weeks before and after, and resist quoting a target percentage from any case study, including the BBC one. Watch strain as well as flow, because the interruption research says the cost of overload is carried by people even when the clock looks fine.


Next steps: pull the current board and count every item started but not finished, then set a per-stage WIP cap below that count and stop starting work when a stage is full. Unblock the items already in the system before adding new ones. Record cycle time, throughput, and WIP over the next few weeks and adjust the limits against the observed queues rather than anyone's discomfort. If you have been quoting a 23-minute switching cost, drop it and make the case on flow and strain instead.

References

  1. "Lean Software Management: BBC Worldwide Case Study" (single-team case, 2009; 37% lead-time improvement and related figures; no control group) —  http://core.ac.uk/download/pdf/10062417.pdf 
  1. Agile Alliance, "Actionable Metrics" (cycle time = WIP / throughput; definitions and assumptions) —  https://www.agilealliance.org/resources/experience-reports/actionable-metrics-siemens-health-services 
  1. Rubinstein, Meyer, and Evans, "Executive Control of Cognitive Processes in Task Switching" (four experiments; millisecond-scale switching costs with reported p values) —  http://umich.edu/~bcalab/documents/RubinsteinMeyerEvans2001.pdf 
  1. Mark, Gudith, and Klocke, "The Cost of Interrupted Work: More Speed and Stress," CHI 2008 (interrupted tasks finished faster, with higher strain) —  http://www.ics.uci.edu/~gmark/chi08-mark.pdf 
  1. Analysis of the "23 minutes to refocus" claim, tracing the figure to the observational 2005 study (24 workers, 700+ hours; analytical secondary source) —  https://worldtimedata.com/productivity/what-the-research-actually-says-about-the-23-minutes-to-refocus-claim/ 
  1. American Psychological Association, "Multitasking: Switching Costs" (authoritative summary of the switching research) —  https://www.apa.org/research/action/multitask.aspx 

Series: developer-tools (8 articles)

  1. …
  2. 7
    Stop Starting More Work: Limit WIP to Shorten Cycle Time This article
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article