Shared specialist
Two teams both need the same database expert. The board shows the migrations as parallel. The chain sequences them and puts a feeding buffer on the second, so the expert is no longer the invisible reason every date slips.
Topic 03 · Critical Chain
The date is not a task.
Roadmaps slip even when every estimate looked generous. The safety was hidden inside the tasks, then spent on late starts, work that grew to fill the time, and people switched across too many initiatives. Critical chain pulls that safety out and puts it where it protects the date.
01
Why the date moves
The estimate was not too short. The safety inside it was used up before the surprise arrived.
Student syndrome
People start late because they know the estimate has padding.
Parkinson’s Law
Work expands to fill the days you gave it. A five-day task rarely comes back in three.
Bad multitasking
Switching initiatives costs the focus the estimate assumed you still had.
A person, not a bar
The critical path pretends the senior engineer, the database expert, or the product owner can be in two meetings at once.
1
Wait
2
Wait
3
Work
4
Work
5
Work
6
Work
7
Work
The estimate was five days, with two days of safety buried inside it. You finish on day 7. The date is missed. The padding was spent before the work started, so the two-day surprise had nowhere to land.
02
Chain, not path
The critical path is the longest sequence of tasks. The critical chain is that sequence once the same person cannot do two of them at once.
API4 days
Migration5 days · Sam
Launch3 days
Permissions, “in parallel”4 days · Sam
Critical path: 12 days, if Sam can review the migration and the permissions change in the same week. They cannot.
03
Where the safety goes
Estimate the focused effort, roughly the time you would bet on half the time. Put the protection at the end of the chain, and where side work joins it.
FeedingDocsFeeding buffer
ChainMigrateAPILaunchProject buffer
Migration8 days
API10 days
Launch6 days
Each task is padded to feel safe. The plan says day 24, but the safety is spent on late starts and on work that grows to fill the estimate. Nothing is left when the chain slips.
Project buffer
At the end of the critical chain. This is what protects the delivery date.
Feeding buffer
Where a side path merges into the chain, so that side work does not steal the project buffer.
Resource buffer
A warning so the scarce person is ready when the chain needs them. It is not extra days.
04
Watch the buffer, not every ticket
Once the plan is set, stop asking whether each task hit its original date. Ask whether the buffer is still healthy for how far the chain has moved.
Chain complete →↑ Buffer consumed
Healthy. Chain 60%. Buffer 30%. Buffer consumption is behind progress. Continue. Do not re-plan every ticket.
05
More than one initiative
The usual organisation does not have one project. It has a portfolio, and one scarce person inside all of them. Stagger the starts.
- Billingday 24
- Searchday 24
- Mobileday 24
Every initiative looks active. None finishes before day 24, and each switch spends the focus the estimates assumed you still had. The portfolio is late because it was all started.
A platform rewrite
Tasks were padded so the six-month plan felt safe. Cut them to focused effort, put a project buffer at the end, and stop pulling the team onto three other urgent features. Review the buffer each week. Do not re-plan every slip.
A multi-team launch
The chain runs through integration and launch, which depend on a few people. Feeding buffers protect those merges. Other squads do not start new feature work until that chain has capacity.
06
One initiative, one buffer
Pick a date that matters. Find the people who are on the chain twice. Move the safety into the open.
- One initiative that is late or unpredictable.
- The tasks, and who is needed on more than one of them.
- The critical chain, not the path that assumed infinite people.
- Focused estimates. A project buffer. Feeding buffers at the merges.
- No new work that competes for those same people.
- Buffer health once a week. Act when it is watch or recover.
- Across the portfolio, fewer initiatives in progress at once.
This week
0/6
Cheat sheet
- Why dates slip
- Student syndrome, Parkinson’s Law, and bad multitasking. Not estimates that were simply too short.
- Critical chain
- The longest dependent path once people are finite. Often longer than the critical path.
- Estimates
- About half the confidence. The time with focused effort and no major surprise.
- Buffers
- Project buffer at the end. Feeding buffers at the merges. Resource buffers are a warning, not more time.
- Fever chart
- Chain complete against buffer consumed. Manage that, not every task date.
- Portfolio
- Stagger starts. Limit how many initiatives are live, or the scarce person becomes the bottleneck for all of them.
Check yourself
5 questions. No timer.