The idea in one paragraph
In any project where tasks depend on each other, some chains of work are longer than others. The critical path is the longest chain of dependent tasks from start to finish, and its total length IS the minimum duration of your project. Delay any task on that chain by a day and the whole project ends a day later. Delay a task that is not on it, and (up to a point) nothing happens to your end date.
A small example
Say you are launching an app. Design takes 5 days. After design, backend work takes 10 days and app screens take 6, and they run in parallel. Testing needs both and takes 4 days.
- Chain A: Design (5) then Backend (10) then Testing (4) = 19 days
- Chain B: Design (5) then App screens (6) then Testing (4) = 15 days
Chain A is the critical path: the project cannot finish in less than 19 working days. The app-screens task sits on the shorter chain, so it has 4 days of float (also called slack): it can slip up to 4 days before it matters. Backend has zero float, a one-day slip there is a one-day slip for everyone.
Why it changes how you manage
Without the critical path, every delay looks equally scary and every task looks equally urgent. With it, you know exactly where to spend attention:
- Protect critical tasks. Staff them with your steadiest people, clear their blockers first, and treat their estimates seriously.
- Relax about float. A non-critical task running late is a note, not an alarm, as long as its float holds.
- Compress where it counts. If the deadline must move up, shortening a critical task moves the end date; shortening anything else changes nothing.
How the critical path is found
The classic method (CPM, the critical path method) walks the schedule twice: a forward pass computes the earliest each task can start and finish, a backward pass computes the latest each can start and finish without delaying the end. Tasks where earliest equals latest have zero float, and together they form the critical path. Nobody does this by hand anymore; scheduling software recalculates it on every change, counting working days and dependency types with lag along the way.