New to this topic?
We recommend reading these guides first to get the most out of this one:
1958
Developed by DuPont
Longest
Path = Min Duration
Float
Measures Task Flexibility
0
Float on Critical Path

What Is the Critical Path Method?

The Critical Path Method (CPM) is a scheduling technique that identifies the longest chain of dependent activities in a project. That chain — the critical path — determines the shortest possible project duration. Every day saved on a critical-path task shortens the project. Every day saved on a non-critical task changes nothing.

CPM was developed in 1957-58 by DuPont and Remington Rand for managing plant maintenance shutdowns. It has since become the foundation of modern project scheduling, used in construction, manufacturing changeovers, new product introductions, and facility expansions.

The Fundamental Insight

If you have 100 tasks and 15 of them form the critical path, accelerating any of the other 85 tasks will not shorten your project by a single day. All project compression effort must focus on the critical path. This is why identifying it correctly matters so much.

Key Terminology

TermDefinitionWhy It Matters
ActivityA task that consumes time and resourcesThe building blocks of your schedule
PredecessorAn activity that must finish before another can startDefines the dependency logic
Network DiagramA graph showing all activities and their dependenciesThe visual model CPM operates on. This is a DAG.
Early Start (ES)Earliest an activity can begin (from forward pass)Tells you the soonest you can start a task
Early Finish (EF)ES + DurationEarliest an activity can complete
Late Start (LS)Latest an activity can begin without delaying the project (from backward pass)Tells you the deadline for starting
Late Finish (LF)Latest an activity can finish without delaying the projectYour hard deadline for the task
Total Float (Slack)LS – ES (or LF – EF)How much a task can slip without delaying the project. Zero float = critical path.
Free FloatTime a task can slip without delaying its immediate successorMore granular flexibility measure
Critical PathThe longest path through the network — all activities with zero total floatDetermines the minimum project duration

How CPM Works: The 5 Steps

Define Activities & DependenciesBreak the project into discrete tasks (WBS). For each task, identify which tasks must finish before it can start (predecessors). Estimate each task's duration using a single deterministic value (unlike PERT which uses three estimates).
Build the Network DiagramDraw the activities as nodes and dependencies as arrows. This creates a directed acyclic graph (DAG). Every path from start to finish represents a possible sequence of work.
Forward Pass — Calculate ES and EFStart at the first activity (ES = 0). For each activity: EF = ES + Duration. The ES of a successor is the maximum EF of all its predecessors. Work left to right through the entire network. The EF of the last activity is the minimum project duration.
Backward Pass — Calculate LS and LFStart at the last activity (LF = project duration). For each activity: LS = LF – Duration. The LF of a predecessor is the minimum LS of all its successors. Work right to left through the network.
Identify the Critical PathCalculate Total Float for each activity: Float = LS – ES. Every activity with zero float is on the critical path. The critical path is the longest path through the network and determines minimum project duration.

Worked Example

Consider a small manufacturing changeover project:

TaskDuration (days)Predecessors
A: Plan changeover2–
B: Order parts5A
C: Train operators3A
D: Install equipment4B
E: Update SOPs2C
F: Trial run1D, E
Path 1: A(2) → B(5) → D(4) → F(1) = 12 days ← Critical Path
Path 2: A(2) → C(3) → E(2) → F(1) = 8 days (Float = 4 days)
The critical path (A→B→D→F) is 12 days. Tasks C and E have 4 days of float — they can slip up to 4 days without delaying the project.

Forward & Backward Pass Detail

TaskDurationESEFLSLFFloatCritical?
A202020Yes
B527270Yes
C325694No
D47117110Yes
E2579114No
F1111211120Yes

Dependency Types

CPM supports four relationship types between activities:

TypeNotationMeaningExample
Finish-to-Start (FS)Most commonB cannot start until A finishesInstall equipment → then trial run
Start-to-Start (SS)Parallel startB cannot start until A startsExcavate → Lay pipe (starts 2 days after)
Finish-to-Finish (FF)Parallel finishB cannot finish until A finishesTesting → Documentation (must finish together)
Start-to-Finish (SF)RareB cannot finish until A startsNew system start → Old system shutdown

Leads and Lags

A lead accelerates a successor (e.g., FS – 2 days: start successor 2 days before predecessor finishes). A lag adds delay (e.g., FS + 3 days: wait 3 days after predecessor finishes). Use lags for curing time, drying time, or mandatory waiting periods. Avoid excessive leads — they often mask poor dependency logic.

What the four days are worth The float on path 2 and the crash budget for the whole project are the same four days. Crashing B and D shortens the project day for day until day four and nothing after it; crashing C and E — even deleting them outright — never moves it off 12.

Every duration and every date here is the page's own. A 2, B 5, C 3, D 4, E 2, F 1 days with the predecessors as listed come from the Worked Example table, and each bar is drawn at the early start and early finish printed in the Forward & Backward Pass Detail table: A 0–2, B 2–7, C 2–5, D 7–11, E 5–7, F 11–12. Path 1, A→B→D→F, is 12 days and critical; path 2, A→C→E→F, is 8 days with 4 days of float — the diagram box beneath the table, and the stat grid's "0 — Float on Critical Path". On the early schedule path 2 has finished its own work by day 7 and cannot start F until day 11, so the hatched stretch is 11 − 7 = 4 days wide; that is the same 4 the table gives as C's float, LS − ES = 6 − 2, and as E's, 9 − 5. It falls under D because D is the only work path 1 still has in that window. The right-hand panel adds one piece of arithmetic and nothing else: project duration = the longer of the two paths, recomputed after each day of compression, which is the page's own step 3 — "you must recalculate the network after each compression step" — carried out. B and D hold 9 days between them, so removing k of them gives max(12 − k, 8): 12, 11, 10, 9, 8, 8, 8 for k = 0 to 6. C and E hold 5, so removing k of those gives max(12, 8 − k): 12, 12, 12, 12, 12, 12 — 12 days at every step, including k = 5, where C and E have gone from the project entirely. The two paths become equal at k = 4, because 12 − 4 = 8, which is why the payoff stops there and why the bracket under the first 4 days is drawn at exactly the width of the float gap opposite — both panels use 28.5 pixels per day. The chart runs to k = 6 only to make the flat tail visible. No crash costs or crash limits are used, because the page gives none.

Crashing the Schedule

When the critical path is too long, you "crash" it by adding resources to shorten critical activities:

Identify crash candidatesOnly critical-path activities matter. Rank them by crash cost per day saved (lowest cost first).
Crash the cheapest firstAdd overtime, extra crew, expedited materials. Each activity has a maximum crash limit — you cannot compress it beyond its crash duration.
Recalculate after each crashCrashing one activity may create a new critical path. You must recalculate the network after each compression step to identify if the critical path has shifted.
Stop when cost exceeds benefitAt some point, the cost of crashing exceeds the value of finishing sooner. That is your optimal project duration.

CPM in Manufacturing Operations

ApplicationHow CPM Helps
Plant shutdowns / turnaroundsIdentify the sequence that determines total downtime. Crash only critical-path work to minimize lost production.
New line installationCoordinate equipment delivery, construction, electrical, commissioning. Find the path that governs go-live date.
New product introduction (NPI)Map design → tooling → validation → ramp. Identify which engineering delays actually push out launch.
Large changeoversWhen changeovers involve multiple teams and steps, CPM finds the fastest parallel sequence.
Facility expansionConstruction, permitting, equipment procurement — each has dependencies. CPM identifies the binding constraint on opening day.

CPM vs. PERT vs. Critical Chain

FeatureCPMPERTCritical Chain
Duration estimatesSingle (deterministic)Three-point (probabilistic)Aggressive + buffers
FocusTask dependenciesUncertainty in estimatesResource constraints + buffers
Float/bufferFloat per activityProbability distributionPooled project buffer
Resource constraintsNot built inNot built inCore consideration
Best forWell-understood, repeatable projectsR&D, novel projects with high uncertaintyMulti-project environments with shared resources
✅ CPM Best Practices
  • Use realistic durations based on historical data, not optimistic wishes
  • Update the network as actual durations come in — the critical path can shift
  • Focus management attention on zero-float tasks
  • Use a proper WBS to ensure no activities are missed
  • Revisit dependencies regularly — some assumed sequences can be parallelized
❌ Common Mistakes
  • Trying to accelerate non-critical tasks — it changes nothing
  • Using optimistic durations that make the plan look good but fail in execution
  • Ignoring resource constraints — CPM assumes unlimited resources
  • Setting the schedule once and never updating it
  • Confusing activity duration with effort (person-days ≠ calendar days)

CPM's Blind Spot: Resources

CPM calculates the critical path based purely on task dependencies and durations. It does not account for resource availability. If two critical-path tasks need the same electrician, one must wait — but CPM will not tell you that. For resource-constrained scheduling, see Critical Chain and Resource Leveling.

🎯 Key Takeaway

The Critical Path Method gives you the single most important insight in project scheduling: which tasks actually determine your project's duration. Focus your management attention, your resources, and your crash budget on critical-path activities. Everything else has float — and float is freedom. Master the forward and backward pass, track float religiously, and remember: when the critical path shifts (and it will), shift your attention with it.

Interactive Demo

Build a project network and identify the critical path. Adjust task durations to see how the critical path shifts.

⚡
Try It Yourself
Critical Path Method Simulator
▼
Adjust task durations to see how the critical path shifts. Red nodes and edges mark the critical path — any delay on these tasks delays the entire project.
A4dFloat: 0dB6dFloat: 3dC5dFloat: 0dD3dFloat: 4dE4dFloat: 0dF2dFloat: 4dG3dFloat: 0dH1dFloat: 0d
TaskNameDurESEFLSLFFloatCritical?
ADesign404040YES
BProcure Materials606393no
CBuild Frame549490YES
DElectrical3478114no
EInstall Parts49139130YES
FWiring27911134no
GTesting3131613160YES
HHandoff1161716170YES
4d
1d12d
6d
1d12d
5d
1d12d
3d
1d12d
4d
1d12d
2d
1d12d
3d
1d12d
1d
1d12d
17 days
Project Duration
5 / 8
Critical Tasks
11 days
Total Float
Ready for the full knowledge check? Test your understanding with guided scenarios and data export.
PROTake the Pro Knowledge Check →
Free forever · Every feature included

Stop reading, start modeling

Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.

Open Workbench →

Take this to a room

The running order

For anyone managing a project schedule. They should leave able to run a forward and backward pass and to say why accelerating a non-critical task changes nothing.

7 beats · 12 min
  1. 1

    One path decides the duration

    The longest path through the network is the project. Everything else has slack.

    • 100 tasks, 15 on the critical path.
    • Accelerating any of the other 85 shortens the project by exactly zero days.
    • All compression effort must go onto the critical path.

    Ask the room Do we know which of our tasks are on the critical path?

  2. 2

    Float is freedom

    Total float is how long a task can slip without moving the end date. Zero float means critical.

    • Float equals late start minus early start.
    • A task with four days of float can absorb four days of trouble.
    • Managing by float tells you where to stop worrying, which is as useful as knowing where to worry.
  3. 3

    Forward pass, backward pass

    Two sweeps through the network and you have everything.

    • Forward: early finish equals early start plus duration. A successor starts at the latest finish of its predecessors.
    • Backward: late start equals late finish minus duration. A predecessor's late finish is the earliest late start of its successors.
    • Float falls out of the difference.
  4. 4

    A worked network

    Two paths, and the numbers make the point without any theory.

    • A(2) to B(5) to D(4) to F(1) is 12 days - the critical path.
    • A(2) to C(3) to E(2) to F(1) is 8 days, so four days of float.
    • Crash C, E or anything on the second path and the project still takes 12 days.
  5. 5

    Crashing, in the right order

    There is a procedure, and the last step is the one people skip.

    • Only critical activities are candidates. Rank by crash cost per day saved.
    • Crash the cheapest first, up to its crash limit.
    • Recalculate after every compression - the critical path may have moved.
    • Stop when the cost of a day exceeds what a day is worth.
  6. 6

    CPM's blind spot

    It assumes unlimited resources, and that assumption breaks quietly.

    • Two critical tasks needing the same electrician: one waits, and CPM will not tell you.
    • The result is a schedule that is logically valid and physically impossible.
    • For that, you need resource levelling or critical chain.

    Ask the room Does our schedule assume anyone is in two places at once?

  7. 7

    Update it, or it lies

    A network set once and never revisited stops being true within weeks.

    • Update with actual durations as they come in.
    • The critical path shifts. Shift your attention with it.
    • Use realistic durations from history, not optimistic ones that make the plan look good.

    Ask the room When was our network last recalculated with actuals?