Critical Path Method

Replace a date list with a logical schedule: calculate which dependent activities currently determine the earliest achievable finish and where flexibility exists.

In one minute

Critical Path Method (CPM) models project activities, durations and dependencies as a network. A forward pass calculates earliest start and finish. A backward pass calculates latest start and finish without delaying the required project finish. Total float is the scheduling flexibility between those dates. The longest duration path through the network normally determines the minimum calculated project duration; activities on it have zero total float under the current assumptions.

The critical path is not “the most important work,” a permanent label or a resource plan. It can change when actual dates, logic, constraints or durations change.

Best for: a defined project with estimable activities and meaningful dependencies.
Avoid when: work cannot yet be decomposed, durations are highly uncertain without ranges, or scarce resources make the logic-only schedule infeasible.

The problem it addresses

Milestone plans often hide sequence. Every task appears urgent, buffers are consumed invisibly and a delay is discussed without tracing its effect on the finish date. CPM makes schedule logic inspectable and supports targeted monitoring and controlled compression.

Its result is conditional on network completeness, duration assumptions, calendars and constraints. A mathematically correct schedule can still be operationally impossible.

When to use it

Use CPM when:

• a project has a defined completion milestone;
• activities have estimable durations and finish-to-start or other explicit logic;
• parallel work and merge points matter;
• the team needs float, forecast impact or schedule scenarios;
• actual progress will update remaining duration and logic;
• resource, calendar and governance constraints can be checked separately.

When not to use it

Do not use it:

• as a prioritisation label for “important” tasks;
• before scope and deliverables are sufficiently defined;
• with arbitrary dependencies added only to fit a desired date;
• as proof that a deterministic duration will occur;
• instead of resource-capacity planning;
• instead of Theory of Constraints when the main problem is a recurring system bottleneck;
• to compress safety, quality or regulatory gates without authority.

Inputs required

Prepare:

• completion criteria, milestones and project calendars;
• activities expressed as observable work packages;
• duration estimates, basis and uncertainty;
• logical predecessors and dependency types;
• external, contractual and control constraints;
• available resources and known contention;
• status date, actual starts/finishes and remaining durations for updates.

Step-by-step process

1. Define completion and calendars

Specify what “finished” means, the scheduling unit, working calendars and imposed constraints. Record which date is an external requirement and which is a calculated forecast.

2. Decompose deliverables into activities

Create activities small enough to estimate and status but large enough to manage. Include validation, hand-offs and approvals rather than only production work.

3. Estimate durations transparently

Record the basis, assumptions, range or confidence and calendar for each duration. Separate effort from elapsed duration.

4. Define dependency logic

Connect genuine technical, physical, informational or control dependencies. Prefer explicit logic over unexplained date constraints. Challenge unnecessary serialisation.

5. Draw and validate the network

Ensure every non-start activity has a predecessor, every non-finish activity reaches completion and there are no loops. Walk the logic with the people doing the work.

6. Run the forward pass

For finish-to-start logic: earliest start equals the maximum earliest finish of predecessors; earliest finish equals earliest start plus duration. At merge points, the latest predecessor governs.

7. Run the backward pass

Starting from the project finish, calculate latest finishes and starts. At branch points, use the minimum latest start of successors so no downstream activity is delayed.

8. Calculate float and identify paths

Total float equals latest start minus earliest start, or latest finish minus earliest finish. Trace continuous zero-float activities to identify the current critical path. Check near-critical paths with little float.

9. Test feasibility and scenarios

Apply resource, calendar and control constraints. Model delay, overlap, resequencing, crashing or scope change. Record cost, risk and assumption effects; compression is a decision, not a formula.

10. Update from actual progress

At each status date, record actuals, remaining duration, changed logic and approved scope changes. Recalculate all paths. Do not preserve an obsolete baseline forecast as the current schedule.

AI automation lens

AI can extract candidate activities and dependencies from approved plans, flag open ends and circular logic, run schedule scenarios and explain changed paths with source references.

It must not:

• invent dependencies or durations to meet a requested date;
• confuse effort with elapsed time;
• hide resource over-allocation or calendar differences;
• remove regulatory, safety or quality gates;
• treat zero float as certainty of failure;
• commit to a recovery plan without responsible owners.

Human project owners validate scope, logic, estimates, resources and the acceptable cost and risk of compression.

Visual model

Text alternative: activity A splits into A–B–D–Finish and A–C–E–Finish. Their durations are calculated separately; the longer path currently governs completion while the shorter path has float.

Calculation example

If the day-zero convention is used, A–B–D lasts 8 days and A–C–E lasts 10 days. The second path is critical and the first has 2 days of total path float. State the chosen date convention because software may display day numbers differently even when elapsed duration agrees.

Interactive example

Scenario

A data migration includes schema mapping (3 days), extraction after mapping (4), interface build after mapping (6), cleansing after extraction (5), integration test after interface and cleansing (3), user acceptance after integration test (4), and cutover after acceptance (1).

Your move

Find the schedule-driving path and assess a two-day delay in cleansing.

Worked answer

The extraction and cleansing branch lasts 3 + 4 + 5 = 12 days before integration test. The interface branch lasts 3 + 6 = 9 days, so it reaches the merge three days earlier. The full critical path is mapping → extraction → cleansing → integration test → user acceptance → cutover, lasting 20 days. A two-day cleansing delay moves calculated completion to 22 days unless another approved response changes logic or remaining duration. Interface work initially has three days of float, subject to calendars and resources.

Facilitation notes

• Draw logic before discussing the desired finish date.
• Ask what tangible output enables each successor.
• Record duration basis and calendar beside the estimate.
• Distinguish a hard external constraint from management preference.
• Track near-critical paths and shared resources.
• Recalculate after real progress; the path can migrate.

Expected output

• a validated activity network;
• duration assumptions and calendars;
• earliest and latest dates;
• total float and current critical and near-critical paths;
• resource and control feasibility notes;
• scenarios, decisions, baseline and update cadence.

Common mistakes

1. Important equals critical. Critical is a schedule calculation under current logic.
2. Missing work. Reviews, integration and approvals vanish from the network.
3. Desired-date logic. Constraints are invented to make a plan look compliant.
4. Duration as certainty. One point estimate hides meaningful uncertainty.
5. Ignoring resources. Parallel tasks may require the same unavailable person.
6. Frozen path. Actual progress can create a different critical path.

Quality checklist

• Completion criteria, status date and calendars are explicit.
• Activities produce observable outputs.
• Dependencies have a defensible mechanism.
• Network open ends, loops and merge points are checked.
• Durations preserve basis and uncertainty.
• Forward and backward calculations are reproducible.
• Resource and mandatory-control feasibility is tested.
• Actual progress triggers full recalculation.

Template

Use the structured CPM workspace to preserve network logic, calculations, assumptions and update decisions.

Knowledge check

Question: A critical activity finishes two days early, but a different near-critical path is delayed by four days. What should the project manager do?

A. Keep the original critical path unchanged.
B. Recalculate the complete network using actual and remaining information.
C. Add two days to every activity.
D. Remove the delayed path.

Answer: B. Criticality is dynamic and belongs to the current network, not the original label.

Related tools

• Process Decision Program Chart prepares responses to uncertain failure branches.
• Theory of Constraints improves a recurring system around its limiting constraint.
• RACI Matrix validates ownership and authority for schedule activities and gates.
• FMEA examines failure modes when schedule compression changes process risk.

References

1. Project Management Institute. “Critical Path Method Calculations.” PMI learning resource (opens in a new tab). Accessed 7 September 2026.
2. NASA. “6.1 Technical Planning.” NASA Systems Engineering Handbook reference. NASA (opens in a new tab). Accessed 7 September 2026.
3. Kelley, James E., and Morgan R. Walker. “Critical-Path Planning and Scheduling.” Proceedings of the Eastern Joint Computer Conference, 1959. Historical primary account; modern software conventions vary.

Method profile

• Primary output: calculated and updateable schedule network.
• Decision level: project or programme.
• Evidence strength: high when logic, calendars, estimates and actuals are governed.
• Review trigger: every status cycle, approved scope or logic change, or material variance.