Critical path method (CPM)

CPM is a model of sequence, not a promise

The critical path method turns a plan into a network calculation. Activities describe the work, durations describe how long the work is expected to take, calendars define when work may occur, and relationships describe which activities drive others. The schedule engine combines those inputs to calculate early dates, late dates, float, and one or more paths to a completion milestone.

That calculated finish is conditional: it is true only if the network represents the intended means and methods, the calendars are correct, the remaining durations are credible, and future work follows the encoded relationships. CPM can calculate a poor plan with perfect arithmetic. Professional review therefore asks two questions separately: did the software calculate the network correctly, and does the network describe a buildable plan?

• Activities should represent measurable work or decision points.
• Relationships should describe real handoffs rather than merely eliminate open ends.
• Calendars should match the work pattern, holidays, shifts, and weather assumptions actually intended.
• Constraints should represent genuine external or contractual boundaries, not force a preferred answer.

The forward pass, backward pass, and float

The forward pass starts at the beginning of the network and calculates the earliest each activity can start and finish. A successor cannot begin until its driving predecessor relationship permits it, interpreted through both activities’ calendars and any applicable lag. The latest early finish at the completion point becomes the network’s earliest calculated completion.

The backward pass works from the required or calculated completion back through the network. It calculates the latest dates that preserve the controlling completion condition. The difference between early and late dates is float. Zero or negative float can identify time-sensitive work, but float values are outputs of the whole model; a changed constraint, calendar, relationship, or project finish condition can alter them without any physical work changing.

What “critical” means in Primavera P6

P6 supports two principal ways to identify critical activities. The project may flag activities whose total float is at or below a configured threshold, or it may identify the longest path: the driving chain that determines the selected completion. Those definitions can produce different activity sets. A three-day float threshold can mark several paths critical; longest path can isolate the controlling chain even when constraints produce unusual float.

For that reason, “the red activities” are not self-explanatory. A reviewer should record the scheduling options used, the critical-float threshold, the milestone at which paths terminate, whether multiple float paths were calculated, and whether the project was scheduled alone or with external relationships. Without those settings, two reviewers can open the same file and report different critical work while both are faithfully reading their own configuration.

Practical rule: name both the method and the endpoint—such as “longest path to contractual substantial completion”—instead of reporting an unexplained critical-activity count.

Calendars, constraints, and lags can change the story

A “day” is not a universal unit inside a multi-calendar schedule. Five days on a seven-day calendar and five days on a five-day calendar end on different dates, and a relationship between them must be translated through the scheduling engine’s calendar rules. Review float in hours when precision matters, and verify the display conversion before comparing two paths that use different calendars.

Constraints and lags deserve similar attention because they can interrupt otherwise transparent logic. A legitimate permit date may be modeled as an external milestone; an unexplained mandatory finish can hold an activity on a date regardless of predecessor performance. A long lag may represent curing, review, or delivery, but a named activity is usually easier to status, assign, and audit. The issue is not that constraints or lags are forbidden—it is that their effect must be visible and justified.

The longest path can move without the project finish moving

Projects rarely have one permanently critical chain. Parallel design packages, procurement, permitting, utility work, testing, and turnover paths compete as their float is consumed. One path can lose ten days of float while another gains five through revised logic, leaving the reported completion unchanged. The unchanged date does not mean the period was healthy; it means the network still had enough flexibility to absorb the movement.

A useful update review therefore compares the current longest path, the prior longest path, and several near-critical paths. Note which activities entered or left each path and why. A path change supported by actual progress or an approved resequence may be entirely credible. A path change caused by deleted logic, a new constraint, or unexplained remaining-duration reductions is a schedule change that requires an answer before the finish date can be trusted.

How to review a reported critical path

Start at the contractual completion milestone and trace driving predecessors backward. Confirm the chain reaches a valid project start or external interface without a break. Then move forward through the same chain and ask whether each handoff is physically and contractually plausible. Review actual dates, remaining durations, calendars, constraints, lags, out-of-sequence handling, and external relationships along the path.

Finally, reconcile the path to field reality and the narrative. If the schedule says permanent power is controlling, the update should identify the current design, procurement, installation, inspection, and energization facts that make it controlling. If the narrative discusses façade access while the calculated path runs through commissioning, either the narrative or the network is incomplete. The purpose of review is not to find a red bar; it is to explain the causal chain behind the forecast.

Common failure modes

The most common mechanical failure is a broken chain: missing logic, a constraint-driven date, or an external relationship that is absent from the review file. The most common interpretive failure is equating low float with importance. Safety, long-lead procurement, owner decisions, and high-cost work may deserve management attention even while they carry float.

Other warning signs include a critical path dominated by level-of-effort activities, long unexplained lags, actual work beyond the data date, negative float created solely by an obsolete constraint, and a critical path that jumps to a different work area every month without a corresponding change narrative. None proves manipulation on its own. Each is a prompt to trace the inputs and require a documented explanation.