gantt
On a construction job the question is rarely what the critical path is in theory. It is which of the forty things happening this month will actually move the handover date, and whether the answer changes once the long lead items are ordered. A schedule that shows every activity in the same weight of line answers none of that. The critical path method answers exactly that question, and on site it tends to produce an answer that surprises the people who assumed the crews were the constraint.
What follows is the method applied to a schedule rather than explained in the abstract: a worked example small enough to check, the four things that reliably break a construction critical path, and why float turns into a contractual argument rather than an arithmetic one.
The critical path is the longest chain of dependent activities through the schedule, measured in calendar time on the project's working calendar. Because it is the longest chain, its duration is the shortest the job can take. Every activity on it has zero float, so a day lost on any of them is a day lost on the handover date unless something else changes.
Three inputs produce it and nothing else does. A complete activity list, a duration for each, and the logic between them. The logic is the part that gets short-changed. Finish to start is the default relationship, but a construction schedule that uses only finish to start will be longer than the job actually is, because trades genuinely overlap. Start to start with a lag is what describes drywall following framing through a building by two floors, and a schedule that cannot express it forces the planner to either invent sequential activities or break the work into artificial zones.
The activity calendar matters as much as the durations. A 10 day activity on a five day working calendar spans two weeks, and one on a six day calendar spans less than that. Weather calendars, statutory holidays and shutdown periods all change the length of the chain without changing a single duration. This is why two planners can enter the same activities and durations and produce completion dates a fortnight apart.
Take an interior fit out of a small commercial unit. Durations are in working days, and each activity lists what has to finish before it can start.
| Activity | Duration | Cannot start until |
|---|---|---|
| A. Permit approval | 15 days | nothing |
| B. Demolition and strip out | 5 days | A |
| C. Order and deliver the HVAC units | 30 days | A |
| D. Framing | 8 days | B |
| E. Rough electrical and plumbing | 10 days | D |
| F. HVAC installation | 6 days | C and E |
| G. Rough inspection | 2 days | F |
| H. Drywall and finishes | 12 days | G |
| J. Flooring | 5 days | H |
| K. Final inspection and handover | 3 days | J |
There are two routes to the handover. Add them up.
The critical path is the second one, and it runs through a purchase order. The activity holding the handover date is a 30 day delivery window that nobody on site can work faster. The whole trade sequence from demolition through rough in carries 7 days of float, which means the crews could lose a week between them and the date would not move.
That result is typical rather than a contrived example. On fit outs and refurbishments the critical path very often runs through permits, long lead equipment or a utility connection, because those durations are set by somebody outside the contract. Two consequences follow. Pressuring the framing crew to pull back three days buys nothing, and the single highest value action available in the first week of the job is placing the HVAC order a week earlier, which takes a purchasing decision rather than any site effort.
The float figure is also the number to use when a trade asks for a week's extension. Seven days exists on that chain as a whole, not seven days for each activity on it. If demolition takes the week, framing and rough in have none left, and the trade sequence becomes a second critical path. From that point every further slip on either chain moves the handover.
The arithmetic is not the fragile part. These four conditions are where a schedule that was correct on paper stops describing the job.
A submittal, review, fabrication and delivery sequence is often entered as a single activity with an optimistic number because the planner is thinking about the fabrication time. Review cycles are waiting time and they belong in the duration. Splitting long lead items into submit, approve, fabricate and deliver, with the approval period shown as its own activity, changes which chain comes out longest on a surprising number of schedules.
A dependency should mean the successor physically cannot start until the predecessor finishes. Schedules accumulate arrows that actually mean the same crew is doing both jobs, or that this was the order somebody preferred. Those are resource constraints and sequencing preferences respectively, and recording them as hard logic produces a critical path that shifts every time a crew is reassigned.
A single block of float parked before the handover milestone hides where the exposure actually is. Weather risk sits on the activities that are outside the building envelope, and putting the allowance against those activities rather than at the end of the schedule keeps the critical path pointing at the work that is genuinely at risk.
The critical path is calculated from current durations and logic, so it changes as the job runs. A schedule updated monthly for a progress claim, with no interim status, means the path being discussed in week three is describing week one. The updating discipline, not the software, is what decides whether the critical path is information or decoration.
On a small internal project, float is arithmetic. On a contract it is money, because whoever gets to use the float is the party protected when something slips.
Contracts differ on this, and the answer is in the document rather than in the method. Some forms treat float as belonging to the project and available to whoever needs it first. Others make it the contractor's, on the reasoning that it was created by the contractor's own sequencing. Where an extension of time or a claim is in play, the specific wording of the contract and advice from the appropriate professional are what govern, not a general rule about scheduling practice.
What the method contributes is the evidence. Two things make a schedule usable in that conversation and both have to be in place before the dispute, not after.
The first is a baseline that was accepted, with the logic and durations recorded as they stood at acceptance. A critical path can only be shown to have moved if there is a version to compare it against. The second is a run of dated updates showing actual starts, actual finishes and remaining durations. An as-built comparison drawn from monthly updates is a much stronger position than one reconstructed from photographs and day sheets a year later.
This is also the reason construction schedules tend to be heavier than schedules elsewhere. A twelve person software team can keep a plan that is useful and slightly wrong. A main contractor cannot, because the schedule is a contractual record as well as a working tool.
The full critical path is a planning and reporting artefact. The version the job runs on is shorter.
A three week look ahead pulled from the schedule is where the critical path becomes operational. Activities on the critical path in the next three weeks get their constraints checked first: material on site, drawings released, permits in hand, preceding inspection passed. That check is the whole point. Everything that stops a crew starting on a zero float activity is a date slip, and almost all of it is visible a fortnight ahead if somebody looks.
The weekly work plan then commits to what will actually be done, and the useful discipline is that commitments on zero float activities are treated differently from commitments with a week of slack behind them. A trade declining to commit to a critical activity is a scheduling event, not a conversation to have later. A trade declining to commit to something with 7 days of float is a note.
Two reports do most of the work in a weekly meeting: the activities with zero float in the look ahead window, and the activities whose float has dropped since the last update. The second one catches the near critical chains before they tip over, which is where most recoverable delays are found.
Scheduling tools split into three groups, and the distinction worth understanding is what each one is actually charging for. Prices checked on 26 September 2026.
| Tool | Entry point relevant to CPM | Published price |
|---|---|---|
| GanttPRO | Core plan, with auto scheduling | $7 per user per month, billed annually |
| TeamGantt | Base plan, including critical path scheduling | $59 per month flat, 1 manager and 10 collaborators |
| TeamGantt Builder Edition | Construction specific plan, unlimited tasks per project | From $199 per month, 10 projects included |
| Procore | Construction platform, priced per organisation | Custom quote, with unlimited users stated on the pricing page |
The pricing shape matters more than the headline number. A per user price rewards a small planning team and penalises putting trade partners in the system. A flat or per project price does the opposite, which is why the entry plans above and a per seat tool produce completely different totals for a job with a dozen subcontractors who all need read access. Procore does not publish a per seat figure at all and states unlimited users on an annual contract, so the comparison there has to be done against a quote.
For heavy programme work with claims exposure, dedicated CPM software is not really optional. For a fit out contractor running four jobs where the schedule mostly needs to be current and visible to the people doing the work, the binding constraint is usually whether anyone updates it, not whether float is computed to two decimals. The comparison pages set out which tools include a timeline view from the start and which put it behind a paid tier, and the feature list covers keeping a Gantt view and the crews' own task board on the same cards rather than as two copies of the plan that drift apart.
Rebuild the logic for the job currently running, with long lead procurement broken into submit, approve, fabricate and deliver as separate activities, and see whether the critical path still runs where the last schedule said it did. If it moves onto a purchase order, the highest value decision available this week is a purchasing decision rather than a site one. Where the real gap is that nobody can see the current schedule without asking the planner for a PDF, put the timeline on the same board the trades already update, which is what the Gantt view in Pinateca is for.
Because delivery and approval durations are usually longer than any single trade activity, and nobody on site can compress them. A 30 day equipment lead time with a 15 day permit period ahead of it is a 45 day chain before a crew has done anything. This is normal on fit outs and refurbishments, and it means the earliest purchasing decisions are worth more to the date than any amount of site pressure.
Weekly status against a three week look ahead is the common working rhythm, with a formal monthly update tied to the progress claim. Monthly alone is too slow, because the path being discussed will describe conditions from three or four weeks earlier. The test is simple: if the float figures cannot be trusted at a Monday meeting, they are not being updated often enough.
That depends on the contract rather than on scheduling practice. Some standard forms treat float as a project resource available to whichever party needs it first, and others treat it as belonging to the contractor who created it through their own sequencing. Where an extension of time claim is involved, the contract wording and advice from the appropriate professional are what decide, so the safe step is to check the document before relying on either assumption.
A Gantt chart that records dependencies and calculates float is doing critical path scheduling, whatever it is called. What dedicated CPM tools add is baseline comparison, resource levelling, calendar handling for weather and shifts, and the reporting formats used in claims. If the job carries contractual milestones and delay exposure, those features are the reason to pay. If the main problem is that nobody can see the current plan, a simpler shared timeline solves more.
Yes, and it is common late in a job. Any two chains with the same total duration both have zero float and both control the handover date. It also happens when a chain with float uses it up, which is why activities whose float has fallen since the last update are worth reporting separately. Watching only the zero float list means finding out about the second path after it has already slipped.