Construction scheduling field guide: CPM, look-ahead, and the critical path

What construction scheduling is, and how it finds a delay early

Construction scheduling is planning the sequence and timing of the work so the crew, the materials, and the other trades line up, then tracking the actual work against that plan so a slip shows up while you can still do something about it. The schedule is how you find a delay before it costs you. A pour that slides a week is a problem you can recover. The same week discovered at substantial completion is a liquidated-damages letter.

Three pieces do most of the work. The critical path is the chain of dependent activities that sets the finish date, so a slip there moves the end. The look-ahead is the three to six week window the field actually builds from, with constraints cleared and crews assigned. The approved baseline plus regular updates is the record that lets you measure where you really are and prove a delay when the cause was not yours.

Scheduling sits next to two other disciplines this guide does not repeat. What the time and material actually cost as the job runs belongs in the job-costing guide. Getting paid for work the contract never covered, including the time a change adds, belongs in the change-order guide. Scheduling is the time side: when each piece happens, in what order, and how you keep the order from coming apart.

What the schedule actually buys you

A real schedule does six things, and a bar chart taped to the trailer wall does maybe one of them. It coordinates the trades so the electrician is not roughing in the wall the drywaller is closing. It sequences material so the rebar lands before the forms are set, not after. It surfaces a delay early, while there is still float to absorb it or a re-sequence that saves the date.

The other three are about protecting your position. The schedule protects the finish date, because you cannot defend a completion date you never planned to. It supports a delay claim, because an approved baseline and honest updates are the evidence that the delay landed on the critical path and was not your fault. And it drives cash flow, because the schedule of values and the draw track the work in the order the schedule lays out.

The job that runs without a schedule is not running without one. It is running on the super's memory, which works until the super is out sick on the week three deliveries collide. Put the sequence on paper and the collision shows up in the look-ahead instead of in the field.

Schedule types and what each one is for

There are four kinds of schedule you will meet, and they answer different questions. The bar chart, or Gantt, shows activities as bars on a calendar. It reads at a glance and the owner loves it, but a pure bar chart has no logic underneath, so it cannot tell you what moves the finish when one bar slips. Use it as the dashboard, not the engine.

The CPM network is the engine. Activities are linked by relationships, the software calculates the longest path, and you get the critical path, the float, and the dates that fall out of the logic. Most contract schedule specifications that ask for a real schedule are asking for CPM. Line of balance, sometimes called a linear or location-based schedule, plots location against time and shows each trade as a sloped line whose slope is its production rate. It is the right tool for repetitive work, parking decks, podium levels, identical floors, where CPM and Gantt hide the production-rate clashes that stack trades on top of each other.

Pull planning and the Last Planner System are not a competing chart so much as a method for building and running the near-term plan with the people who do the work. The CPM gives the framework; pull planning fills in how the field hits it. Most jobs run a CPM for the contract and the claim, a look-ahead for the field, and pull planning to connect the two.

The work breakdown structure and the right level of detail

A work breakdown structure splits the job into the activities you will schedule. The skill is the level of detail. Too coarse and the schedule is useless. A single activity called foundations that spans six weeks tells you nothing about whether you are on track until the six weeks are gone. Too fine and you are managing a thousand-line monster that nobody updates, which rots into the same uselessness from the other direction.

The working rule most schedulers settle on is activities in the range of roughly one to fifteen working days, broken so that a single crew and a single area own each one. For concrete, that usually means breaking by area and by operation: form the east footings, set rebar east footings, pour east footings, then the same chain west. You want each activity to be something a foreman can call done or not done without an argument.

Break the job the way you build it, by area and by trade hand-off, not by CSI division on paper. The division list is for the estimate. The schedule has to mirror the actual sequence on the deck, because that is what the look-ahead pulls from and what the update measures against.

Activities and realistic durations

Each activity needs a duration, and the honest way to get one is crew size times production rate against the quantity, not a number that makes the bar fit the space left on the calendar. Take the quantity from the takeoff, apply a production rate the crew actually hits, and let the duration fall out. Then sanity-check it against the last job that did the same work.

This is the garbage-in problem, and it sinks more schedules than bad logic does. Durations padded to look safe build slack the field then spends, so the job drifts late while every bar reads green. Durations squeezed to hit a date the estimate wanted produce a schedule nobody in the field believes, so nobody works to it. Either way the schedule stops describing the job. A duration is a forecast of crew output, so it deserves the same rigor as the estimate.

Account for the calendar too. A five-day duration is not five days if the work week is five days and a holiday lands in the middle, and concrete carries its own clock: cure time before you strip, before you backfill, before you load the slab. Build those waits in as activities or as lag, because they are real time the next trade has to wait through.

Logic and relationships drive the dates

The dates in a CPM schedule are not typed in. They fall out of the logic, the relationships that say which activity has to happen before which. Get the logic right and the dates calculate themselves. Get it wrong and every date downstream is wrong with it, no matter how good the durations are.

Most relationships are finish-to-start: you finish setting the rebar before you start the pour. The others earn their keep in specific spots. Start-to-start with a lag runs two activities in parallel after a head start, so the second crew can follow the first down the slab a few days behind instead of waiting for the whole thing to finish. Finish-to-finish ties two activities to wrap together. Finish-to-start with a lag is how you carry cure time: pour finishes, then a two-day lag, then strip starts.

The predecessors are where the sequence lives, and they are also where the games get played. Open logic, activities with no predecessor or no successor, lets the software float work anywhere and hides the real path. Hard constraints, a date typed onto an activity to force it, override the logic and quietly create or mask float. A clean schedule earns its dates from relationships, not from dates nailed on by hand.

What is the critical path?

The critical path is the longest chain of dependent activities through the schedule, and its length is the shortest time the job can finish. That sounds backward until you sit with it: the longest path is the one with no slack, so any day it slips is a day added to the finish. Activities on the critical path have zero float. Delay one of them and the end date moves, full stop.

This is the single number to carry in your head. Know your critical path and you know which delay matters. A two-week slip on an activity with three weeks of float costs you nothing on the finish date. The same two weeks on a critical activity costs you two weeks of completion, and possibly liquidated damages. The field instinct to chase whatever is loudest is exactly wrong. Chase what is critical.

The critical path is not fixed. It shifts as the job runs and the updates come in, so the path that ran through the foundations in month one can run through the curtain wall in month four. Recheck it every update. The contract schedule specification and the software define how the path is identified, longest path versus total-float threshold, so confirm which method governs before you argue about what is critical.

What is float in a schedule?

Float is the slack on non-critical work, the time an activity can slip before it starts to matter. Total float is how long an activity can be delayed without pushing the project finish. Free float is how long it can slip without delaying the very next activity. Critical activities have zero total float by definition, which is what makes them critical.

Float is a resource, so use it on purpose and do not waste it. The crew that burns three weeks of float on an easy activity early has handed that cushion away, and when the hard activity downstream needs it, the float is gone. Treat float as the buffer that protects the date, not as permission to start late because the bar says you can.

Negative float means you are already behind. When the calculated finish lands past a contract date or a deadline constraint, the activities feeding it show negative float, and the number is how many days you are in the hole. Negative float is a recovery flag, not a status quo to update around. Who owns the float, the project or the contractor, is set by the contract schedule specification, and on a job with liquidated damages that ownership question decides who gets paid when float runs out. Read the spec before you assume the float is yours to spend.

What is a baseline schedule?

The baseline is the approved schedule, the agreed plan you measure everything against. Once the owner or the CM accepts it, that version is frozen as the reference, and the actual progress on every update gets set against it. No baseline, no measurement, because you have nothing to compare today's reality to except an opinion.

Changes to the baseline need approval. This is the part contractors lose money on. You cannot quietly re-baseline to bury a slip, and you cannot let the owner re-baseline to erase a delay they caused. The baseline is the contract's picture of the plan, so a revision to it is a contract matter that gets submitted, reviewed, and accepted, with the reason documented. Many schedule specifications require the baseline be submitted and accepted within a set number of days of notice to proceed, and they spell out what a revision requires.

Keep the original baseline even after revisions. When a delay claim comes, the as-planned baseline against the as-built actual is the comparison that shows what happened and what it cost in time. Lose the original and you have lost the before picture, which is the half of the claim you cannot reconstruct after the fact.

What is a look-ahead schedule?

A look-ahead is the three to six week window pulled from the CPM that the field actually builds from. The master schedule lives in the office and runs the whole job. The look-ahead is the slice the foreman works this month, with the activities broken to the day, the constraints cleared, the materials staged, and the crews assigned. It is where the schedule stops being a document and becomes the plan for Tuesday.

The look-ahead earns its keep by making the near term real. You take the activities coming up in the window, confirm the predecessor work will actually be done, confirm the submittals are approved and the material is on site, confirm access and inspections are lined up, and assign the crew. Anything not ready is a constraint to remove before the activity is supposed to start, not a surprise the morning of.

Run the look-ahead every week and pull it straight from the master so the two never drift apart. A window most jobs find workable is six weeks of visibility with the first one to two weeks committed in detail. The field cannot plan around a Gantt chart that shows a month as one bar. It can plan around a look-ahead that says which crew is on which footing line Thursday and what has to land before then.

Pull planning and the Last Planner System

Pull planning is the trades planning the sequence together by working backward from a milestone. You put the foremen who actually do the work in a room with the milestone on the wall, and each trade states what it needs handed to it and when, building the sequence back from the finish instead of pushing it forward from a wish. The hand-offs get negotiated by the people responsible for them, which is why the plan tends to hold.

The Last Planner System wraps this into a weekly rhythm. The phase plan comes from pull planning, the look-ahead makes the next several weeks ready by removing constraints, and the weekly work plan is where the foremen make real commitments to each other about what gets done and handed off. The last planner is the foreman, the last person who can commit the crew, which is where the name comes from.

The number that tells you whether the system is working is percent plan complete, PPC: the count of committed activities actually finished as planned, divided by the count committed, times 100. A PPC that sits low means the commitments are not real or the constraints are not getting cleared. Track the reasons for the misses, not just the percentage, because the recurring reason is the thing to fix.

Removing constraints before they reach the crew

A constraint is anything that has to be cleared before an activity can start: the submittal approved, the material delivered, the permit issued, the inspection passed, the access opened, the predecessor finished. The make-ready work is finding those constraints in the look-ahead window and knocking them down ahead of the date, so the crew shows up to work that is actually ready.

This is the part that separates a schedule that runs from one that just exists. Keep a constraint log: the activity, the constraint, who owns clearing it, and the date it has to clear by, which is the activity's start date minus the lead time to resolve it. Work the log every week in the look-ahead meeting. An activity with an open constraint inside the next two weeks is a red flag, because if the rebar submittal is not approved, the rebar is not on site, and the pour does not happen no matter what the bar chart says.

The discipline is to look far enough ahead that the constraint can still be cleared. A submittal that takes three weeks to approve has to be flagged more than three weeks before the work needs it. Find it the week before and the activity is already dead. Make-ready is early or it is nothing.

Updating the schedule, because a stale one is useless

A schedule you do not update is a schedule that lied to you the moment the field changed. Updating is recording what actually happened, the actual start, the actual finish, the percent complete, and the remaining duration on the work in progress, then letting the software recalculate the dates and the critical path from where you really are.

Update on a regular cycle, usually monthly for the contract CPM and weekly for the look-ahead, and update the same way every time. The remaining-duration question is the one people botch. An activity is not 50 percent done because half the calendar time passed. It is 50 percent done when half the work is in place, and the remaining duration is your honest forecast of what is left, which may be longer than the original duration if the crew is struggling. Status the work the field actually did, not the work the baseline hoped for.

The update is also where the critical path moves and where negative float appears, which is the early warning the whole exercise exists to give you. A schedule updated honestly every month catches the slip in month two. A schedule that gets a real update once, at the start, catches nothing, and the first time anyone notices the job is late is when it is too late to recover.

Recovering a job that is slipping

When the update shows negative float, you have two jobs: recover the time where you can, and document the delay you could not control. Recovery on the critical path is the only recovery that moves the finish, so the recovery plan starts there. Re-sequencing is the cheapest lever, finding work that can run in parallel or in a better order to shorten the critical chain without spending money.

When re-sequencing runs out, the levers cost money: add crew, add overtime, add a second shift, or bring the long-lead delivery in faster. Each has a ceiling. Crews stacked too tight trip over each other and production per worker falls, so more bodies past a point buys less than the timesheet suggests. Overtime past about a month bleeds productivity as the crew wears down. Run the recovery against the cost, because the fastest fix is rarely the cheapest, and the acceleration cost may itself belong in a claim if the delay was excusable.

If the delay was not yours, an excusable delay, the recovery and the documentation are separate tracks. You may still accelerate to protect the date, but you document the excusable delay in parallel so the time and the acceleration cost are recoverable. What counts as excusable and what notice it requires is set by the contract, so the contract language governs, not the general rule.

Delay claims and time impact analysis

A delay claim stands on two schedules and one analysis. The baseline shows what you planned. The as-built shows what happened. The time impact analysis shows that a specific delay event landed on the critical path and pushed the finish, because a delay that did not touch the critical path did not delay the job, and the owner's lawyer knows it. Time impact analysis inserts the delay into the schedule as a fragnet at the point it occurred and measures the move in the finish date.

The categories decide who pays. An excusable delay earns more time. A compensable delay, usually owner-caused such as late access or slow approvals, earns time and money. A non-excusable delay, your own poor planning, earns neither. Concurrent delay, when an owner delay and a contractor delay overlap, usually means neither side recovers for the overlap, which is why the timeline has to be clean enough to separate causes.

Notice is where good claims die. Most contracts require written notice of a delay within a set number of days of the event, often somewhere in the range of 7 to 21 days, and a claim filed late is a claim waived no matter how valid. The pricing of the extra time, the extended general conditions and the acceleration cost, runs through the change-order process, which the change-order guide covers. Verify the notice period and the claim procedure in your specific contract, because they vary and the contract controls.

Weather days and contingency

Weather is a known unknown, so it gets planned for, not reacted to. Many contracts give a number of anticipated weather days built into the contract time, and a delay only becomes excusable when the weather exceeds that anticipated baseline. Confirm how your contract counts a weather day, because abnormal weather measured against the historical average for the month is the common test, not simply a day it rained.

For concrete the weather risk is sharper than a lost day. A pour you cannot protect from a hard freeze or place in driving rain is not just delayed, it is a quality problem if you push it. Carry a little float on weather-exposed critical work and log the lost days as they happen, because the weather claim you reconstruct from memory at the end of the job is the weather claim you lose.

Long-lead procurement drives the job

The long-lead items are the ones with delivery times measured in months, and they drive the schedule more than the field work does. Switchgear, generators, elevators, structural steel, large mechanical units, and specialty embeds can carry lead times long enough that the order date, not the install date, is the activity that controls the finish. Miss the order date and no amount of field hustle recovers it.

Put procurement in the schedule as real activities, not as an assumption that the material will be there. Show the submittal, the approval, the fabrication, and the delivery as a chain ahead of the install, and tie the chain to the install with logic so a slip in approval moves the install date automatically. The procurement log and the schedule have to agree, because the most expensive delays on a commercial job are usually a long-lead item that nobody scheduled the order for until it was already late.

For concrete the long-lead risk is quieter but real: specialty rebar, post-tensioning supplies, embeds and anchor bolts set by others, and on a fast job the mix design approval and the batch plant's capacity on your pour day. Confirm the plant can supply the volume on the date before the date is the only thing standing between you and a cold joint.

Coordinating trades, owner milestones, and inspections

The schedule is what coordinates the trades, the owner's milestones, and the inspections into one timeline instead of a series of collisions. Every trade wants the same area at the same time, and the schedule is the referee that decides who gets it when. The hand-off from one trade to the next is where jobs slip, so the schedule has to show the hand-offs, not just the work, and the look-ahead has to confirm the predecessor trade actually finished before the next one mobilizes.

Owner activities belong in the schedule too. Owner-furnished equipment, owner decisions, design responses, and the owner's own milestone dates are predecessors to your work, and putting them in the schedule with logic is how a delay on the owner's side becomes visible and, when it lands on the critical path, claimable. A schedule that only shows your work cannot prove the owner delayed you.

Inspections and tests are activities, not afterthoughts. The rebar inspection before the pour, the special inspection, the AHJ sign-off, each one is a gate that stops the next activity until it clears. Schedule them with the lead time the inspector actually needs to show up, because the pour that waits a day for an inspection nobody scheduled is a self-inflicted delay.

Resource loading and cash flow from the schedule

A schedule with the labor, the equipment, and the cost loaded onto the activities tells you more than dates. Resource loading shows where the crew demand spikes past what you can staff, the histogram that reveals you have three activities all needing the same crane in the same week, so you can level the work before the field discovers the conflict the hard way.

The schedule also forecasts cash. Spread the cost across the activities by their timing and you get the cash-flow curve, the S-curve that tells the owner when to expect each draw and tells you when you will be carrying the most cost before payment catches up. Many schedule specifications require a cost-loaded schedule precisely so the schedule of values and the monthly draw track the planned progress.

This is where scheduling and job costing meet without overlapping. The schedule forecasts when the cost will land and in what order. Job costing captures what the cost actually was once it lands and sets it against the estimate, which the job-costing guide covers. Run them together and a slip in the schedule shows up as a cash-flow problem before it shows up in the bank, which is the early warning a contractor lives or dies on.

Milestones, contract dates, and liquidated damages

Milestones are the dates that carry consequences: substantial completion, phased turnover, the date the owner has tenants or production waiting on. The contract usually ties money to them, and the most common form is liquidated damages, a set dollar amount per day the job runs past the date. Liquidated damages are why the critical path is not an academic exercise. Every critical day late is a number on an invoice you owe.

Build the contract milestones into the schedule as constraints and watch the float to them on every update. Negative float to a milestone with liquidated damages attached is the most expensive number on the job, and it has to drive the recovery decision the moment it appears. Interim milestones with their own damages, common on phased and occupied work, each get the same treatment, because you can hit substantial completion and still owe damages on a phase you let slip.

Whether the liquidated-damages amount is enforceable, how substantial completion is defined, and what relief the schedule gives you are contract questions, and the contract language and the jurisdiction control them. Read the milestone and damages provisions before you assume any of it, because the definition of done in the contract is rarely the field's definition of done.

The contract schedule specification

Most commercial contracts include a schedule specification, usually in Division 01, and it dictates how you schedule, not just that you do. Read it before you build the schedule, because it commonly sets the method, often CPM, the software, frequently Primavera P6 or Microsoft Project, the level of detail, the update frequency, the float ownership, and the deadline to submit and get the baseline accepted.

The spec also sets the rules that decide your claims later. How float is owned, whether the schedule must be cost-loaded and resource-loaded, what a time extension request has to contain, and which delay-analysis method the owner will accept are all spelled out in the spec on a well-written contract. Building a schedule that does not meet the spec is building evidence that will not be accepted when you need it.

When the spec is silent or vague, the recognized scheduling practice fills the gap, the CPM and delay-analysis conventions that the courts and the industry treat as standard. The contract schedule specification governs first, the project's own requirements next, and the general practice last. Confirm what your spec actually requires before you assume the default, because the default is only the default until the contract says otherwise.

Field look-ahead versus office CPM

Two people own the schedule and they own different parts of it. The office, the scheduler or PM, owns the CPM: the logic, the durations, the monthly update, the baseline, and the claims. The field, the superintendent and the foreman, owns the look-ahead: who is on what this week, the constraints, the daily sequence, the commitments. The job runs right when those two agree, and it comes apart when they do not.

The classic failure is two schedules that have stopped talking. The office CPM says one thing, the field is building to a look-ahead that drifted off it weeks ago, and nobody reconciles them until the monthly update reveals the gap. The fix is mechanical: the look-ahead pulls from the CPM, the field's actual progress and re-sequences feed back into the CPM update, and the weekly look-ahead meeting is where the two get squared.

The office cannot status the schedule from a desk, and the field cannot see the whole critical path from the deck. The scheduler who never walks the job builds durations the crew laughs at. The super who ignores the CPM chases the loud problem instead of the critical one. They need each other, and the schedule is the thing they meet over.

Records and the as-built schedule

The records are the evidence, and the as-built schedule is the most valuable one. Keep every update, not just the latest, because the sequence of updates is the story of how the job actually unfolded and where the delays landed. The as-built schedule, the record of when each activity actually started and finished, set against the baseline that showed when each was planned, is the comparison that proves a delay and quantifies it in time.

The records that back the as-built are the daily ones, and they are the ones that go missing. Daily reports with crew counts and what was worked, delivery dates, inspection dates, the constraint log, photos with dates, and the weather are what let you reconstruct the timeline when a claim comes a year later. Memory does not hold up against a paper schedule, and the contractor who logged the field reality day by day wins the argument the one who did not cannot even start.

Capture the field record where the work happens, not on a clipboard that never makes it back to the office. A field tool like FieldOS that ties dated photos, daily logs, and the look-ahead to each activity turns the daily grind into the as-built record automatically, so the evidence exists before anyone knew they would need it. The schedule you can prove is worth more than the schedule you merely kept.

Field checklist

What to document

The schedule record answers the questions that come months out, when a delay surfaces and the only thing that settles it is what you wrote down as the job ran. Capture enough that someone who was not there can reconstruct what was planned, what happened, and why the two parted.

Common mistakes

• Running no real schedule, or just a bar chart with no logic, so nothing tells you what moves the finish.
• Durations that are padded or squeezed, and logic that is missing or faked with hard date constraints.
• Ignoring the critical path and chasing the loudest problem instead of the one that moves the end date.
• No field look-ahead, so constraints surprise the crew the morning the activity was supposed to start.
• Never updating, or statusing by calendar time instead of work in place, so the schedule goes stale and lies.
• No accepted baseline and no kept records, so there is nothing to prove a delay when the claim comes.
• Long-lead procurement, inspections, and owner activities left out of the schedule until they are already late.
• An office CPM and a field look-ahead that have drifted apart and that nobody reconciles.

Standards and references

CPM scheduling is a recognized practice rather than a single code, and the bodies that define it are AACE International, which publishes recommended practices for planning, scheduling, and forensic delay analysis, the Project Management Institute, whose PMBOK frames the schedule-management process and the float and critical-path definitions, and ASCE, which has published standards for schedule preparation. These describe accepted practice; none of them overrides your contract.

The Last Planner System and lean look-ahead and pull-planning methods come from the Lean Construction Institute, and the production-rate logic behind line-of-balance scheduling sits in the same lean and location-based tradition. Use them as the field method on top of the contract CPM, not as a replacement for the schedule the contract requires.

The document that actually governs your job is the contract schedule specification, usually in Division 01, which sets the method, the software, the update frequency, the float ownership, and the delay-analysis approach. Forensic delay analysis follows AACE Recommended Practice 29R-03 and similar references for the accepted methods, including time impact analysis. Cite the practice that fits the point, and let the contract specification and the jurisdiction control where they speak, because they govern over any general rule of thumb.

Units and terms

Scheduling carries its own vocabulary, and the same idea reads differently across a spec, a software export, and a delay report. These are the terms that have to mean the same thing to the field and the office for the schedule to work.

CPM

Critical path method: a schedule of logic-linked activities whose dates and critical path the software calculates from the relationships and durations

Critical path

The longest chain of dependent activities, and therefore the shortest time to finish; a slip on it moves the end date

Float (slack)

The time an activity can be delayed before it matters; critical activities have zero float

Total float

How long an activity can slip without delaying the project finish

Free float

How long an activity can slip without delaying the start of the next activity

Negative float

A signal the work is behind: the calculated finish lands past a deadline by that many days

Baseline

The approved as-planned schedule, frozen as the reference everything is measured against

Look-ahead

The three to six week window pulled from the CPM that the field builds from, with constraints cleared and crews assigned

WBS

Work breakdown structure: the job split into the activities you schedule, broken by area and operation

PPC

Percent plan complete: committed activities finished as planned divided by committed, times 100

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is job costing?
• What is a change order?
• What is lean construction?
• What is the critical path in a construction schedule?
• Why do bridge decks deteriorate faster than the rest of the bridge?