CAPINTELLIGENCE
Data-driven. Field-proven.

CAP·SAR · Schedule Assurance Review

Is your critical path real, and does it reach the in-service date you promised?

The integrated schedule is the one artifact the sanction rides on: every package, every purchase order, the tie-in window, and the in-service date resolve to it. CAP·SAR works that schedule in full and hands your team a prioritized correction plan before the baseline locks at sanction, while the logic can still be fixed.

Deep analysis in. A short, prioritized plan out.

Schedule assurance scoreIllustrative
69 / 100
Quality: Developing Readiness: Conditional

Shared core + project extension · 3 gates
Graded against public standards, not a vendor tool

The problem

Most project schedules get baselined because they look complete.

A schedule with broken logic is a paper map in a nice frame: it shows the route, but it cannot tell you the bridge ahead is out. Sound logic is what lets a schedule behave like a live map instead, rerouting and re-forecasting the in-service date the moment a vendor slips. Most project baselines lock at sanction without anyone stress-testing the logic underneath, and three failures hide there in plain sight.

A path that does not reach the finish
The critical path dead-ends in a payment milestone, not the in-service date

If the driving chain does not run through procurement and the window to the business milestone, the date on the cover is not the date the logic supports.

A procurement gap
The long-lead item is on the schedule, but nothing drives its order

Delivery sits on the plan as a date, not as the end of a chain from IFC through order to receipt, so a late drawing never moves it until the crane is waiting.

A window nobody bounded
The in-window scope is not fenced by the window start and end

On a coupled project the tie-ins have to fit the outage. If the fragment is not bounded, the schedule cannot tell you it does not.


What makes it different

A CPM engine, not a checklist.

CAP·SAR rebuilds your schedule’s logic from first principles with a calendar-aware critical-path engine, then reads it three ways that a settings audit cannot.

01 · Procurement chain

The order behind the delivery

Every long-lead item traced from issued-for-construction through the purchase order to receipt on site, so a late drawing shows up as a late crane before the crane is booked.

02 · Longest path

The path to the in-service date

We trace the driving chain backward from the business milestone, not the network’s last bar, and report how far it actually reaches through real work and through the window.

03 · Contingency

Reserve measured, not assumed

We separate the planned duration from the real built-in reserve to the date you committed to, and flag a finish that logic does not actually drive.

Engineering
IFC drawings
Procurement
Exchanger order
Delivery
On site
Window
Tie-in & install
Records
MC & PSSR
Business promise
In service
We trace the driving chain backward from the in-service date you committed to, not the network’s latest bar. Illustrative: the path runs through procurement into the tie-in window and reaches the promise. When it dead-ends in level of effort or a payment milestone instead, the critical path is not real, and we flag it.

Leveling, and the real driver

Where the duration comes fromIllustrative
Logic 12.1 moLeveling +2 mo
A leveling delay of this size marks the project resource-driven through construction: more crews would compress it. Below five percent, the network governs and more crews will not.

Contingency to the committed date

Planned duration and reserveIllustrative
Planned 12.6 moReserve 6 wk
Six weeks of genuine reserve sit between the logic-driven completion and the committed in-service date. When the finish floats with nothing driving it, that is a defect, not contingency.

Beyond the checks

The numbers a checklist cannot produce.

The same engine that grades the schedule also stress-tests it. These are the two numbers a steering team actually asks for at sanction: how likely is the date, and what do we watch once the critical path is confirmed.

The schedule, simulated

14.1 mo
P50 in service
15.3 mo
P80 in service
61%
odds of the
committed date

Duration risk simulated down the driving path, thousands of runs. Against a fourteen-month commitment the P80 lands five weeks late, driven by the exchanger delivery: exactly the finding you want before sanction, not after. Illustrative.

Near-critical exposure

Exchanger delivery38 d / 2 d
Tie-in window16 d / 4 d
Control-system cutover11 d / 6 d
Days of work carried / days of protecting float · ranked by compression risk · illustrative

After “is the critical path real,” the next question is which chain becomes it. These are the ones a two-week vendor slip promotes, ranked, so your team watches the right work.


Every check we run

The shared core, plus the checks a project needs and a turnaround does not.

The core is the same public-framework set the turnaround practice runs, DCMA 14-point, GAO, AACE, and PMI, because topology, float, and the critical-path cascade are the same discipline on any schedule. The project extension carries what turnarounds lack: engineering and procurement logic, gate milestones as anchors, cost loading, payment linkage, and the window. Each check is weighted by how much it puts the in-service date at risk, and the calibration bands are a project’s own. Nothing is a proprietary score you cannot see inside.

CheckWhat it catches
Configuration & integrityshared core
A1Scheduling settingsThe scheduling options are set so the forward and backward pass can be trusted.
A2CalendarsWork patterns are real, and no calendar quietly runs more hours than its name implies.
A3File & status integrityOne data date, clean actuals, nothing corrupting the calculation.
Network logicshared core
B1Open endsEvery activity has something driving it, so a slip actually propagates.
B2Dangling logicNo activity tied on only one side, hiding its true impact on the finish.
B3Relationship typesFinish-to-start dominant; overlaps modeled honestly, not by convenience.
B4LagsDelays built into the logic are justified, not padding the plan.
B5Leads / negative lagsNo negative lags papering over missing sequence detail.
B6Hard constraintsDates driven by logic, not pinned by constraints that mask the real risk.
B7Redundant & circular logicNo loops or duplicate ties distorting the calculated path.
Float & critical pathshared core
C1Negative floatThe plan is achievable as drawn, not already behind on day one.
C2Unexplained high floatNo activity floating longer than the whole project with no reason it should.
C3Critical-path validity & reachThe critical path is continuous, real, and reaches the in-service date you promised.
C4Near-critical & merge exposureThe paths about to become critical, and the points where many converge, are visible.
C5Float distributionThe criticality profile is sane, not a network with no real spine.
Durations & milestonesshared core
D1Activity durationsNo oversized activities hiding sequence, detail, and risk.
D2Level of effortHammocks and LOE are bounded, not silently inflating the critical path.
D3MilestonesGates, mechanical completion, PSSR, and in-service milestones are structured and tied on both sides.
Resources, leveling & densityshared core
E0Leveling stateWhether the schedule was resource-leveled, determined forensically from the file.
E1Resource loadingThe work that should carry crews and hours actually does.
E2Assignment integrityNo broken, zero-unit, or impossible resource assignments.
E3Leveling coherenceThe leveled dates and the underlying logic agree with each other.
E4Resource densityThe manpower curve is physically achievable, not a wall of parallel work.
Coding & traceabilityshared core
F1Code completenessPhase, area, system, discipline, and package coding present, so the plan can be sliced and audited.
F2WBS structureA clean, meaningful hierarchy, not empty scaffolding.
F3TraceabilityActivities tie back to work packages and purchase orders and carry unique, legible names.
Project extensioncapital projects
P1Engineering & procurement logicIssued-for-construction before fabrication, purchase order before delivery, delivery before install: the sequence the field actually depends on.
P2Gate milestones as anchorsG3, G4, mechanical completion, PSSR, and in-service exist as milestones, tied on both sides, and the plan re-lays from them.
P3Cost & resource loadingThe baseline carries the hours and the money the control estimate says it does, so earned value can be read against it later.
P4Milestone-payment linkagePayment milestones tie to real completion, not to dates a contractor can hit without finishing the work.
P5The window fragmentOn a coupled project, the in-window scope is bounded by the window start and end, and the pre-window work is verified complete before it.
P6Permit & weather windowsSeasonal and permit constraints are modeled as what they are, not as constraints that mask the real driver.
Gate 1 · Configuration

The engine settings are valid, so the calculated dates can be trusted at all.

Gate 2 · A real critical path

The critical path is continuous, made of real work rather than level of effort, and reaches the in-service date.

Gate 3 · Status integrity

One data date and clean actuals, so the measurement rests on solid ground.


How it reads out

Two answers, never blended into one.

A schedule can be well built and still not ready, so we report quality and readiness separately. You see how good the model is, and, independently, whether it is safe to lock as the sanction baseline.

Quality · how well the schedule is built (0 to 100)
UnreliableDeficientDevelopingSoundBest-in-class
045607585100

An absolute maturity grade across every element. Project schedules are longer and lower-density than turnaround ones, so the bands are calibrated to projects and the grade is honest.

Readiness · is it safe to lock at sanction
Clear – safe to lock the baseline at sanction
Conditional – fix the named items first
Failing – a structural showstopper
Not Assessable – basis must be corrected

When readiness is withheld, the report says exactly why, in the language a scheduler can act on rather than a code.


Why you can trust the output

Findings that survived a challenge.

Every figure in the review is computed by the engine, never by a language model. The interpretation on top has to earn its way out, twice, before it reaches your team.

01 · Deterministic

The engine writes every number

The CPM engine and the checks produce every figure in the report. The model reads them and explains them; it cannot write one.

02 · The numeric wall

Untraceable figures kill the finding

A finding may only cite numbers that exist in its own evidence. A figure that cannot be traced back to the engine removes the finding, automatically.

03 · The adversarial verifier

Refuted findings never render

A second pass is built to argue against every finding from the same data. Only the findings it cannot knock down reach your report.

Ask any vendor what happens to an AI finding that is wrong. Here, it never renders.


What you get

A correction plan handed over before the baseline locks.

The analysis is exhaustive. What comes back to your team is not: a clear, prioritized plan that maps each issue to its fix, early enough to make the changes before sanction freezes the baseline and the date is committed.

In
Your P6
schedule
→
Engine
CPM + core
+ extension
→
Out
Correction plan
+ workbook

The correction plan is the decision document: each finding, why it matters, and the specific fix, in priority order. The workbook behind it is the line-by-line repair manual, every flagged activity with its issue and acceptance criteria.

And the findings do not expire with the review. At closeout, CAP·SPR traces the defects flagged here to the specific as-built activities they delayed and the days they cost. The defects we flag now are the ones we can later prove mattered.

CAP·SAR runs as a standalone review, or as the schedule engine inside the CAP·AR3 sanction readiness review, where it feeds the final investment decision.


What changes

The difference before the baseline locks.

Without CAP·SAR

  • ×A critical path no one confirmed actually reaches the in-service date
  • ×A long-lead delivery that nothing drives, found when the crane is waiting
  • ×Contingency everyone assumed but the schedule never carried

With CAP·SAR

  • ✓A driving path verified to run through procurement and the window to the date you committed to
  • ✓Logic corrected before sanction, while it is cheap to change
  • ✓The real reserve to your date, measured and stated

Your guide

We have sat in the seat, on both sides of the table.

CAP Intelligence is an independent assurance practice built from the owner’s side, across refining, petrochemicals, industrial gas, LNG, and power, as the owner accountable for the capital outcome and as the consultant brought in to protect it. It is the capital-projects practice of STO Intelligence LLC, and shares one spine with its turnaround practice.

Our methods stay at public-framework level, AACE, CII front-end planning practice, PMI, DCMA, and GAO, so every finding is defensible and nothing is a black box. You own the standard, not a proprietary score you cannot see inside.

Why CAP Intelligence

Independent Owner-side Public-framework method Refining · Petrochem · LNG · Power Veteran-owned

Empathy for the seat you are in, sanctioning a project on a schedule you cannot see inside. Authority to check it: a real CPM engine, public frameworks, a project-calibrated standard, and owner-side experience of what a late long-lead item does to a window.


Where this goes

One review fixes one gate. The system is the prize.

The correction plan fixes this schedule. The same engine runs inside the CAP·AR gate reviews when you want the whole package checked, CAP·SPR proves at closeout which findings mattered, and the discipline that prevents them lives on in CAP·PATH + CAP·CORE, the standard you own, run live.

Run an independent engine over your schedule.

Send us the P6 file for the project you are taking to sanction. We will show you where the logic holds, where it does not, and what to fix before you lock the baseline.