01 / REVIEW
Check the governing clauses.
Review a submitted schedule against the specification you name. See findings with activity IDs, the relevant clause, and the evidence behind each result.
Explore schedule checks Construction schedules. Open to scrutiny.
Every finding.
Every assumption.
Nothing hidden.
Your schedule tells a story. Schedule Forensics shows the working behind it: the clauses checked, the choices made, and the questions a person still needs to answer.
An independent research project for schedule reviewers and delay analysts. Explore the evidence, then decide whether it fits your work.
Schedule reviewGenerated example
Bridge widening · Update 03
Baseline + three updates · Contract terms supplied
A computation, not an approval. The reviewing authority makes the determination.
The importer could not read all supplied material as written. Read the qualifications with the findings.
81%22 of 27 scored clauses passed
24%27 of 113 applicable clauses scored
5 failedFindings tied to the clause and the evidence.
6 need a personJudgment is identified, never filled in silently.
68 not evaluatedMissing inputs named so the review can move forward.
Plus 12 observations · SummarySee the working
Evidence, up front
The engine matches P6's stored values on 67.85% of 584,687 measured field comparisons. It is not a replacement for P6. Recheck computed dates and floats before relying on them. Read the full measurement ↗
From a file to an answer you can inspectKnow what the schedule says.
Know what the schedule says.
And what it doesn't.
One workflow for understanding the submission, checking its governing requirements, and tracing the basis of an analysis.
02 / ANALYSE
Make the method explicit.
Run a named delay method with its assumptions in view. All nine AACE MIPs are implemented, alongside prospective time impact analysis. Their evidence and limits are published.
Understand the methods03 / VERIFY
Leave a trail someone can follow.
Keep the inputs, settings, digests, and results together in a reproducibility record. Re-run the calculation and identify where an answer diverges.
See how verification worksThe workflowFrom “is this right?”
From “is this right?”
to “here is the basis.”
The engine reads the file, runs the checks, and exposes the unanswered questions. You bring the contract, the project context, and the judgment.
See what running it involvesStart with the schedule and its context.
Supply the baseline and updates, select the project and jurisdiction, and name the governing specification.
Read the findings with their qualifications.
Assumptions come before the results. Passed, failed, referred, and unevaluated clauses stay distinct.
Find out what would settle the gaps.
The unanswered command groups missing evidence into a worklist. Supply the contract terms and run the checks again.
Make the answer reproducible.
Export Markdown, HTML, or JSON. Record the inputs and settings so a later reviewer can retrace the computation.
Inside the generated example81% passed.
81% passed.
24% was scored.
Both numbers are true. The first describes the outcome of scored clauses. The second shows how much of the applicable review received a score.
All 113 applicable clauses27 scored / 113 applicable = 24% coverage
- 22 passed
- 5 failed
- 12 observations
- 6 need a person
- 68 not evaluated
Only the 27 scored clauses22 passed / 27 scored = 81% score
An observation, a referral, or an unevaluated clause is not a pass. Inspect the complete results ↗
One finding, traced
Does the reported path span the schedule?
The example reports a five-activity path beginning at CURE. Its start is later than the earliest activity in the network—a finding worth inspecting, with an arithmetic qualification attached.
Read the finding and its basis ↗- CURE
- PAVE
- STRIPE
- PUNCH
- OPEN
- Earliest activity in network
- 2 March 2026
- Reported path starts at CURE
- 1 June 2026
Sequence only; not a duration chart. This finding depends on computed arithmetic. Recompute in P6 before relying on it.
Designed for scrutinyA confident number
A confident number
isn't the whole answer.
A review can pass most of the clauses it checked and still leave most of the schedule unchecked. That is why score and coverage always travel together.
Missing evidence never becomes a pass.
A check that lacks a required input reports “not evaluated” and names what it needs. It does not quietly insert a default.
Professional judgment stays with a person.
A clause registered as judgment cannot return a pass. The tool assembles evidence and states the question for the reviewer.
The method is part of the answer.
An ambiguous method name is refused rather than guessed. Different methods can give different results on the same events; the method and choices belong beside the number.
Limitations belong in the open.
Whole-day arithmetic, limited evidence for some file formats, and the current validation gaps are documented. Read those limits before using a computed date or float.
Grounded in the documents you work withThe clause.
The clause.
The context.
The qualification.
Checks reference published scheduling specifications and analysis standards. Naming a guide does not make it contractually binding; the governing documents and their precedence must be stated.
UFGSAACEASCE 67-17SCL ProtocolGAOState DOT specifications
Read formats. Unequal evidence.
Readers cover Primavera P6 XER, P6 XML, Microsoft Project XML, and USACE SDEF. The external corpus evidence is XER; the other readers rely on the project's own tests. Binary .mpp files are not supported.
Explore formats and coverageTake a closer lookBring your questions.
Arrange a walkthrough Bring your questions.
We'll bring the working.
Walk through the generated example, discuss the clauses you review, and decide whether this approach is useful. No file, signup, or purchase required.
Source: web/pages/index.md. Source commit date: 2026-09-09.