Frequently asked questions

Short answers to the questions people arrive with. Where an answer rests on a measurement, the measurement lives on Is this real? and How it works rather than here, and this page links there instead of restating it.

What is forensic schedule analysis?

It is the work of establishing, after the fact, how much of a project's delay each event accounts for, using the schedule as evidence. It differs from schedule quality checking in that the question is not whether the programme is well built but what it shows about what happened.

It is not one technique. AACE RP 29R-03 organises the field into nine method implementation protocols, and they answer different questions, need different data, and produce different numbers on the same file. Choosing between delay methods is about where they diverge; Methods treats each one in turn.

What does this tool do?

It reads a schedule file without opening Primavera, computes the dates and floats itself, and then does two things with the result. It checks the file against the clauses of the specification or standard that governs the contract, citing the clause with each finding. And it runs delay analysis methods by their AACE number, reporting what each one computes along with the choices that produced it.

Everything it emits is traceable to an input: the clause it checked, the field it read, the option the file itself declared. See What it does and How it works.

What does it not do?

It does not conclude. It will not tell you a delay is excusable or compensable, who is responsible, whether there is concurrency governing under a US jurisdiction, or whether a schedule is reasonable. Under US jurisdictions the concurrency module reports every theory's finding and returns no governing answer, because the authority that governs there endorses none.

It does not approve or reject a submission. The reviewing authority determines that, and every report says so at the top. It never writes a schedule file, never modifies what it is given, and will not hand back a corrected schedule.

Where a clause asks for a judgement no software can make, it refuses rather than returning a green tick — a rule classified as judgement cannot return a pass at all. Limitations is the full list, including the things it gets wrong.

How is it different from a schedule-health checker?

A health checker asks whether a schedule is well built, against a general-purpose library of metrics with configurable thresholds. This asks whether the schedule satisfies the clause in your contract, and cites the clause.

That difference has a practical edge. "Seven activities have no predecessor, which the section governing this contract does not permit" is a statement someone can check against their own document. A generic threshold — a percentage of relationships carrying lags, say — is somebody's house rule, and the reasonable first response to it is "says who?".

Three other differences. Checks that could not run are counted and printed as coverage beside the result. Clauses needing a supplied contract term report as not evaluated and name the term, rather than substituting an industry default. And it reads how the file was computed, from the schedule options the file itself records, so a submission that satisfies every visible requirement while being optimistic by exactly the logic that was dropped is visible.

The health-checking ground is well served and we do not claim otherwise. Oracle's own documentation describes a schedule health score in Primavera Cloud with configurable criteria, and a DCMA-criteria scoring surface in Construction Intelligence Cloud Advisor. Several vendors publish products with far larger metric sets, longer production histories and security certifications this project does not have. If you already run one, the useful question is what it did not catch. See the DCMA 14-point checks.

Does this replace an expert?

No, and it is built so that it cannot. It is a tool, not an opinion. It computes and reports; it does not render a professional judgement, and on the questions that decide a dispute it refuses outright or returns every position the authorities take.

No professional engineer licensed under any state practice act has reviewed or adopted its output. Where a specification asks for exactly the judgement that requires one, the tool abstains and names what it would need. The correct use today is to find the mechanical defects before a qualified person spends an afternoon on them — not to decide anything.

What file formats does it read?

Four: Primavera P6 XER, Primavera P6 XML, Microsoft Project MSPDI, and USACE SDEF. The format is decided by the contents of the file rather than by its extension, so a renamed export is still read correctly and a file matching none of the four is refused as unreadable rather than guessed at. It does not read a binary .mpp file; export from Microsoft Project as XML instead.

They are read; they are not equally evidenced. Everything the engine has been measured against is XER. The other three readers rest on this project's own tests only, and two of them carry no resources or costs. Limitations states this plainly, and the XER format explains what a P6 export actually contains.

Do I need Primavera installed?

No. It reads the exported file directly and computes the schedule itself — no Primavera dependency, no automation of the desktop application, and no licence of anyone else's software required to run it.

That independence is the point of the exercise as well as a convenience: an engine that asked P6 for the answer could not be used to check P6's answer. It is also why any figure it produces should be reconciled against the source scheduling application before it is relied on — the report says so itself.

How can the other side check a finding?

By re-running it. Each run can emit a record binding three digests: the bytes as received, the network parsed out of those bytes, and every date, float and flag computed from that network. A verify step either reproduces the answer or names the activity and the field where the two runs disagree — "the digests differ" is useless in a dispute, because it tells neither side where to look.

The record is append-only, hash-chained and independently verifiable. Verification describes the format and also states the things no record of any kind can establish, which is the more important half.

Conformance findings are checkable without any of that machinery: a finding names the clause and names the activities, and anyone with the file open can look.

What happens when it cannot decide something?

It abstains, and names what it would need. That is the most common outcome, by design.

Three shapes of refusal are worth knowing. A clause with no threshold supplied reports as not evaluated and names the term; no industry default is substituted for a bracket the contract was supposed to fill in. A rule that turns on judgement cannot return a pass. And a request for a delay method by an ambiguous name raises rather than guessing: ask for a "time impact analysis" and it names the two methods that phrase could mean and asks which you meant.

Abstaining cannot raise the score; it lowers coverage instead, and both numbers print together. See Reading a report.

Which standards and specifications does it read?

The analysis-method side reads AACE RP 29R-03 and RP 52R-06, ANSI/ASCE/CI 67-17, and the SCL Delay and Disruption Protocol. The conformance side reads federal sections including UFGS scheduling specifications, a set of state DOT scheduling specifications, GAO-16-89G's scheduling best practices, and the DCMA 14-point criteria.

Which of them governs is a fact about the contract rather than a preference. Where two governing documents answer the same question differently, the report prints the disagreement rather than resolving it silently. See the standards crosswalk, UFGS scheduling clauses, state DOT specifications and the GAO schedule assessment guide.

If a standard is listed, is every clause of it automated?

No, and the gap is published rather than glossed. A substantial minority of the clauses in the sections implemented here are deliberately not automated, each with a stated reason — most often that the clause asks for a judgement, or depends on a document the schedule file does not contain, or requires a contract term nobody supplied.

Naming a standard means the rules that exist were written against that document and cite it by clause. It does not mean a passing report is a statement that the schedule satisfies the whole section. Conformance carries the clause-by-clause coverage, and Limitations carries the count of what is unimplemented.

Is it correct?

That question has a measured answer and it belongs to the page that owns it, with its population, its caveats and the several occasions the figure moved down. Read Is this real? before forming a view, and Limitations beside it.

The short version: it is not validated against Primavera at scale, a schedule arithmetic error is possible, and every report carries that limitation on its own face rather than leaving you to find it here.

Is my project data safe?

The engine runs locally and sends nothing anywhere. It has no network calls and no third-party runtime dependencies — enforced as a build failure rather than as a policy — and its browser interface binds to the local machine. This website serves generated demonstration material and saved outputs; it does not receive or analyse schedule uploads.

What is not claimed: no SOC 2, no ISO 27001, no FedRAMP authorisation, no penetration test. If your office requires any of those before a file can be shared, we would rather you applied that bar than waived it. See privacy.

If the project is in dispute, or a dispute is anticipated, do not send a file directly. Route anything through your counsel of record. A schedule handed to a third party outside that routing is a disclosure your opponent can ask about, and that is not a risk for us to take on your behalf.

What does it cost?

There is no published price, no trial and no order form. Nothing is sold from this site and no software is downloadable from it. A request is the beginning of a conversation about scope, not a purchase and not an automatic licence grant. See enterprise access and deployment.

Who is behind it, and who else uses it?

It is an independent research project carried out by one person, with no company behind it. There are no customers, no pilots and no agency users; no professional engineer has reviewed any output, and no state DOT has seen any rule pack. That is the weakest fact about the project and it is stated first rather than buried. See About.

How do I arrange a demonstration?

Start with the material already published: the sample reports, the worked example including the clauses the engine refused to decide, and the walkthrough. Those are real runs of the real command and will tell you more in ten minutes than a call will.

If you want to talk after that, email is the whole process — no form and no scheduler. Talk to us has the address and says what happens to what arrives. No project file is needed for a first conversation, and for a project in dispute none should be sent without counsel.

Source: web/pages/faq.md. Source commit date: 2026-09-11.