Valeur.See your contract’s weak clause before you sign.
For the contracts managers, in-house counsel and project owners who sign FIDIC and public-procurement contracts. Every contract is a load-bearing structure; Valeur finds the weak member before signature — not after a dispute begins. Under FIDIC 20.1/20.2 the notice window is 28 days from awareness; when it closes, your claim is time-barred.
A contract clause shows its price on site. Valeur works it out at the signing table.
A bridge opens to traffic only after a load test. A contract, by contrast, carries millions of dollars of load every single day untested — until one clause gives way, at the worst possible moment. Valeur runs that test before signature.
Every figure in this band is sample analysis output · simulation
Nine out of ten such projects have cost overruns.
Bent Flyvbjerg · Project Management Journal 45(2) · 2014. The paper gives no sample size, so this page does not say how many projects were counted.
Valeur makes no claim on that figure. Valeur changes what you hold on signature day: every notice obligation in the contract becomes a dated calendar line, and every clause becomes an assessment with its source printed on it.
Legal, finance and operations look at the same contract through different windows. Valeur aligns all three on a single decision surface.
When the other side sends a new draft, the text diff is converted into decision impact: which sentence moved which range.
Instead of a single-point estimate, the range, the scenarios and the assumptions sit on one card — every number arrives with the ground it rests on.
The daily record from site drops straight into the claim file as evidence; the gap between site and contract closes.
Tenant isolation, role-based views and the audit trail are on by default.
A four-step decision line instead of a long consultancy cycle.
A contract’s price is usually learned late. Teams see where it costs them at the moment the clause bites: after the notice day has passed, after the delay has taken root, after the counterparty has taken its position. The four steps and five functions below put the same knowledge on the table before signature.
Upload
Main contract, appendices, specifications and revisions are brought into a single tenant space.
Analyse
The clause graph is extracted: parties, obligations, dates and cross-references are linked.
Attack
The counterparty rehearsal and the scenario simulation keep pressing on the structure; every point that breaks becomes a finding with its assumptions written next to it.
Bring to the gate
The Signature Gate decision comes with an evidence file: findings, range, recommendation and audit trail.
Under steps 02 and 03 sit five separate functions.
The contract’s notice obligations are written into the project’s own calendar. The 23 clauses of the FIDIC 2017 Red Book list arrive with whether they carry a time bar and who the notice goes to. When an event is logged the last day is computed and its basis is printed; where a clause carries no fixed period the claim is anchored to the 20.1 time bar. Every obligation rests in one of four urgency bands.
The time, programme and notice card →The simulation that plays out delay and dispute scenarios before signature — sold under the name Quantum Foresight — turns the project’s drivers into probability bands, produced by a 10,000-iteration Monte Carlo run. Each driver carries the origin of its input as one of three values: real project data, an evidenced record, an assumption. Headline bands are computed from the first two; assumption-labelled drivers are listed in their own section under their own label.
Provenance labels →The dispute is played as a three-stage ladder: direct negotiation, mediation/DAAB, arbitration. The size in dispute is resolved first from the contract’s real numbers; where an estimate is supplied the result returns under an assumed label. Every assumption used is listed with its value in an open ledger.
The dispute model →13 FIDIC clause topics are treated separately across 4 legal traditions; in the legal-tradition mapping table, 46 countries are bound to one of those four. Each answer carries the assessment, the procedural steps, the 1999 and 2017 clause references, and the source name with its link. Under each answer sits a mandatory record: the output is an assessment framework and the final decision belongs to counsel.
Jurisdiction and disclaimer →Every reviewed and approved rule is tested against all clauses on each subsequent upload and rescan. The test is deterministic: the rule’s detection hint is sought as a case-folded substring of the clause text. This layer stays deterministic end to end; the model reading is a separate layer and the two appear under separate stamps.
The signature library →If this goes to dispute, where does the other side stop? Valeur plays it out before you sign.
A dispute’s cost grows along two arms. One is preparing the claim; the other is the time your money spends tied up. The outcome usually favours not the side with the best legal argument, but the side that worked out in advance how the other will behave at each stage. Valeur plays the dispute as a three-stage ladder — direct negotiation, mediation/DAAB, arbitration — and labels every number it uses with where it came from: evidence resolved from your contract, or an assumption stated in the open.
The average value of disputes in North America was $60.1 million. In 2024 the average time to resolve a construction dispute was 12.5 months.
Arcadis · 15th Annual Construction Disputes Report — North America · 2025. The scope is North America only. The value here is the amount a party claimed, not money lost. The clock starts when the claim is formally submitted.
Those two figures are sector averages; they are not your file’s value, and Valeur makes no claim to shrink them. What Valeur does is resolve your file’s size from the contract’s real numbers and play the three-stage ladder on that size.
Not a single forecast but a fan of scenarios. Move the cursor; read the range for any month.
Delay is argued in days and paid in money. That is why it stays in dispute longest. Each side does the day-to-money conversion in its own favour, and what the table cannot agree on is usually not the delay itself but its price. The simulation that plays out delay and dispute scenarios before signature — sold under the name Quantum Foresight — runs ten thousand walks, moving the contract parameters one at a time, and prints under each driver whether its input was real project data, an evidenced record, or an assumption. The cone below draws 64 of those walks over 24 months: steel-blue P10–P90 band, gold median, vermilion worst-case trace.
On average, a one-year delay correlates with a 4.64-point increase in percentage cost overrun.
Bent Flyvbjerg · Project Management Journal 45(2) · 2014. The source’s own language sets up a correlation, not a cause.
Valeur promises you no such coefficient; that is a published finding about how delay turns into money across the sector. What Valeur does is turn your project’s own delay drivers into probability bands, and print under each band the data it came from.
Which clause costs you what — on one screen, with the evidence.
Clause Analysis
Clause, party, governing law and wording differences in one context; the cross-reference chain is visible.
Risk Engine
Every finding is tied back to its driving clause with the notice window beside it, so a time bar becomes visible before it expires.
Financial Impact
A clause change is converted into a cash impact: a set of scenarios, with the assumptions behind them.
Revision Delta
When the counterparty’s draft lands, what gets reported is the commercial delta.
Signature Gate
The GATE decision comes with an evidence file: management sees what was signed and why, in the audit trail.
Workers spend more than 14 hours a week looking for project information, resolving disputes and correcting errors; 5.5 of those hours go to searching for project data alone.
PlanGrid & FMI · Construction Disconnected · 2018. A self-reported survey of 599 respondents, sponsored by PlanGrid, a construction software firm.
Those hours are the sector’s own reporting; Valeur makes no claim to reduce them. What Valeur does is make a judgement permanent, and apply the same rule the same way on every scan that follows.
From contract to site and site to contract: one spine, both directions.
A subcontract template is produced in-house, used once, and then becomes everyone’s starting point. A template that circulates ahead of its review multiplies the same risk across dozens of files through a single clause — and it is the slowest kind of loss to notice. Contract Management governs what was signed; Construction Management governs how the signed contract lives on site. Both halves are built on the same data spine.
Only 31% of all respondents’ projects came within 10% of budget in the past three years.
KPMG International · Climbing the Curve — Global Construction Project Owner’s Survey · 2015. A self-reported owner survey of executives at more than a hundred organisations; the report is dated 2015.
Valeur makes no claim on that ratio. What Valeur does is bind the production decision for your own templates to a measurable review state, and re-read that state on every call.
Six teams, six questions, one contract. Valeur answers all six from one clause–risk–cash spine.
The one-page truth. For every contract in the portfolio: the range, the gate status and the three largest drivers — each one tied to an evidence file. The signature decision reaches the board as a defensible package.
The one who lives the clause. Clause graph, notice windows and revision delta on one screen; as a time bar approaches, the system raises the alarm before the Contracts Manager does.
Guardian of the margin. Claim strength, variation orders and the cash range in one narrative; the scenario engine shows which negotiating move moved the range by how many millions.
Engineer of the wording. Every gap the stress test breaks open arrives together with alternative clause language; legal moves risk from opinion to evidence.
Owner of the range. A P10–P90 cash band instead of a single-point forecast, with escalation and FX sensitivity; the CFO walks into the financing meeting with a scenario set.
The signal from site. Daily reports and evidence capture feed claim strength directly; site sees what the contract requires, with the clause reference.
Contract exposure is largest in EPC. Valeur starts there; the clause–risk–cash spine is the same at every scale.
This is where a single contract carries the highest load: FIDIC-based, multi-jurisdiction, ten-year commitments. The value of a load test is clearest at this table.
The site and product language shifts with the main line of work, the sub-sector and that sector’s critical pressure — supply and FX in solar, expropriation and permits in metro, power commitment in data centres.
From a residential contractor to a multinational petrochemical Employer, the same clause–risk–cash spine; only the module set and the depth change.
These four are set out on their own pages, each with its own pressure.
-
Offshore Oil & Gas Construction
LOGIC and EPCI ground: metocean-driven extensions of time, vessel availability, and SURF-to-topside interface claims.
The sector page → -
Mining Construction
EPCM and Silver Book ground: the geotechnical report against actual ground, process capacity guarantees, and shaft advance rate.
The sector page → -
Nuclear Construction
Heavily amended EPC ground: regulatory approval delay, NQA-1 compliance, and the nonconformance notification chain.
The sector page → -
Tunnel & Underground Construction
Emerald and NEC4 ground: the GBR against actual ground, TBM advance productivity, and dewatering cost from water ingress.
The sector page →
Seven questions to ask before you buy contract analysis.
Is my contract read by AI?
There are two readings and they stay apart on screen. The first is the model reading: it assesses the clause text on four dimensions and arrives under its own stamp. The second is the reviewed-rule scan: the detection hint of a rule your legal team approved is sought as a case-folded substring of the clause text, producing a record on an exact hit. That second layer contains no model call. Each record is stamped with the reading it came from, so you never have to guess which is which.
Can I sign a contract on this output?
The output is an assessment framework and it says so on itself. Every answer from the legal knowledge base carries a mandatory record: the legal outcome depends on the concrete facts, the governing law and the tribunal, and the final decision requires qualified counsel. Reviewed-rule hits carry their verification status openly too: the record states that it comes from the library and remains open to lawyer confirmation. Valeur’s job is to put a sourced, traceable framework in front of your counsel.
How real is the forecast number on screen?
Every driver of the forecast carries the origin of its input as one of three values: this project’s real data, an evidenced record, or an assumption. Headline bands are computed from the first two only. Assumption-labelled drivers sit outside the headline, in their own section, under their own label. So you read separately which part of the number rests on measurement and which on assumption - you take that split on the evidence, not on trust.
Where do the behavioural assumptions in the counterparty rehearsal come from?
Every parameter of the counterparty profile is marked as an assumption, and the defaults are defined as illustrative. Each assumption used is listed with its value and origin in an assumptions ledger. The module’s own note says to calibrate against your real dispute data before relying on it. The size in dispute is left to no assumption: where possible it is resolved from the contract’s real numbers and returns under an evidenced label; with no basis for it, the field stays empty rather than invented.
How many countries do you cover?
There is no single true answer, so we tell you exactly what we counted. Our legal-tradition mapping table holds 46 countries, each bound to one of four traditions: 23 civil law, 12 common law, 10 GCC civil law, 1 mixed. The assessment is produced through the tradition a country belongs to, so a country’s answer is as detailed as that clause-tradition cell. A country-specific note is defined for the United Arab Emirates and covers the onshore/offshore distinction.
If I scan the same contract twice, do I get the same result?
For the reviewed-rule scan, yes - by definition. A hit is the rule’s detection hint occurring as a case-folded substring in the clause text; the same text and the same library give the same hits. The notice calendar is deterministic in the same way: the last day is the event date plus the clause’s notice period, and the basis of that arithmetic is printed inside the output. The model reading is a separate layer under its own stamp, so you can tell the two apart on screen.
When does subcontract template production open?
When the live state of the review reaches the full programme. The programme is 19 packs; at decision time the gate reads the count of distinct reviewed packs and opens when that count reaches 19. The gate’s answer describes itself: reviewed count, programme size, progress ratio and each pack’s status come back together, so you can see which pack is next. If the read fails, the decision stays on the protective side and the state is logged.
Tenant isolation
Each client’s data stays in its own space; cross-access is closed off by architecture.
Role-based views
Legal, finance and site see the same data through their own permission window.
Audit trail
Every finding, decision and revision is recorded with a timestamp.
Multiple documents and jurisdictions
Multilingual contract sets and differing governing laws in a single analysis.
KVKK / GDPR aligned
Data residency and processing are configured to meet enterprise compliance requirements.
Let us walk your next contract clause by clause, before you sign.
The most expensive clauses are the least read. The closer the signature gets, the truer that becomes. Give us thirty minutes and we will walk a set of clauses from one of your own contracts: what each clause carries, which assumptions the range rests on, and what the Signature Gate puts on the table.