The evidential architecture of contemporaneous records

A record is not a document count

Large projects routinely produce millions of files and still fail to prove elementary propositions. The problem is rarely absolute absence. It is fragmentation, ambiguity and missing linkage.

A photograph shows water on a slab but not the date, the location, the planned activity or the work prevented. A daily allocation shows forty operatives on site but not their tasks or their output. A programme update moves an activity without explaining the logic change. A cost code holds overtime, plant and supervision but not the event that required them. A notice is contractually prompt and factually generic. The project has data and no coherent evidential proposition.

A short record can be powerful when it answers the right questions: what happened, why it changed the work, what it cost, and what it did to the programme. A useful record lets another person, who was not there, test all four against independent sources.

Contemporaneity matters because memory is reconstructive. A diary written at the end of a shift is not automatically correct, and an account written two years later is not automatically false. But a timely record reduces the room for unconscious rationalisation, preserves detail, and gives the other party the chance to investigate and mitigate. Its weight rises when it is routine, specific, corroborated, and created before anybody knew what the dispute position would be.

Which is why an unfavourable entry strengthens a record system. If the daily reports acknowledge contractor-caused congestion on the Tuesday and record employer-restricted access on the Wednesday, they read as operational observation. A system that only ever records client delay is not a record system. It is a pleading drafted one day at a time.

What the principal guidance actually requires

Technical guidance The SCL Protocol identifies six record groups: programme, progress, resource, cost, correspondence and administration, and contract and tender documents. Its purpose is to push the parties to deal with time and disruption contemporaneously rather than reserve everything for retrospective analysis. It is guidance, not legislation, not a contract term, and not a universal statement of law, and it yields to the contract and the law of the forum.

Technical guidance The RICS material on ascertaining loss and expense, rewritten and reissued in 2024, emphasises factual, contemporary records kept in date order with unique references. That serves a specific quantum purpose: additional cost has to be identified, evidenced and connected to the matter said to have caused it. Account records establish that money was spent; operational and contractual records establish why it may be recoverable.

The schedule guidance reinforces a point that gets lost in disputes: a critical path is not a static coloured line preserved in the baseline. Criticality migrates as work proceeds, so native schedules, status dates, actual starts and finishes, calendars, constraints, logic changes and update narratives all matter. A PDF bar chart communicates a result and cannot expose the assumptions that produced it.

None of these publications creates entitlement. Notice requirements, burden and standard of proof, admissibility, causation and damages remain matters of contract and governing law. That qualification bites hardest on international projects: an English judgment may be persuasive learning for a project in the Gulf and it does not displace the applicable civil code, the contract, the arbitration law or the tribunal’s directions.

Build an evidence graph, not an archive

The basic unit is a controlled event, not a folder. Each material event needs a stable identifier used across operational, programme, commercial and correspondence systems, and the project builds a graph around it.

• Date, time, shift, location and work-package identifiers.
• The planned activity, method, resource and output at the relevant time.
• The observed event or changed condition, stated factually.
• The instruction, design issue, access restriction or other source relied on.
• Labour, plant, subcontractor and material resources affected.
• Installed quantities and measurable output before, during and after the effect.
• The immediate consequence: idle time, resequencing, stacking, rework, extra shift, changed method.
• Mitigation attempted, by whom, with the decision and the result.
• The programme activity, status date and path said to be affected.
• The cost code, estimate, actual transaction and forecast consequence.
• The relevant notice, continuing particulars, response and determination.
• Source attachments with author, timestamp, version and custody history.

That is not a demand that a supervisor complete a legal questionnaire. Capture is layered. The supervisor records observable fact and resource and output data at the workface. The planner connects it to sequence and forecast. The commercial manager classifies the possible mechanism and segregates cost. Document control protects provenance. The event owner closes the gaps at the weekly review.

Three controls decide whether the graph is trustworthy. Preserve the original, so corrections create an audit entry rather than silently overwriting history. Maintain a common taxonomy, so that Area B-04 in the programme, Level 4 East in the diary and Zone 7 in the ledger are known to be the same place. And retain negative evidence: cancelled work, plant breakdown, low attendance, late design response, successful mitigation. Completeness is more credible than curation.

When claims go wrong

Van Oord: a model cannot outrun its foundations

Judgment In Van Oord UK Ltd v Allseas UK Ltd [2015] EWHC 3074 (TCC) the claimant advanced a claim of roughly £10 million, later reduced. The programme at the centre of the disruption case was described by its own preparer as fundamentally flawed, with no coherent alternative offered. The quantum expert had not verified key components and did not understand how parts had been prepared; the judge disregarded that evidence entirely. The factual evidence also attracted serious criticism, and the recovery was a small fraction of the sum claimed against payments already made on account.

The failure mode is cumulative. An unreliable programme weakens causation. Unverified calculations weaken quantum. Confused or partisan witnesses weaken the court’s confidence in both. A polished report cannot repair foundations the expert never tested.

Lucas Earthmovers: expertise is not independent proof

Judgment In Lucas Earthmovers Pty Ltd v AngloGold Ashanti Australia Ltd [2019] FCA 1049 the claimant’s expert accepted pleaded quantities and summary information without independently measuring the work. Underlying calculations, source material and an adequate witness explanation of the spreadsheet method were missing, and the Federal Court identified fundamental deficiencies in proof.

The lesson is narrower and more useful than "spreadsheets are bad". A summary schedule is persuasive only if its lineage can be shown: source record, calculation rule, reconciliation, competent witness. An expert can test that chain and opine on it. An expert cannot become the missing fact witness by adopting the result.

White Constructions: method labels do not prove causation

Judgment In White Constructions Pty Ltd v PBS Holdings Pty Ltd [2019] NSWSC 1166 the New South Wales Supreme Court found both sides’ delay methodologies unsatisfactory, returned to the factual sequence and common-sense causation, and dismissed the proceedings.

It is a warning against methodological substitution. As-planned versus as-built, windows, time impact: these name analytical structures. They do not establish that a respondent caused a particular delay, that the affected activity was critical, or that alternative causes were excluded. Records of access, design, resources, sequence and decisions remain the evidential substrate.

None of these cases creates a rule that imperfect records defeat every claim, and tribunals can infer facts from incomplete evidence. What they show is that uncertainty is not distributed neutrally. The party carrying the burden is the party exposed when its causal chain rests on assertions nobody can audit.

When records work

Walter Lilly: detailed proof survives complexity

Judgment Walter Lilly & Co Ltd v Mackay [2012] EWHC 1773 (TCC) is usually cited on global claims and concurrency. Its practical lesson is about proof: the contractor relied on its cost system, detailed voluntary particulars and extensive factual evidence, and the court engaged with individual heads rather than treating complexity as a licence for approximation.

It should not be reduced to "good records win". Entitlement and valuation varied issue by issue, and the judgment turns on English law and that contract. It does show that contemporaneous cost systems, explained by credible witnesses and tied to factual events, let a court engage with substance instead of speculating from a top-down difference between bid and actual.

Howard Contracting: a practical programme can still persuade

Judgment In Howard Contracting, Inc v G.A. MacDonald Construction Co, 71 Cal.App.4th 38 (1999), a California appellate court upheld major elements of a delay award where the contractor’s bar-chart evidence was based on critical-path reasoning, identified the controlling sequence and demonstrated critical delays. The court rejected the argument that only a computer-generated CPM schedule could support the analysis.

This is jurisdiction-specific, and it is not permission to abandon schedule discipline. Its enduring point is that substance governs form: a contemporaneous programme that accurately expresses logic and is supported by fact can beat sophisticated software fed unreliable inputs.

Yew San: visual progress records with disciplined context

Judgment The Singapore High Court’s account in Yew San Construction Pte Ltd v Ley Choon Constructions and Engineering Pte Ltd [2019] SGHC 285 describes monthly reports running to dozens of pages: progress by phase and gridline, planned against actual in bar charts, hundreds of activities identified, progress marked on drawings, and photographs selected to show location and viewing direction. A factual witness accepted the reports’ accuracy, which is what permitted the expert to rely on them.

The litigation had mixed outcomes, so it is not a claim won by records alone. It is a model of evidence design. Photographs become analytical evidence when anchored in space, time and scope. A progress percentage becomes testable when connected to quantities and drawings. An expert’s reliance becomes safe when a knowledgeable witness can establish how the record was produced.

Santos v Fluor: strong first-instance findings, appellate caution

Judgment In Santos Ltd v Fluor Australia Pty Ltd [2025] QSC 184 the Queensland Supreme Court adopted referee findings in major LNG-project litigation, and the reasons show the value of transparent costing, source records and a reasoned reconciliation of enormous project data. An appeal followed. As at the research cut-off of 3 August 2026 no public appellate judgment had been identified, with cross-appeals recorded as part-heard in late July 2026.

That qualification is the point. It is strong first-instance evidence of how a court may approach a data-heavy quantum dispute, not a final authority, and it cautions against simple success stories. Large outcomes come from contract construction, liability findings, factual evidence, expert method and recoverable quantum together, not from record keeping alone.

Platforms are infrastructure, not proof

Vendor literature in this space, our own included, promotes costed site diaries, daily resource allocation, photographs, programme visibility and contract-linked variance capture. Those capabilities answer genuine defects in paper diaries and disconnected spreadsheets. They also come with a commercial context, so buy against evidence requirements rather than dashboards.

• Offline, low-friction capture at the workface.
• Mandatory location, activity, resource and quantity fields.
• Timestamp, author, approval and immutable version history.
• Original photographs and metadata retained, not just compressed copies.
• Transparent calculation rules and controlled taxonomies.
• Export in human-readable and analysable formats, without vendor lock-in.
• Role-based access, information security and privacy controls.
• Links to programme, cost, contract and document systems.
• Preservation and legal-hold capability.
• A clear policy on artificial intelligence, including provenance and human verification.

Automated drafting can turn structured facts into a diary narrative or flag omissions. It must not invent causation, classify responsibility without review, or silently change the primary record. The original observation, the machine transformation and the human approval have to remain distinguishable, which is the same rule this site applies to every other use of AI in claim work.

The standard for a serious project

Governance begins with ownership. The project director owns the reliability of the project account. Operations own workface fact. Planning owns schedule integrity and time explanation. Commercial management owns contract process and valuation. Cost control owns ledger reconciliation. Document control owns custody. Legal counsel advises on privilege, disclosure, data protection and strategy, and labelling ordinary project records privileged does not make them so.

Training should focus on observation. A good entry reads: access to grid C5 unavailable 07:10 to 11:40 because stored facade panels occupied the approved crane exclusion zone; eight installers reassigned to C3; two idle for 1.5 hours; removal requested under RFI-184; photograph set EV-0421 attached. A poor entry reads: client delay, lost day, claim to follow. The first can be tested. The second announces a conclusion without its premises.

Six questions belong on a monthly board agenda.

1. Can every material time or cost variance be traced to an event and a source record?
2. Are labour, plant and quantities allocated to work, rather than recorded as attendance?
3. Does every programme movement have a factual narrative and a retained native update?
4. Do the notices and continuing particulars align with what the records actually show?
5. Are contractor-responsible and neutral events recorded with the same discipline as employer-risk events?
6. What value and time exposure currently lacks adequate evidential coverage?

The last one matters most. Evidence risk is manageable while the work is live. Once the site team disperses, the same gap becomes an expert assumption, a disclosure argument, or a discount to settlement value.

The executive conclusion

The failures in Van Oord, Lucas Earthmovers and White Constructions were not caused by a shortage of professional vocabulary. They happened where a court could not safely travel from assertion to source, from method to fact, or from cost to cause. The stronger evidence in Walter Lilly, Howard Contracting, Yew San and the first-instance Santos decision shows the opposite: organised facts let a tribunal test difficult propositions without having to trust the expert because the model looks elaborate.

The standard is easy to state and demanding to meet. Record the fact quickly. Locate it precisely. Measure resources and output. Link it to programme and cost. Preserve its provenance. Record mitigation and competing causes. Then let the later analysis stay faithful to what the project knew at the time.