Documentation that wins disputes

Every lesson in this course has come back to records. Timely notice is proved by a dated letter. Entitlement is proved by the contract, the drawings and the correspondence. Delay is proved by schedule updates that were maintained honestly at the time. Disruption is proved by quantities and hours recorded by activity. Quantum is proved by cost records. Negotiators, adjudicators, arbitrators and judges all decide on the evidence, and in construction the evidence is mostly the project's own records. The party with better records usually wins, or settles on better terms.

This final lesson sets out what records to keep, how to keep them so they are believed, how to preserve them when a dispute arises and how to avoid creating documents that hurt. It ends with a capstone that follows the Ashford truss claim from the day the problem appeared to its resolution, drawing together the methods of the whole course.

Contemporaneous records

A contemporaneous record is one made at the time of the event by someone with knowledge of it, in the ordinary course of business. Tribunals give such records much more weight than recollections or reconstructions prepared later for a claim, because they were made before anyone knew they would matter. A daily report written on June 18, 2025 saying 'crane down, hydraulic failure, no trusses on site to lift' is worth more than any witness statement written a year later.

Quality matters as much as quantity. A daily report that says only 'work progressing' every day is almost useless. One that names the crews and their sizes, the areas they worked in, what stopped them and what they were waiting for is evidence. Training superintendents and foremen to write factual, specific reports, and checking them weekly, is one of the cheapest forms of claim protection available.

Auditing records before they are needed

A contractor that intends to claim should check its records while the project is still running, when gaps can still be filled by other evidence such as photographs, emails or the owner's own records. An audit compares what exists with what will be needed.

# Audit of Kestrel Ridge's daily reports for the truss redesign window (seeded illustrative data)
import random
from datetime import date, timedelta
random.seed(42)
start = date(2025, 6, 2)
days = [start + timedelta(days=i) for i in range(70) if (start + timedelta(days=i)).weekday() < 5]
reports = {}
for d in days:
    if random.random() < 0.9:
        reports[d] = {"crew": random.random() < 0.85, "areas": random.random() < 0.8,
                      "delays": random.random() < 0.7, "photos": random.random() < 0.75}
print(f"Working days in window: {len(days)}; daily reports found: {len(reports)} ({len(reports)/len(days):.0%})")
for field in ("crew", "areas", "delays", "photos"):
    n = sum(r[field] for r in reports.values())
    print(f"  reports recording {field:<7} {n:>3} ({n/len(days):.0%} of working days)")
gaps, run = [], []
for d in days:
    if d not in reports or not reports[d]["delays"]:
        run.append(d)
    else:
        if len(run) >= 2: gaps.append((run[0], run[-1], len(run)))
        run = []
if len(run) >= 2: gaps.append((run[0], run[-1], len(run)))
print("Runs of 2+ days with no delay record:")
for a, b, n in gaps:
    print(f"  {a} to {b} ({n} working days)")

Working days in window: 50; daily reports found: 44 (88%)
  reports recording crew     37 (74% of working days)
  reports recording areas    39 (78% of working days)
  reports recording delays   31 (62% of working days)
  reports recording photos   30 (60% of working days)
Runs of 2+ days with no delay record:
  2025-06-09 to 2025-06-11 (3 working days)
  2025-06-18 to 2025-06-19 (2 working days)
  2025-07-03 to 2025-07-07 (3 working days)
  2025-07-11 to 2025-07-14 (2 working days)

The audit shows that daily reports exist for 88 % of working days in the window, but only 62 % record delays and only 60 % have photographs. One of the gaps, June 18 to 19, falls inside the crane breakdown, which is exactly the period where Kestrel Ridge needs to prove that no trusses were available to lift (lesson 5). Kestrel Ridge should now look for other evidence for those days, such as delivery records showing no trusses arrived, emails with the steel fabricator and the city inspector's own reports, and should tighten its reporting for the rest of the job.

Making records credible

Records are believed when they look like normal business records made at the time. Several practices help.

• Consistency: the same format, completed every day by the same role, signed or logged in the system with a timestamp.
• Facts, not argument: 'Tile crew of 8 could not start natatorium deck; mechanical crew of 6 occupying area' rather than 'Owner's incompetent scheduling caused chaos again.'
• Corrections in the open: a mistaken record is corrected by a dated note, not by editing the original. Electronic systems should keep audit trails.
• Cross-checks: daily reports, payroll, delivery tickets and photographs should agree. Inconsistencies invite challenge.
• Native files: schedules should be kept in their native format with every update, so experts can test the logic; PDF prints are not enough.

Records created after the event for a claim, such as 'as-built' schedules reconstructed months later, are not worthless, but they need support from contemporaneous sources, and their author must be ready to explain how they were prepared.

Documents that hurt

Projects also create records that damage their authors. Internal emails speculating that 'we underbid the concrete badly', a superintendent's text joking about the crew's slowness, or a project manager's note that 'the owner has a point on the formwork' are all likely to be produced in a dispute. The answer is not to stop writing, and certainly not to destroy documents, but to write internal communications as if the other side will read them, keep speculation out of email, and route sensitive assessments through counsel where privilege may apply.

In the US, attorney-client privilege protects confidential communications between a client and its lawyer for the purpose of legal advice, and the work-product doctrine protects materials prepared in anticipation of litigation. Neither protects underlying facts, and copying a lawyer on a business email does not make it privileged. Claim consultants' work may or may not be protected depending on how they are retained and for what purpose.

Preserving records when a dispute arises

Once litigation is reasonably anticipated, parties have a duty to preserve relevant documents, including electronically stored information such as email, text messages and schedule files. Federal Rule of Civil Procedure 37(e) allows courts to order measures to cure prejudice from lost electronic information and, where a party acted with intent to deprive, to instruct the jury that the lost information was unfavorable or to enter judgment. A party should issue a written litigation hold to everyone who may have relevant records, suspend routine deletion and preserve phones and devices of key people.

Project records must also survive the project. Contracts and statutes require records to be kept for set periods, often several years after completion, and limitation periods for latent defects can be longer. Records should be archived in a searchable form with the native files intact, not left on a departing project manager's laptop.

Capstone: the Ashford truss claim from event to resolution

The truss claim has appeared in almost every lesson. Following it from start to finish shows how the pieces fit.

Several features of this story are worth drawing out. The claim succeeded because notice was given on time; without it, the best schedule analysis in the world might not have mattered. The schedule analysis was credible because the updates had been kept honestly through the project, including the contractor's own two days of delay. The quantum was accepted largely because it conceded what it should and avoided double counting. The settlement worked because both sides had valued their alternatives and found that a deal was better. And the whole process took about nine months, not five years, because the parties used the cheaper rungs of the ladder before the expensive ones.