The engineer report runs forty pages. Page six names the engineer who inspected the roof, and that name appears nowhere in the adjuster notes. An email in the same file says a roofing contractor saw water coming through the attic ceiling. The insured's recorded statement mentions a household member who was home during the storm, and nobody captured that person's name or a phone number.

Three relevant people, three documents, one file that still looks organized. That's the normal state of a claim file, and it's why a witness list from an insurance claim file has to work as an evidence register, not a contact sheet. A contact sheet tells you who to call. An evidence register tells you who a person is, what they may know, which document proves their involvement, when they entered the timeline, and what's still missing.

The sequence below is the one I'd run on any property or auto file headed toward discovery: scan for names, record fields that keep gaps visible, test the result against a hypothetical roster, split fact witnesses from retained experts, link people to documents, then check the whole thing before it supports anything outside the team.

Why Witnesses Hide Across a Claim File

Claims files grow by accretion. The first notice of loss arrives from the agent, the recorded statement comes back from a vendor, the engineer report lands six weeks later, and correspondence with the public adjuster runs in its own thread. Nobody owns the step where names from those documents get merged into one list, so the merge never happens.

Each source also names people differently. The adjuster writes "R. Vela" in a note, the estimate lists a company rather than a person, the engineer report uses a full name with a license number. Matching those three references to one human being is work the file never assigned to anyone.

A name alone can also misstate what the person knows. A household member quoted in a recorded statement may have watched the ceiling collapse, or may only be repeating what the insured said afterward. Until someone separates attributed statements from confirmed firsthand observation, the roster reads cleaner than the file underneath it.

Regulators treat this as part of the record, not an administrative extra. The National Association of Insurance Commissioners guidance on filing an auto claim advises collecting each witness's name and contact information and recording the dates and substance of conversations with adjusters. Market-conduct standards point the same direction: a claim file should be complete enough that pertinent events and dates can be reconstructed after the fact.

That standard is hard to meet when identity, availability, and dates were never captured. If a person can't be tied to a document and a date, they aren't on the roster yet, and the roster is only as good as the scan behind it.

How to Find Every Person Named in the Record

Work the file in a fixed order. Order is what stops names from slipping past, because a scan driven by curiosity skips the attachments that look boring.

1. Build a dated timeline first. Date of loss, notice, inspection, each statement, each payment or denial. Without it you can't tell whether two names are two people or one person at two stages.

2. Scan by document type. First notice of loss, recorded statements, adjuster notes, emails, engineer and contractor reports, photo logs, estimates, invoices, public adjuster correspondence. Notes are where names hide.

3. Pull every name, including the margins. Signature blocks, cc lines, sign-in sheets, "present at inspection" lines, and initials all count.

4. List who should be there but isn't. The household member, the neighbor, whoever reported the loss to the agent.

5. Log negative results. A neighbor the adjuster tried and failed to reach is an entry, dated, not a blank.

6. Merge duplicates. Nickname, company name, misspelled surname, one person.

7. Check each name against the timeline. Every entry needs at least one dated reference.

The document-review discipline that makes this scan reliable is the same one described in this guide to AI document review for P&C claims, where the point is coverage of the whole file rather than spot reading.

Where Names Get Missed Most Often

Three misses recur. Initials-only notes, where "M.H." in the margin is a contractor nobody opened a row for. Names buried inside attachments that were catalogued but never read, such as a signed subcontractor sheet inside a PDF bundle. And names that exist only in a photo caption, where a field photo is labeled with a first name and a date.

A name without a source reference goes on a separate holding list until it has one.

What to Record for Each Witness

A witness roster is an evidence register, not a contact list. Each row records what the file supports, ties every assertion to its source, and leaves gaps written out where a guess would otherwise go. Eight fields carry the entry. The third column matters as much as the second, because the honest answer is often that the file doesn't say.

Attributed Is Not the Same as Witnessed

The source reference field is where the register earns its name. Each entry needs a document name, a date, and a page or Bates range that a reviewer can pull up directly. A phone number in a coverage letter and a phone number in an adjuster note are two different records, so the contact stays tied to whichever document produced it.

Then comes the distinction that gets a roster in trouble. An email saying a contractor "saw water entering the attic" establishes that the email attributes that account to the contractor. It does not prove the contractor personally watched water come through the ceiling. So the entry needs three separate pieces: the statement attributed to the witness, whether direct observation is confirmed, and the basis for that confirmation. A valid basis is a photo timestamp, the witness's own words in a recorded statement, or a record placing the witness on site. Anything less stays marked unconfirmed, and the row still counts.

The household member follows the same logic. Potentially relevant, presence unconfirmed, no statement, no contact details. That row reads not in file, which is an entry, not a blank. It preserves the uncertainty and tells the next reviewer exactly what's missing.

A Sample Witness Roster With Gaps Marked

Same property claim, fictional names, four rows. The table is hypothetical and illustrates the fields above.

Read the rows as four different problems. Reyes is a relevance question with no contact information. Hale is an attribution question with a usable address and nothing confirming observation. Nand is an identity question, since the role field can't be filled without contradicting or confirming the chronology. Ellis is a documentation question: no contact, no CV, and an assumption waiting to happen.

Gordon Ellis's missing CV is the row most likely to be misread. He is named on page six of an engineer report, which makes him a person with technical opinions in the file. Whether he was retained for this claim, or is an employee whose regular duties include inspections, or something else, the file doesn't say. Writing CV not in file and assigning a request keeps that open. Treating his report as background reading and classifying him as a fact witness because no CV appeared is a guess dressed as a conclusion.

Copy these five columns onto the people the scan already found, and enter only what the file supports today.

Fact Witnesses and Retained Experts Need Different Records

A job title doesn't settle the question. A structural engineer can be a fact witness who inspected a roof before suit, a retained expert engaged to opine on causation, or an employee whose regular duties include expert testimony. The title stays the same in all three files. What changes is the engagement, the proof expected from that person, and the calendar pressure attached to disclosure.

Record the basis rather than the label. Four values cover most entries: retained, specially employed, an employee whose regular duties involve expert testimony, or an unretained expert. Rule 26(a)(2)(D) gives that classification dates: absent a court order or stipulation, expert disclosures are due at least 90 days before trial or the date the case is ready for trial, and a disclosure offered solely to rebut another expert is generally due within 30 days after the opposing disclosure. Check both against the governing rules and any scheduling order before building a calendar on them.

Fact witnesses need a leaner record: role, basis for knowledge, firsthand observation separated from statements only attributed to them, contact source, and the gaps still open. Expert rows need more: expected opinions, the claim-file materials those opinions rest on, and whether a CV, report, or summary is present. If the CV is absent, write the absence as an entry, cite the document naming the person, and assign follow-up.

That separation pays off later. Interrogatories often ask for witnesses and the substance of their knowledge; these sample property claim interrogatories show the format those answers have to fit.

Linking People to Documents and Flagging Open Work

Every witness row carries its documents: photographs, engineer reports, repair estimates, emails, recorded statements, invoices, inspection notes. Keep the source reference precise enough for someone else to find the exact item, meaning document name, date, page or Bates range. Mark proposed exhibits as proposed exhibits for review rather than implying they're admissible.

Each linked document does one of two jobs. It either attributes a statement to the person, or it independently confirms something they claim to have observed. Record which job it does. A repair estimate that corroborates a bystander's account belongs in a different field than the recorded interview where the bystander described the loss, and conflating the two is how rosters turn into guesswork.

The Quality Gate Before Anyone Relies on the List

Status is an output. It comes from comparing each account against timestamps, photographs, weather or scene data, repair records, and independent witnesses, and flagging contradictions instead of reconciling them. Treat confidence as a screening signal only. A self-assured witness can still be wrong, so a confident account earns corroboration work, not a shortcut past it. The IRMI commentary on adjusting liability claims recommends documenting a witness's location and any physical limitation that could affect observation, which are exactly the fields that let a reviewer weigh an account later.

Contradictory or unreliable witnesses stay on the roster with their reason and version history attached. An overwritten entry is harder to defend than a flagged one.

Give every open item an owner and a due date, or the people with no contact details, no statement, and no deposition sit unnoticed for months.

The flags also have to move as the file does. Wamy audits the claim file continuously as it evolves and re-evaluates completeness each time a new document is uploaded, so a gap closed by an incoming statement stops appearing as open work and a fresh contradiction surfaces instead of sitting in a static spreadsheet.

Rebuilding the underlying sequence from the documents themselves is covered in this walkthrough on building an insurance claim chronology from documents.

Check Your Roster Against the Claim File Today

Four checks before the roster supports anything outside the team.

• Completeness after reasonable inquiry. Every row traces to a source document someone can pull.

• Consistency with the claim file. No name, date, or role contradicts the adjuster notes or the timeline.

• Duplicate identities. One person, one row, one spelling flagged.

• Privilege separation. Factual account, expert material, and attorney impressions kept in different fields.

An internal roster is not automatically a compliant disclosure and not a final trial witness list. Any procedural requirement needs current primary authority and an identified jurisdiction. What the roster does offer is a clean mapping step. Under Rule 26(a)(1), a party identifies each individual it may use to support a claim or defense along with the subjects of that person's discoverable information, which are the name and subjects-of-knowledge fields already built. Expert rows feed the separate expert disclosure schedule.

The Federal Rules of Civil Procedure as published by the U.S. government also matter for production questions: a person may obtain their own previous statement about the action without the heightened showing normally required for trial-preparation materials, and protective orders remain available where disclosure creates privacy concerns, undue burden, or expense. That rule is the reason to classify each entry as identity and contact information, factual witness account, expert or consultant material, or attorney impressions. Separating those four categories lets a carrier, TPA, or subrogation counsel answer discovery without producing protected analysis by accident.

From there, the same entries serve witness-list drafts, expert designations built on the material in the file, interrogatory answers, deposition outlines, and mediation preparation. Wamy builds that roster from the people already named in the file, recording name, company, role, retaining party, and timeframe; it separates fact witnesses from retained experts and flags missing contacts, statements, or depositions. When new evidence lands, it folds that into the claim analysis and creates a notification task for the relevant parties with the updated findings.

Before a demand, a disclosure, or a deposition outline goes out, the file has a habit of reminding you what was never collected. The pre-suit property claim checklist covers the rest of that file-side groundwork.

Wamy: Connect the People in Your Claim File to the Evidence

Wamy builds a cast of characters and witness roster from people identified in the claim record. It captures names, companies, roles, retaining parties, and involvement periods, connects individuals to relevant documents, and flags missing contact information, statements, or depositions.

Your team can use that source-linked record to prepare witness-list drafts, interrogatory answers, deposition outlines, and expert designations using the material available in the file. Uncertain identities and conflicting accounts remain questions for review.

Book a demo to see how Wamy turns references scattered across reports, emails, and statements into a witness roster your team can verify and use.