Expert Drafts and Attorney Communications

What the draft rule protects, and the phrase that does the work

Rule 26(b)(4)(B), in the form adopted in the 2010 amendments, is short: Rules 26(b)(3)(A) and (B) “protect drafts of any report or disclosure required under Rule 26(a)(2), regardless of the form in which the draft is recorded.” The text is on Cornell's copy of Rule 26.

The operative clause is the last one. “Regardless of the form in which the draft is recorded” reaches well past a file named report_v3.docx. It reaches a draft section pasted into the body of an email, a version circulated with tracked changes and margin comments, a shared document with a revision history the platform keeps automatically, and a paragraph worked out in a chat thread. Protection follows the content, not the container. That was the point of the amendment: before it, an expert's editing history was a standing discovery target, and the result was that experts and counsel stopped putting anything in writing, which served nobody.

Note what the provision does not do. It protects drafts of the report. It does not protect the underlying data, and it does not create a privilege over facts. Rule 26(a)(2)(B)(ii) still requires disclosure of the facts or data considered, and an analysis that happens to live inside a draft is still an analysis whose inputs are disclosable. Communications with counsel are a separate provision with exceptions of their own, and those exceptions are where the real work is.

The three carve-outs are the whole ballgame

Rule 26(b)(4)(C) protects communications between the party's attorney and any witness required to provide a report, regardless of the form of the communications, except to the extent that the communications:

1. (i) relate to compensation for the expert's study or testimony;
2. (ii) identify facts or data that the party's attorney provided and that the expert considered in forming the opinions to be expressed; or
3. (iii) identify assumptions that the party's attorney provided and that the expert relied on in forming the opinions.

Read those three as a running checklist against your own file rather than as a footnote to the protection. The general rule is broad enough that it is easy to stop reading at “protect communications,” and that is the reading that causes trouble, because the exceptions are not narrow technicalities. They are the three categories most likely to exist in a real engagement.

Carve-out (i) is straightforward and largely duplicated by Rule 26(a)(2)(B)(vi), which puts compensation in the report anyway. Carve-outs (ii) and (iii) are the interesting ones, because they are created by ordinary, well-meant conduct — usually in the first week, usually by an attorney trying to be helpful.

The spreadsheet problem

Here is the fact pattern, and some version of it recurs in nearly every technical engagement I have seen. Counsel forwards a workbook: here are the traffic numbers. Somebody at the client pulled them. There is a tab per year, a column of monthly totals, and no record of which property they came from, which view or data stream, which date range, which filters, whether internal traffic was excluded, or whether the figures are sessions, users, or pageviews.

The moment the expert considers that workbook it is a Rule 26(b)(4)(C)(ii) item: facts or data the attorney provided and the expert considered. The communication conveying it loses protection to that extent. The workbook is produced, the transmittal is produced, and at deposition the expert is asked where the numbers came from — a question the expert cannot answer, because the expert did not pull them and does not know how they were built. The follow-up, which is the question that does the damage, is whether the expert verified them.

The clean practice is the same practice that makes the opinion defensible. Pull the data yourself from the first-party source. Record the property, the date range, the filters, the segment, and the date of the pull in the work log. Treat the client's own figures as a cross-check to be reconciled rather than an input to be adopted, and where the two disagree, say so in the report and explain why. Where the analysis genuinely has to begin from a number the expert cannot pull — a revenue-per-session figure held by the client's finance team, an internal margin — do not fold it in silently. Label it.

Assumptions, labeled as assumptions

Carve-out (iii) reaches assumptions the attorney provided that the expert relied on. The instinct on first reading is to minimize the number of items falling into that category. That instinct is wrong, and acting on it produces a weaker opinion as well as a worse discovery position.

An opinion in a search matter almost always rests on premises somebody else supplies: the date the disputed conduct began, the boundary of what the defendant controlled, which URLs belong in the disputed set, the period the claim covers, the characterization of a contractual relationship. An analyst cannot determine any of those from crawl data. Counsel can, and frequently must.

State each premise on the face of the report, in its own identified place, as an assumption, with its source. That costs nothing that was ever protected — the rule has already carved it out — and it buys three things. The opinion becomes conditional in a way that is honest and testable. The assumption becomes a variable that can be changed on cross-examination without the opinion collapsing, because the reader can see which conclusions move when it moves. And the expert is never walked into presenting somebody else's premise as their own finding, which is the most damaging moment available in this kind of examination.

The same logic governs the interface with a financial expert. Where a damages expert takes the traffic counterfactual from the search expert, and the search expert takes a conversion value from the damages expert, each is an assumption with a stated source, and each expert should be able to say precisely which inputs came from the other. That is sound division of labor and it is also a (C)(iii) question that will be asked.

Protection is not the same as invisibility

Two consequences follow from how these provisions fit together, and both get missed.

First, protected does not mean unexamined. Drafts and communications are protected; the expert's opinions, methods, data, reasoning, and history of reasoning are the entire subject of the deposition. If a draft contained an opinion the expert later abandoned, the draft is protected — but the expert can still be asked whether they ever held a different view, when it changed, what changed it, and whether anyone suggested the change. The rule protects the document. It does not suppress the question, and an expert who believes otherwise will answer badly.

Second, form is not a hiding place, in either direction. Because draft protection follows content regardless of the form in which the draft is recorded, moving report language into a chat thread does not strip it of protection — but it also does not remove it from the file, and it makes the eventual sorting harder. Keep report drafting in the drafting environment, data work in the data record, and correspondence with counsel in correspondence. Where a specific item sits close to the line, the expert's job is to flag it and counsel's job is to decide it; characterizing what is protected is a legal judgment.

Deposition sequencing, and who pays for it

Rule 26(b)(4)(A) permits a party to depose any person identified as an expert whose opinions may be presented at trial, and provides that where Rule 26(a)(2)(B) requires a report, the deposition may be conducted only after the report is provided. The document comes first and the examination proceeds from it, which is why a loosely written report is expensive later: every ambiguity becomes an hour of questioning.

Rule 26(b)(4)(E) is the provision that appears on almost no page written for attorneys and that first-time retaining counsel always want to know about. Unless manifest injustice would result, the court must require the party seeking discovery to pay the expert a reasonable fee for time spent responding to discovery under (A) or (D). The deposing party ordinarily pays for the expert's deposition time. The obligation is framed as mandatory, subject to the manifest-injustice escape and to the court's view of what is reasonable, and disputes here are usually about the rate or the hours rather than the principle.

That has a budgeting consequence on both sides. Deposing a search expert properly means walking through crawl configurations, export parameters, date ranges, filters, and the treatment of confounds, and none of it is quick. The party asking the questions ordinarily bears the cost of the answers, which is a reason to plan the examination rather than fish through the file on the record.

The file hygiene this all adds up to

None of the above requires a system. It requires six habits, kept from the first day of an engagement rather than adopted the week before a report is due:

• Pull first-party data yourself, and log every pull with its parameters and its date.
• Keep counsel-supplied material in a separate, identified place, so that a (C)(ii) production is a filter rather than a search.
• State every attorney-supplied assumption on the face of the report, with its source.
• Draft in the drafting environment, not across email threads and chat.
• Do not treat draft protection as a license to write carelessly, because an abandoned view is still a view you can be asked about.
• Keep the compensation record current, since both (C)(i) and Rule 26(a)(2)(B)(vi) reach it.

The through line is that the discovery rules and the reliability rules ask for the same behavior from different directions. An expert who pulls their own data, documents the pull, and labels every borrowed premise has produced an opinion resting on facts they can vouch for and a file that sorts cleanly when someone asks for it. An expert who accepts a spreadsheet, adopts a date, and assumes a scope has produced an opinion resting on premises they cannot defend and a file in which the borrowed parts are indistinguishable from the developed ones. The second expert is rarely caught out on the law. They are caught out on the third question of the deposition.

Frequently Asked Questions

Published 2026-08-11