Engaging an Expert

The conflict check, and what to send for it

Nothing substantive should move before the conflict check clears. Send identifying information only: every party and its affiliates and parents, counsel of record on both sides, the jurisdiction and posture, and — the part specific to search matters — the websites and domain names at issue.

That last item is why a search expert's conflict check is not a formality. Conflicts here arise from the practice, not only from the caption. An expert who still takes client work may have audited the plaintiff's site, served a competitor of the defendant, or worked on the migration now being litigated. None of that shows up in the party names. It shows up when you list the domains. Keep that first message free of anything privileged or protected; there is no engagement yet, so nothing for a protection to attach to.

Scope the question before you scope the engagement

The most consequential decision in an engagement is the framing of the question, and in my experience opinions in this field fail far more often on scope than on data. A broad question obliges the expert to defend every link in a long chain, and a chain fails at its weakest link under cross-examination. A narrow question is harder to attack because each step has data behind it. Compare:

• Broad: did the defendant's conduct cause the plaintiff's decline in organic traffic?
• Narrow: did the migration on a stated date break the redirect map — the list mapping each old URL to the new URL replacing it — and is the decline consistent in timing, page-level distribution, and query pattern with that break rather than with the core algorithm update in the same quarter?

The second question is testable. It has a date, a mechanism, and a named alternative explanation that gets ruled in or out on the evidence. It also has an answer that might be no, which is the feature and not the defect.

Scoping also decides what is not the search expert's question. Lost profits and but-for revenue are the damages economist's work, and a search expert who volunteers a revenue figure has stepped outside the discipline in which he is qualified.

Consulting first, testifying later — the conversion runs one way

Retain the expert in a consulting capacity first. Designate as testifying later, if and when the analysis supports an opinion worth disclosing.

The reason is in Rule 26(b)(4)(D). Ordinarily a party may not discover facts known or opinions held by an expert retained in anticipation of litigation who is not expected to be called at trial, absent Rule 35(b) or exceptional circumstances. A testifying expert's position is the opposite: the report is disclosed, the facts or data considered are disclosed with it, and only the narrower protections of Rule 26(b)(4)(B) for drafts and Rule 26(b)(4)(C) for attorney-expert communications remain — subject to the carve-outs in (C) for compensation, for facts or data the attorney supplied and the expert considered, and for assumptions the attorney supplied and the expert relied on.

What that buys you early is the freedom to hear an answer you did not want: that the decline began weeks before the conduct complained of, or that the drop is a broken tracking tag rather than a ranking loss.

The conversion runs one way. A consultant can be promoted to a testifying witness. Material the witness has considered cannot be un-considered, and the consulting protection cannot be restored once the designation is made.

What belongs in the engagement letter

The engagement should run between the expert and retaining counsel rather than with the party directly, which is the posture the work-product and communication protections are built around. The letter is where later arguments are avoided:

• The matter and the parties, identified precisely enough to define the conflict position going forward.
• The assignment stated as a question, not as a conclusion to be supported.
• The fee structure and what is billable: analysis, data acquisition, report preparation, deposition and trial time, travel, and deposition preparation, including who bills whom for deposition time.
• Confidentiality and protective order compliance, including how produced material is stored, retained, and returned or destroyed.
• The disclosure obligations acknowledged, so the four-year testimony list and the ten-year publication list are supplied at retention rather than at the deadline.
• Conditions of withdrawal, including what happens if the record will not support an opinion.

What actually drives cost, and who pays for the deposition

Cost here is not driven by the size of the claim. A large demand over one landing page can take a fraction of the hours a modest dispute takes across four years of an e-commerce catalog. What drives it is the shape of the record.

• Data volume. Access logs are the extreme case — tens or hundreds of gigabytes of raw requests to be parsed, filtered to search engine crawler traffic, and validated before a finding comes out of them.
• The number of URLs. A 40-page brochure site and a 900,000-URL catalog raise the same questions and are not the same job. Crawling, comparison across dates, and page-level attribution all scale with it.
• The length of the period at issue. Every added quarter brings algorithm updates to rule in or out, seasonality to control for, and site changes that are themselves candidate explanations.
• How much of the record survives. The largest and least predictable driver. Search Console data runs on a rolling sixteen-month window, logs live on the host's retention schedule, and analytics properties get deleted with the employee who owned them. Once the direct evidence is gone the work becomes reconstruction and inference — slower, weaker, and more expensive than reading a preserved export.
• The alternative explanations to be excluded. Causation opinions are priced by the confounds, not the conclusion.

One allocation is set by rule rather than by negotiation. Under Rule 26(b)(4)(E), 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 Rule 26(b)(4)(A) or (D). In practice the party taking the expert's deposition ordinarily pays the expert's reasonable fee for that time. Whether preparation time falls within it is treated differently across courts, which is why it belongs in the engagement letter.

The deadlines you are working backward from

Rule 26(a)(2)(D) sets the default. Absent a stipulation or a court order, expert disclosures must be made at least 90 days before the date set for trial or the date the case is to be ready for trial. Where the evidence is intended solely to contradict or rebut evidence on the same subject matter identified by another party, the disclosure is due within 30 days after the other party's disclosure. In practice a scheduling order displaces the default, and the order governs.

The 90-day figure is not a 90-day project window. Work backward: counsel's review of the draft, the report, the analysis, the data acquisition, and — first — the conflict check. Data acquisition depends on other people, and it is the step that overruns.

The 30-day rebuttal window is the harder constraint and the one most often underestimated. Thirty days to obtain the opposing report and its reliance materials, reproduce what that expert did, identify what they did not do, and write a defensible rebuttal is not generous. Where a rebuttal is foreseeable, retain the expert in a consulting capacity before the opposing report lands.

What to send, and in what form

Send data in the format it was produced in — native exports with their date ranges intact, not screenshots or PDFs of dashboards. A screenshot of an analytics chart cannot be re-run, re-segmented, or checked; a CSV export can be all three, which is what makes it usable as the basis of an opinion. What most often matters:

• Google Search Console exports — the reporting property for a verified site, showing the queries, impressions, clicks, and average positions Google recorded — with both the date range requested and the date the pull was made.
• Analytics exports at session and landing-page level, beginning well before the conduct.
• Raw server or CDN access logs, unfiltered.
• The URL inventory and the redirect map, where a migration is involved, with the deployment and CMS records establishing what changed and when.
• Contracts, statements of work, and correspondence, where the standard of care is at issue.
• The pleadings, relevant transcripts, and the opposing report with its reliance materials.

Read-only access to the client's own Search Console and analytics properties beats exports, because it lets a pull be re-run and a filter tested rather than taken on faith. Get the preservation letter out before the rolling windows discard the months you need. And stop treating third-party rank tracking as a substitute for first-party measurement: a rank tracker's history is an estimate from somebody else's sampling method, and an opinion resting on it alone is the weakest in this field.

A realistic timeline

Stated structurally, because every matter's data differs:

1. Conflict check — days, not weeks, if the identifying information arrives complete.
2. Consulting review of the record — what exists, what is missing, whether the theory is supportable at all. This phase saves money, because it is where a case that is not there gets identified.
3. Data acquisition — the long pole, and the one nobody controls. It depends on the client's IT, the host's log retention, third-party platforms, and sometimes on discovery from the other side.
4. Analysis — scales with the drivers described above.
5. Report drafting and counsel's review, ending at the disclosure date rather than beginning near it.
6. Deposition, then trial.

The single most common problem is late retention. An expert brought in after discovery has closed inherits whatever was requested by people who did not yet know what a search analysis needs — and the logs, historical exports, and crawl-level data that would have answered the question were often never requested. By then the honest opinion may be that the record cannot support one.

For the rules cited here, see Federal Rule of Civil Procedure 26.