How an Engagement Works

The conflict check runs before a single case fact moves

A first message should contain names and nothing else. The parties, including parent companies, subsidiaries and trading names. Counsel of record on all sides. The agencies, platforms, publishers and vendors likely to appear in the matter. Any non-party whose conduct is in issue. Not the story, not the theory, and not a summary of what the other side did.

The reason is structural rather than fussy. Whether an expert can be disqualified generally turns on two questions: whether the party claiming to have retained him could reasonably conclude that a confidential relationship existed, and whether confidential or privileged information actually passed. A friendly exploratory call in which counsel describes the matter can satisfy both halves before anything is signed. That is how an expert becomes unavailable to a side that never retained him, and how a side that meant to retain him loses him to a motion. The order of operations is the protection and it costs nothing to observe.

I run the check against past and current engagements and against commercial relationships as well as litigation ones. Experts in this field frequently have live relationships with the platforms and agencies whose records are in issue — program access, partnerships, speaking arrangements. Those are rarely disqualifying, but they are disclosable, and an expert who does not raise them before retention has handed the other side a free cross-examination opening. If the check does not clear, I say so, I do not explain why, and I delete what was sent.

Retention, and the terms that are discoverable by rule

Retention is normally by counsel rather than by the client directly. The engagement letter fixes six things: the retaining firm and the matter; whether the role is consulting or testifying, and the mechanism for converting one into the other, since that distinction decides what is discoverable; the scope; the rate structure, including deposition and trial; how productions are received, stored and disposed of at the end of the matter; and the position on the use of artificial intelligence in the work, stated at the outset rather than discovered in a deposition.

Two of those are worth naming because counsel occasionally assumes otherwise. Compensation is discoverable by rule — a retained expert’s report must state the compensation to be paid for the study and testimony, and communications about compensation are expressly carved out of the protection that otherwise covers attorney-expert exchanges. Write the letter as though opposing counsel will read it, because they will. And this practice does not accept fees contingent on outcome, in any structure, in any matter.

Scope, written so the answer is not built into the question

A scope written as a conclusion damages the engagement before any records arrive. Determine whether the agency overbilled the client is a conclusion in search of support, and it is the sentence a competent cross-examiner reads aloud from the engagement letter. Reconcile agency-invoiced spend against platform-recorded spend for the period stated, identify every divergence, and characterize the divergences by cause is an assignment. The second version produces the same analysis and can be defended.

Scope is also provisional in a way that surprises attorneys used to other kinds of expert work, because what the records can support is not known until the records are seen. It is common for a scope to narrow once the production is inventoried — a question that assumed click-level detail cannot be answered when the click-level window closed eleven months before the complaint. I would rather narrow a scope in writing in week three than defend an over-broad one at deposition, and a narrowed scope stated in the engagement file is a document, not a concession.

What to send first, and the order that costs least

Once the check has cleared and terms are agreed, this order lets useful work start at the lowest cost:

1. The operative pleadings, so the analysis can be aimed at the claims it has to serve.
2. The contract, every statement of work, and every amendment. Without these, no scope or standard-of-care analysis is possible at all.
3. The scheduling order, with the expert disclosure and rebuttal dates.
4. The protective order, before any confidential production moves.
5. What has actually been produced, in native format with metadata intact, plus the production log.
6. The other side’s expert report, if this is a rebuttal engagement, together with the reliance materials that expert was required to identify.
7. Any preservation letters and litigation-hold notices touching ad accounts, analytics properties and marketplace seller accounts.

Native format matters more here than in most kinds of expert work. An export converted to PDF has usually lost the field names, the row-level granularity, the time zone and the date range that make a figure reproducible — and a PDF of a dashboard is a screenshot of a query that nobody can re-run. If the production contains only PDFs of exports, that is itself a discovery issue worth raising early.

The evidence review begins with what still exists

The first substantive question in this field is not what happened. It is what survives. Advertising and analytics records are generated, held and deleted by platforms on published schedules that run whether or not a complaint has been filed and whether or not a hold letter has gone out, and no major advertising or analytics product exposes a control that suspends deletion. A preservation instruction binds the parties and their own systems; it has no effect on a third party’s retention schedule.

So the review opens with an inventory rather than an opinion: which accounts and properties exist, who owns each, which manager or business account they sit under, who was granted what level of access and when, which payment profile actually paid, and what the retention position is for every record type in issue. That inventory drives the discovery requests, because a request drafted against the wrong object returns nothing and burns weeks that the retention windows do not give back.

The first thing counsel receives is therefore usually a memorandum, not an opinion: what exists, what is missing, what has already been lost, what should be requested and from whom, and what should be exported immediately by any party in a position to export it. In a matter caught early that memorandum is frequently worth more than the report, because it changes the requests while there is still time to serve them.

The analysis, and the step that is skipped most often

The core of most engagements is reconciliation: putting two or three independently held records of the same events side by side and accounting for every difference. What the platform billed. What the platform’s own reporting says was spent, by campaign and by day. What the client was invoiced, and what the client’s monthly report said had happened.

Divergences fall into three classes and only one of them is a finding. Definitional differences come from counting different things: conversions against all conversions, one attribution model against another, a different conversion window, viewable impressions against served impressions, account time zone against reporting time zone. Timing differences come from records that restate — conversion lag, invalid-activity credits posted after the fact, modeled figures that fill in later, a report run on the first of the month against the same report run on the thirtieth. Unexplained is what is left after the first two have been eliminated.

The discipline that makes an opinion survivable is eliminating the first two classes in writing, with the definitions and the dates stated, before characterizing anything as the third. An opinion that skips that step is the opinion a rebuttal expert takes apart in an afternoon. Everything is documented so that a competent expert on the other side can reproduce it: the export used, the date range, the time zone, the attribution model and conversion window in force, the filters applied, and the date the record was pulled.

What comes back to counsel, and in what form

Work product moves in stages: working papers and reconciliations, then a memorandum to counsel setting out what the records support and what they do not, then — if the matter proceeds and I am designated — a report.

Drafts of a required report are protected regardless of the form in which the draft is recorded, which covers a marked-up document, a shared file and a version in a repository. Communications between counsel and a testifying expert are protected too, with three exceptions: compensation; facts or data counsel provided that the expert considered; and assumptions counsel provided that the expert relied on. Because those three categories are discoverable, the clean practice is to keep counsel-supplied facts, data and assumptions as an identifiable, dated list from the first day of the engagement. That turns a discoverable category into a document rather than an excavation, and it happens to make the report’s statement of facts considered straightforward to write.

The report, and the rule that governs its contents

A retained expert’s written report must contain a complete statement of all opinions and the basis and reasons for them; the facts or data considered in forming them; any exhibits used to summarize or support them; the witness’s qualifications, including a list of all publications authored in the previous ten years; a list of all other matters in which the witness testified as an expert at trial or by deposition in the previous four years; and a statement of the compensation to be paid for the study and testimony. It is prepared and signed by the witness.

One phrase in that list does more work than it looks like it does. Considered is broader than relied on: material examined and set aside is still considered. In a marketing matter where the production runs to many gigabytes of exports, the report has to identify what was considered with enough specificity that the other side can reproduce the analysis from the same files. Vague reliance statements are where an otherwise sound report starts losing ground.

Admissibility is deliberately not this site’s subject. Rule 702, as amended effective 1 December 2023, requires the proponent to establish the rule’s requirements by a preponderance and requires that the opinion reflect a reliable application of the method to the facts; a separate practice site covers that standard and the case law around it at length. The practical consequence for this work fits in one sentence: do not claim more than the method produced.

Rebuttal engagements, and the deadline that usually is not the rule’s

A large share of the work in this field is rebuttal, and the procedural posture is the first thing to check. The federal default for rebuttal disclosure is thirty days after the other party’s disclosure, and it reaches only evidence intended solely to contradict or rebut evidence on the same subject matter — but the default applies only absent a stipulation or a court order, and in practice the scheduling order nearly always sets its own expert and rebuttal dates. Read the order, not the rule.

A retained rebuttal witness still files a full report. Nothing in the rule creates a lighter one for a witness who is responding rather than opening: the publication list, the testimony list and the compensation statement all apply identically. And the constraint that keeps a rebuttal inside its scope is the one it is easiest to breach — saying and therefore the correct figure is this, when nobody asked for a figure, converts a rebuttal into an affirmative opinion that may be both untimely and outside the subject matter. Stating that an analysis does not support the conclusion drawn from it, and showing why, is a complete opinion on its own.

Deposition, trial, and what the process does not deliver

At deposition the questions in this field are predictable in shape: where each number came from, whether it was counted or estimated, what settings produced it, what was not examined, and what the analysis does not establish. A file kept as described above answers all five without improvisation. The party seeking that discovery ordinarily pays a reasonable fee for the time spent responding to it. At trial the job is to explain a technical record to people who do not work in it, using exhibits that a juror can follow and that a cross-examiner cannot show to be selective.

What the process does not deliver is worth stating as plainly as what it does. It does not deliver a predicted outcome, and no engagement here is structured on one. It does not deliver an opinion specified in advance by the retaining party; if the records do not support the theory, that is what the memorandum will say, and it is better learned before a report than after one. It cannot recover expired data. It does not resolve intent, because no configuration record contains a motive. And it does not answer the legal questions — duty, breach, materiality, and what the matter is worth — which belong to counsel, to the trier of fact, and in quantum disputes usually to an economist as well.