Building a Damages Model That Survives Cross

Two experts, one seam, and the reason to keep them apart

The most durable structural decision in a marketing damages case is made before any number is computed: deciding which expert supplies which half.

The digital marketing expert supplies the record and the mechanism. What was configured, when, and under which login — account structure, targeting, budgets, bid strategy, conversion actions, the attribution model in force, tag deployment, change history. What ran — impressions, clicks, spend, placements, creative, keyword and search-term data, landing pages. What a user actually saw and in what sequence. What the platform's own records do and do not contain, and the retention and definitional limits on them. Whether the observed pattern is consistent or inconsistent with the alleged cause, which alternative explanations exist, and whether the data rule them out.

The damages or accounting expert supplies the but-for financial model, margins, cost structure, fixed versus variable treatment, discounting, present value, mitigation, and the disgorgement computation.

The seam is apportionment. The marketing expert says which explanations the data can and cannot exclude; the damages expert converts that into a number. A report that silently crosses the seam invites a motion under FRE 702, and the site's own credibility is better served by naming the boundary than by hiding it.

The but-for construction, and the isolation requirement it carries

Every damages theory here is a comparison between the world that happened and a world that did not. The Reference Manual on Scientific Evidence, 4th ed. (2025), frames it as requiring "a clear statement of what occurred in the actual world" plus "a description of the but-for world, including the defendant's proper actions in place of their unlawful actions," with the plaintiff's position assessed assuming the harmful act had not occurred.

The obligation that comes with it is the one models fail. The measurement must "isolate the change in the plaintiff's economic position caused by the harmful act and exclude any change … arising from other causes." The Guide notes that disputes commonly turn on whether the model adequately controls for confounding factors, and that a defendant will frequently build its own calculation on the premise that the losses would have occurred anyway.

In a marketing dispute the but-for world is often a hypothetical campaign that never ran, which is where this work is most exposed. An expert can testify to whether the work performed conformed to the contract's scope, to the platform's own published rules, and to objectively verifiable facts of execution. Asserting the revenue a different agency would have produced is a counterfactual claim with no counterfactual behind it, and it does not survive a determined cross.

The categories a court sees, and which the record can reach

Five categories account for most of what is claimed in these disputes, and they make very different demands on the evidence.

• Lost profits — the profit the plaintiff would have earned but for the conduct. Requires the full but-for construction and isolation of the defendant's contribution from every other cause. The most demanding category and the one most often overreached.
• Reliance and out-of-pocket — amounts paid out in reliance: agency fees, wasted media spend, remediation cost. Available without proving what sales would have been.
• Disgorgement of the defendant's profits under the Lanham Act, 15 U.S.C. § 1117(a).
• Corrective advertising — the cost of advertising to repair damage to a mark or a reputation.
• Statutory damages — available for counterfeiting under § 1117(c) and cybersquatting under § 1117(d), and not available for ordinary false advertising or infringement. That limitation is worth confirming early, because a case built on the assumption of a statutory floor changes shape without one.

Section 1117(a) also permits a court, subject to the principles of equity, to enter judgment for any sum above actual damages "not exceeding three times such amount," with the statute stating that such a sum "shall constitute compensation and not a penalty," and permits fees "in exceptional cases." The statutory text is short and worth reading directly.

Out-of-pocket is the category the record usually supports

In an agency or vendor dispute, the strongest damages theory is frequently the least ambitious one, because it does not require a but-for sales figure at all. What the client paid is a documented quantity. What the platform charged is a separately documented quantity, held by a party that controls neither side of the dispute.

The reconciliation compares three things for each period: what the platform charged, from its own billing record; what the platform's reporting interface says was spent, exported from the account; and what the agency invoiced the client, alongside what the agency's own report told the client had been spent. The comparison is objective, is rarely done before litigation, and produces a number rather than a judgment about quality — which is why it survives cross better than most of what gets offered.

The discipline is in the exclusions. A gap between the platform's billing record and the interface's cost column is usually benign and explainable: credits and refunds, invalid-activity credits that appear on the billing record but not in the cost column for the original day, currency conversion, tax and regulatory operating costs, time zone and billing-period boundaries, accrual against invoice timing. An expert who treats such a gap as evidence of wrongdoing without eliminating those in writing will be embarrassed. And where the agency bought media on its own account and resold it, a gap between what the platform charged and what the client was billed is the expected result of the agreed structure, not a deviation from it. The reconciliation determines the fact; the contract determines its meaning.

Disgorgement, and who carries the apportionment burden

Disgorgement is attractive to plaintiffs because of how the burden is allocated. Under § 1117(a) the plaintiff's affirmative showing is the defendant's sales; the defendant must establish all elements of cost or deduction claimed. The Ninth Circuit's Model Civil Jury Instruction 15.31 tracks the statute and adds the part that matters most to a marketing expert: the defendant must also establish any portion of the profits attributable to factors other than the infringement.

That is a marketing question wearing a damages label. Brand strength, price, distribution, unrelated advertising, product quality, retail placement, seasonality — evidencing the non-infringing drivers of a defendant's sales is a legitimate, well-defined defense-side assignment, and it is one of the clearest places in this practice where the marketing record does load-bearing work.

Two precision points. Romag Fasteners, Inc. v. Fossil Group, Inc., 590 U.S. 212 (2020), held that willfulness is not an inflexible precondition to a profits award under § 1117(a) for a § 1125(a) violation; it did not hold that mental state is irrelevant, and the Court said a defendant's mental state is relevant to assigning an appropriate remedy. And the defendant's profits are not the plaintiff's loss. The statute lists them as separate recoveries with separate proof requirements, so a disgorgement figure built from the defendant's sales says nothing about what the plaintiff lost, and a lost-profits figure says nothing about what the defendant gained.

Corrective advertising, and a percentage that is not a methodology

Corrective advertising damages are available in principle in several circuits, and the arithmetic behind the leading case deserves to be read rather than cited.

In Big O Tire Dealers, Inc. v. Goodyear Tire & Rubber Co., 561 F.2d 1365 (10th Cir. 1977), a jury awarded $2.8 million on the theory that the award would let the plaintiff advertise at equivalent volume to correct the record. The Tenth Circuit held that excessive and reduced it to $678,302, arrived at by taking 28 percent of the $9,690,029 the defendant had spent on the campaign — the plaintiff had dealers in 14 of 50 states — and then cutting that by 75 percent to reflect the Federal Trade Commission's practice of ordering roughly 25 percent of an advertising budget spent on correction.

That 25 percent is not a rule of damages law. It is a 1977 borrowing of an agency's remedial practice, applied by analogy to a private award, with no statutory basis and no empirical basis in the record of the case. Adray v. Adry-Mart, Inc., 76 F.3d 984 (9th Cir. 1995), allowed prospective corrective advertising costs but capped them at the damage to the value of the mark, given the danger of overcompensation. And in Zazu Designs v. L'Oreal, S.A., 979 F.2d 499 (7th Cir. 1992), Judge Easterbrook wrote that "using a percentage of some number unrelated to the plaintiff's injury is an unacceptable way to estimate damages," adding that people who want damages "have to prove them, using methodologies that need not be intellectually sophisticated but must not insult the intelligence." A model built as a percentage of the defendant's ad spend should expect to meet that sentence.

Four separate leaps, each needing its own support

The chain most marketing damages models walk has four links, and each is a separate claim requiring separate evidence.

Rankings are not traffic. A position change moves impressions and click-through, but the relationship is not fixed and depends on query mix, the shape of the results page, and how much of the demand was branded.

Traffic is not conversions. Traffic that arrived through a different query, on a different device, with a different intent, converts differently. A decline concentrated in traffic that never converted has a small revenue consequence and a large chart.

Conversions are not revenue. Platform-reported conversion counts are not counts of observed conversions — Google models conversions for cross-device, consent-limited and device-restricted situations, and the export does not mark which. Conversions also cannot be summed across platforms, because each attributes to itself under its own window and its own model, and the overlap is not knowable from the reports.

Revenue is not profit. Margin, variable cost treatment and mitigation sit between them, and they belong to the accounting expert.

A model that traverses all four without evidencing each is the model that produces a headline figure and no defensible basis for it. In Grasshopper House the competing per-visit valuations were roughly $40 on one side and $1.80 on the other — an order-of-magnitude spread that tells you how little constraint the underlying method imposed.

How these models actually come apart

Three failure patterns account for most of what happens to damages opinions in this space, and none of them is about arithmetic.

Causation left unaddressed inside the model. In Grasshopper House, LLC v. Clean & Sober Media, LLC (9th Cir. 20 Aug. 2021), an unpublished memorandum disposition and therefore not precedent under Circuit Rule 36-3, the panel affirmed exclusion of a damages expert whose per-webpage-visit valuation rested on a regression the district court found "was flawed as to the issue of causation" and would be "wholly useless to the jury." It is a digital marketing damages theory excluded on causation grounds in a false advertising case, and it is the closest analogue available.

Competing causes omitted. Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8th Cir. 2000) — a model that "ignored inconvenient evidence" in the record was excludable however sophisticated it was.

The wrong comparison series. In re Executive Telecard Securities Litigation, 979 F. Supp. 1021 (S.D.N.Y. 1997) — an expert who benchmarked against an index with "no meaningful correlation" to the security at issue had a methodological defect, not a presentational one.

There is a large body of decisions excluding economic and accounting experts for failing to isolate the effect of the challenged conduct, and a much thinner body specifically about digital marketing experts. The field is young and most of these matters settle. The general standard applies with full force regardless.

What the model has to disclose on its face

A damages model in this field carries disclosure obligations that a financial model does not, because the inputs are unstable in ways a reader cannot see.

• The extract date of every platform figure. Analytics 4 documents that conversions can be reattributed for up to seven days after the conversion; the same report pulled on day 1 and day 10 can legitimately differ.
• Which figures are modeled and which were counted. The report does not distinguish them on its face, and an expert who cannot say which a given cell is has a problem waiting in deposition.
• The attribution model, conversion window and consent configuration in force in each period compared, and whether any of them changed inside the comparison.
• Every definitional break falling inside the window — the January 2021 Meta window change, the April 2021 tracking permission requirements, the September and November 2023 attribution migrations.
• What was not available and could not be checked. Retention windows in this field are short and rolling; stating what had already expired is both accurate and the honest answer to "did you consider X?"

None of this weakens a model. A figure whose provenance and instability are disclosed is a figure the other side has to argue with. A figure presented as though it were read off an instrument is one they only have to characterize.

Frequently Asked Questions

Published 2026-08-12