Agency and Consultant Disputes

The element that breaks first, and why it should be conceded early

A negligence claim against a service provider requires the plaintiff to establish what a reasonably competent practitioner would have done, that the defendant fell below it, and that the shortfall caused the loss. In a licensed profession the first element arrives ready-made: a board, an examination, a published body of practice, a disciplinary mechanism. Digital marketing has none of those, and no general licensure requirement.

It is worse than an absence. The two trade bodies whose documents would be offered as the industry standard have said in writing that they disagree about whether one of them is a standard. The Association of National Advertisers publishes a master media buying agreement template and describes it as "a starting point for the development of a customized contract" in which "changes in template provisions should be expected." On 1 July 2024 the American Association of Advertising Agencies published guidance addressed to version 3.0 of that template, calling it "a complex and one-sided document" and stating that "the 4A's does not recognize ANA templates as industry standards."

Nor do the credentials in circulation fill the gap. The Media Rating Council runs genuine annual audits, but accredits measurement services rather than practitioners, and platform certifications are free, retakeable vendor product exams. Google's Partner badge requires a 70% optimization score, a 90-day ad spend of $10,000 across managed accounts, and certification of half the account strategists — a measure of how closely an account follows Google's own recommendations, a volume threshold, and an exam. None measures the client's outcome.

Say all of that first. An expert who leads with "the industry standard for account structure" is testifying to his own preferences, and a competent opponent gets there in five questions.

What a court can look to instead

Four sources of duty survive that problem, and they carry nearly every defensible opinion here.

The contract and the statement of work. By a distance the strongest, because it is what the parties actually agreed. Duty, deliverable, metric, approval rights, audit rights, and who owns the ad accounts on exit are all in it or conspicuously absent from it.
The platform's own published rules. Google Ads policies and Meta's advertising standards are enforceable by the platform against the account. They create no duty running to the client, but they can establish that a specific act was prohibited by the system the agency chose to operate in — and account suspension is a harm with a date on it.
Verifiable facts of execution. Conversion tracking that never fired. A tag removed on a datable day. Budget running to a landing page returning a 404. Spend billed that does not match spend delivered.
Internal inconsistency. The agency's own reporting contradicting the platform's record for the same period and the same definitions.

Where industry custom is offered, the answer is the one courts have used since The T.J. Hooper, 60 F.2d 737 (2d Cir. 1932): a whole calling can lag behind reasonable prudence, so proof of universal practice is evidence rather than a defense. That cuts both ways, and it leaves a plaintiff's expert who has nothing but his own view of best practice with no anchor at all.

A poor result is not by itself evidence of negligence. Campaigns underperform for reasons unconnected to the agency: competitor entry, pricing, client-side site changes, fulfillment failure, algorithm and measurement changes, and the plain fact that paid media does not work equally well for every offer. A declining return-on-ad-spend chart is evidence that reported return on ad spend declined. Any opinion that travels from "results were bad" to "the agency was at fault" without an intervening, record-based causal step is the kind this practice exists to take apart.

Reconciling agency-reported spend against platform-reported spend

This is the recurring finding, and it is arithmetic rather than judgment. For each period there are three quantities:

A — what the platform charged, from its own billing record. On Google Ads monthly invoicing the API exposes an Invoice resource carrying adjustments, regulatory costs, taxes and account budgets.
B — what the platform's reporting interface says was spent: the Cost column, by campaign, by day, exported from the account.
C — what the agency invoiced the client, alongside what its own report said had been spent.

A gap between A and B is usually benign: credits and refunds, invalid-activity credits, overdelivery adjustments, currency conversion, tax and regulatory operating costs, the boundary between a billing month and an account time zone, and accrual in the interface against the invoiced period. An expert who calls that gap misconduct without eliminating those first will regret it.

The comparison that matters is C against A. Where the client was billed more than the platform charged and no markup was disclosed, that is a fact neither party can talk away, because the two records were created independently and neither party controls both. It is not automatically a breach. A disclosed markup, an agreed management fee computed on spend, or an agreed principal-buying arrangement produces exactly the same arithmetic as an undisclosed margin. The reconciliation settles the fact; the contract settles the meaning.

One further pattern is easy to miss. Where C equals A but the agency's report does not match B, the invoice is right and the reporting is wrong — and every performance metric reported that period was computed on the wrong denominator.

Principal or agent, and why the distinction decides the disclosure question

An agency buying media as agent is subject to ordinary agency duties, including the duty not to take an undisclosed benefit from a third party in the transaction. An agency buying inventory for its own account and reselling it is a counterparty, and its obligations are the ones the contract creates. Whether the line was crossed, and whether the client consented in writing, is frequently the whole case. Whether an agency owes fiduciary duties at all turns on the forum's law and the contract, and that is a legal question rather than an expert one.

The debate starts from the ANA's Independent Study of Media Transparency, prepared by K2 Intelligence and published 7 June 2016. Its methodology belongs beside every citation of it: 143 interviews with 150 sources between 20 October 2015 and 31 May 2016, of whom 117 were directly involved in media buying and 59 reported experience with non-transparent practices. It found rebates ranging from 1.67% to approximately 20% of aggregate media spending, markups on media resold through principal transactions of approximately 30% to 90%, and that none of the 22 advertisers interviewed reported awareness of rebates flowing back to their agencies. That is an anonymous-source interview study commissioned by the advertiser side, which the 4A's called "anonymous, one-sided" at the time. It is not an audit, not a sample and not a finding of any tribunal; it establishes nothing about any particular agency, and there is no audited public figure for US agency rebate retention.

The argument has since moved to principal media. An ANA survey released 19 March 2026, fielded among 114 client-side respondents self-selected from the trade body's own members, reported 58% using principal media in the past year, up from 47% in 2024, with only 57% holding guidelines governing it. The neutral formulation is that the question is rarely whether principal media is permitted, but whether the parties agreed the rules for it.

One document settles the posture on its face. The 4A's/IAB Standard Terms and Conditions for Internet Advertising, Version 3.0 (2010) establishes sequential liability at § III(c): the media company holds the agency liable "solely to the extent proceeds have cleared from Advertiser to Agency," and for uncleared sums holds the advertiser solely liable. An agency buying on those terms is buying as agent, in writing. An agency that has taken the inventory onto its own balance sheet has abandoned that posture, and the insertion orders show it.

The record set a reconstruction is built from

Seven categories, each separately obtainable, each usually held by a different party:

The contract, every SOW and every amendment, plus the compensation schedule and any audit, rebate or principal-media clause.
Account ownership and access records — who owns each ad account, which manager account it sits under, who held what access and from when, and which payment profile actually paid. The Google Ads API documents BillingSetup as the association that "effectively determines who pays for an advertiser's account," which answers the agent-or-reseller question before anyone argues about the contract.
Change history for each account, and invoices against platform spend, as above.
Campaign structure over time, snapshotted across the engagement. Reconstruction from change history plus current state is possible and lossy, and a report should say so.
Communications — email, chat, decks, budget approvals. This is where these cases are usually won and lost, and it is not technical evidence.
Every report the agency gave the client, beside the platform's export for the identical period and settings.

The product is a dated, sourced, reconcilable timeline. The opinions are whatever it will bear.

What none of it settles

Change history records that a setting changed, when, and under which login. It does not record why, does not record who instructed it, does not distinguish a deliberate act from an error or a bulk-applied recommendation, and does not capture every category of change. The defensible formulation is that the record shows X occurred on this date under this login, and does not show why.

A spend reconciliation is arithmetic and nothing more. It cannot show intent, cannot establish that a markup was undisclosed where the disclosure was oral, and cannot separate a deliberate margin from a bookkeeping lag. Where the agency bought as principal and resold, the gap between what the platform charged and what the client paid is the expected result of the agreed structure rather than a deviation from it.

And the production is almost never complete. Retention windows are short and rolling, change history ages out, ad accounts get deleted at the end of an engagement. A report has to state what was unavailable and what could not be checked, which is also the only honest answer to "did you consider X?" on cross-examination.

Rebuttal engagements, which are common here and rarely written about

A large share of the work here is responding to a report that already exists, and three features make it productive. The metric-definition problem: an opening expert who computes return on ad spend or "conversions" without stating the attribution model, the conversion window, the conversion actions included and the extract date has produced a number nobody can reproduce, including the expert. The one-source problem: opening reports are frequently built on the client's copy of the agency's own reporting, while the independent second and third records often still exist. And the causation gap, where a report travels from "the agency did X" to "the client lost Y" with nothing in between.

The discipline that keeps a rebuttal credible is knowing where it stops. Saying that an analysis does not support its conclusion, and setting out why, is a complete rebuttal opinion. A rebuttal expert who volunteers "and therefore the correct figure is Z," without being asked to compute Z, has started offering an affirmative opinion that may be untimely and outside what Rule 26(a)(2)(D)(ii) permits.

Frequently Asked Questions

Is there an industry standard of care for a digital marketing agency?

Not in the sense a licensed profession would recognize. There is no general licensure requirement, no disciplinary body, and no single standards body. The advertiser trade body publishes a contract template and expressly disclaims that it is a standard; the agency trade body stated in July 2024 that it does not recognize those templates as industry standards. The Media Rating Council audits measurement products rather than practitioners, and platform certifications are vendor product exams. What remains is the contract, the platform's own published rules, verifiable execution facts, and internal inconsistency in the agency's reporting.

Can poor campaign results by themselves support a negligence claim?

A poor result is not by itself evidence of negligence. Performance moves for reasons unconnected to the agency: competitor entry, seasonality, pricing, product and site changes made by the client, fulfillment problems, auction dynamics, algorithm changes, policy enforcement, and measurement changes that alter the reported result without altering the actual one. A declining performance chart establishes that reported performance declined. Linking it to conduct requires a record-based step in between — a configuration that was wrong on a datable day, a tag that stopped firing, spend that went somewhere the client did not approve.

How do I test whether an agency billed more than the platform charged?

Put three records side by side for each period: the platform's own billing record, the platform's reporting export for the same dates, and the agency's invoices and client reports. The comparison is objective because the first and third were created independently and neither party controls both. Before characterizing any gap, an expert should eliminate credits and refunds, invalid-activity credits, currency conversion, tax and regulatory operating costs, time-zone boundaries and accrual timing, in writing. A remaining gap is a fact. Whether it is a breach depends on whether the contract authorized a markup, a fee computed that way, or principal buying.

Does a Google Partner badge or a platform certification establish competence?

It establishes what the badge measures, which is not quality of work. Google's published requirements are a minimum optimization score of 70%, a 90-day ad spend of $10,000 across managed accounts, and certification of at least half of the account strategists. Optimization score reflects how closely an account follows Google's own recommendations; ad spend is volume; the certifications are free, retakeable vendor product exams that expire. None is an outcome measure for the client, and the first is arguably a measure of deference to the vendor whose spend the agency is meant to be managing.

What is the difference between an agency acting as agent and acting as principal?

An agent buys media on the advertiser's behalf and is subject to ordinary agency duties, including the duty not to take an undisclosed benefit from a third party in the transaction. A principal buys inventory for its own account and resells it, making the relationship on that transaction buyer and seller, governed by whatever the contract says. The distinction is usually visible in the documents before anyone argues about it: sequential-liability terms on an insertion order describe an agent, and the payment profile that actually funded the platform account describes the posture in practice.

What should counsel send an expert first in an agency dispute?

The operative pleadings, so the analysis serves the claims actually pleaded; the contract, every statement of work and every amendment, without which no scope or standard-of-care opinion is possible; the scheduling order with the expert and rebuttal dates; the protective order, before any confidential material moves; and the production itself in native format with metadata intact, because exports converted to PDF are frequently useless for this work. Add any litigation-hold notices covering the ad accounts and analytics properties. The most common irreversible loss in these matters is data that aged out of a platform while the parties argued about the protective order.

Published 2026-08-12