Paid search in litigation
Abstract ladder rung illustration representing Agency and Client Disputes

What it turns onData plus discoveryThe account shows part of it; the rest comes from the other side.

Agency and Client Disputes

Short answer
The account shows what was done and what it cost; approval and markup come from discovery
Where it comes from
Ownership and access history, change history, platform cost against invoiced cost
Who holds it
Split — the advertiser, the agency's own manager account, and the agency's ledgers
Retention window
2 years of interface change history; 37 months of daily cost data, read Aug 14, 2026
What will not work
The change log names the user who made a change, never whether the client approved it
Applies to
Advertiser against agency, consultant or in-house manager over spend, access and results

Half of an agency dispute sits in the account; the other half sits in the agency's ledgers and arrives only through discovery

What the account settles, and what has to come from the other side

An agency dispute divides cleanly in two, and the division decides how much of it an examination of the account can carry.

On one side sit the facts the platform records: who created the account, which manager account holds owner status, every user who had access and at what level, every change with a timestamp and a login, what each campaign cost by day, what conversion tracking existed and whether it fired, and what condition the account was left in. Those are exportable, reproducible and not seriously arguable once produced.

On the other side sit the facts the platform never sees: whether the client approved a budget increase, what the agency was paid by media suppliers, what margin sat between media cost and invoice, and what anyone intended. Those live in email, signed authorizations, supplier agreements and the agency's own ledgers.

So the honest answer to "will the account prove my case" is that it proves the operational half and usually tells you which documents to demand for the rest. Whether any of it breaches the agreement is for counsel. The rest of this page is what the record shows and where it stops.

Ownership is a status in the platform, not a consequence of paying

The most useful thing to learn in week one is that Google Ads has a specific technical meaning for ownership, and it has nothing to do with who paid for the media. From Google's manager account documentation:

  • An owner manager account has full administrative and data access privileges, which may include personal information.
  • A client account can have only one owner.
  • Ownership is transitive: if a manager account owns a client account, every manager above it in the hierarchy also has ownership.
  • Ownership is created automatically when a manager account creates a new client account. A manager linking to an account that already exists does not get ownership by default; the client has to grant it.
  • Only owner admins can transfer ownership, turn it off, and change certain security settings. Admins on an owner manager account can invite and remove users — the mechanism by which an advertiser gets locked out of an account it believed was its own.

The consequence is direct. If the agency created the account, its manager account became the owner by default and nobody signed anything for that to happen; an advertiser that assumed it owned the account because it paid the invoices may hold no owner status at all. If the advertiser created the account and merely linked the agency, it retained ownership and can sever access unilaterally. Which history occurred is a binary, discoverable fact, visible in the creation record and the manager-link history.

The compensation model decides what evidence exists

Before anything is examined, establish how the agency was paid: the model determines both the incentive and whether a platform-native record of the media exists in the client's name.

ModelWhere the dispute comes from
Percentage of media spendIncentive to increase spend regardless of return; budget increases the client says it never approved
Flat monthly feeIncentive to minimize labor; neglect, unmonitored automation
Performance, per conversion or leadIncentive to shape the conversion definition; what was counted, and whether tracking inflated it
Non-disclosed or principal buyingThe agency buys as principal and resells at an undisclosed margin; the client never sees platform cost
Rebates and volume incentivesSupplier payments based on volume, invisible in any ad platform export

The conversion-definition row deserves attention in a performance-priced engagement, because it is measurable. When a conversion action was created, what it counts, whether its count setting is every conversion or one per click, whether it changed mid-engagement, and whether one user action was counted by two tags are all recorded. Pricing tied to conversions plus a definition that changed in month four is a documentable sequence, whatever it turns out to mean.

When the media was never bought in the client's account

The non-disclosed model is the one most attorneys underestimate. In a principal buy the client is not the platform's customer — the agency is. What the client received as "reporting" is an agency document, not a platform export, and there may be no account in the client's name at all.

Reconciliation is then the whole case, and the question is narrow: does the invoiced media cost tie to any figure the platform can produce, month by month and campaign by campaign? If the agency cannot produce a platform-native export that ties to its invoices, the gap is the finding — measurable work with a clean output, and usually the first thing to do.

For the scale of what has been documented in the industry generally, the citable source is the study the advertiser trade body commissioned from K2 Intelligence, released June 7, 2016. Its reported findings: cash rebates from media companies based on spend volume that advertisers said they neither received nor knew about; rebates as free inventory; rebates structured as low-value service agreements tied to spend; dual rate cards for principal versus agent roles; and principal transactions with markups of approximately 30 to 90 percent. The methodology belongs alongside the number: 150 completed interviews between October 20, 2015 and May 31, 2016, all anonymized. The announcement of that study is the primary reference. It establishes that these practices occur and are known. It names no agency and quantifies no prevalence, and an opposing expert will say so, correctly.

What the named matters actually establish

Named US appellate authority on advertiser-versus-agency paid-search disputes is thin, and the reason is structural. Master services agreements almost universally carry arbitration clauses, so disputes resolve confidentially and produce no published opinion; those filed usually settle before a dispositive ruling. No appellate decision squarely addressing negligent paid-search account management was located for this site. So the dispute gets litigated on the contract, the record and expert testimony rather than on precedent. Three matters are worth knowing accurately; two are routinely described backwards.

Uber Technologies, Inc. v. Fetch Media Ltd., filed September 2017 in the Northern District of California, alleged the mobile agency had billed for fraudulent clicks and fake installs, with roughly $40 million at issue; the agency counterclaimed for about $20 million in January 2018. Uber voluntarily dismissed on December 22. There is no merits ruling, and coverage describing it as a win is wrong.

Phunware, Inc. v. Uber Technologies, Inc., No. CGC-17-561546, San Francisco County Superior Court, is the instructive one, and the reason is the sanction rather than the conduct. Uber obtained terminating sanctions for spoliation, having established that files from a whistleblowing employee and logs showing where ads were actually placed had been destroyed. That account comes from trade coverage rather than from the order, which is worth obtaining before anyone relies on it. The matter was effectively decided on preservation failure: the party that fails to preserve platform logs, placement reports and internal messages is at severe risk, and the destruction is itself provable.

Outside litigation, the 2016 Dentsu disclosure is a benchmark for what improper billing looks like when an agency admits it: 633 suspicious digital advertising transactions affecting 111 advertisers, dating from November 2012, with admitted practices including fees charged where no placement was made and reports misrepresenting performance. About ¥230 million was repaid.

A handover, measured

There is no tort of returning an ad account in bad condition. Claims of that shape get pleaded as breach of the transition or wind-down clause, as conversion or trespass where access or data is withheld, occasionally as breach of fiduciary duty where state law and the contract support it, and sometimes under computer-misuse statutes where a departed agency keeps using access. Which fits is for counsel. What I can do is measure the condition, and the measurement is the same whatever the theory:

  1. Ownership and access history. Who created the account, which manager holds owner status, the history of manager links and unlinks, and every user with access, at what level, granted and revoked on what dates.
  2. Change history across the handover. Mass pauses, campaign deletions, removal of conversion actions, deleted negative keyword and audience lists, and bid strategy resets in the final days appear with user identity and timestamp — if examined before the window closes.
  3. Conversion tracking integrity. Whether conversion actions existed, when they were created, whether the tag was present on the site, and whether the action was ever attached to the campaigns reported on. A conversion action created but never fired is a specific, provable finding.
  4. Portability. Whether remarketing lists, offline conversion imports and analytics linkage survived the transition or left with the agency's manager account.
  5. Structural condition. Negative keyword coverage, location targeting and the presence-versus-interest option, network settings, and whether campaigns were left on automated bidding with no target.
  6. Reconciliation. Platform cost against invoiced cost, by month and campaign.

The overspend claim, and what Google's documentation does to it

Budget claims are the most common opening theory and the most common early casualty. Google publishes the rule:

On a given day, your campaign might spend up to twice your average daily budget to take advantage of fluctuations of traffic. … At the end of the month, you will have spent no more than 30.4 times your average daily budget.

PeriodDocumented ceiling
Any single day2× the average daily budget
A calendar month30.4× the average daily budget

Source: Google Ads Help, About overdelivery and your average daily budget, read August 14, 2026. A claim built on a single day's overspend that does not address this is not a claim, and it is better to learn that in week one than in a deposition.

What survives is different in kind: spend above an authorized total, budget changes with no written authorization behind them, spend concentrated in campaigns outside the approved plan, or a monthly charge exceeding the documented cap. Each is provable from budget change history with user and timestamp, joined to daily cost and read against the authorizations counsel produces.

What no export settles, and how long the record lasts

The limits matter as much as the findings. Each is where the record hands off to discovery.

  • Whether the client approved a change. Change history records the login that made it, never the consent behind it. Approval lives in email, meeting notes and signed authorizations.
  • Undisclosed markup. If media was bought in the agency's own account, the client's platform record may not exist. Proving margin requires the agency's records and billing — a discovery problem before an analysis problem.
  • Rebates from media suppliers. Invisible in platform data; only supplier agreements and the agency's ledgers show them.
  • Intent behind a destructive change. That every campaign was deleted on the last day is provable; that it was malicious rather than instructed is not, from the log alone.
  • Causation of lost revenue. That performance declined after a handover is measurable. That the handover caused it, rather than seasonality, a competitor or site changes, requires a counterfactual and is contested.
  • Whether the agency was worth the money. Not a data question.

The clock decides more of these matters than the merits do. Google Ads change history in the interface reaches back two years, and Google states data older than that will not be available; the API returns field-level change detail for only the past 30 days. Since June 1, 2026, reporting data has been retained 37 months at hourly, daily and weekly granularity and 11 years at monthly and coarser granularity. Those figures are from Google Ads Help and the API documentation, read August 14, 2026. One exception is worth raising on the first call: Search Ads 360 change history goes back to May 21, 2016 or later, so if the advertiser used it, a far longer record may exist.

Frequently Asked Questions

Who owns a Google Ads account, the advertiser or the agency?

Whoever holds owner status in the platform, which is not the same as whoever paid the invoices. Google's manager account documentation states that a client account can have only one owner, that ownership is transitive up the manager hierarchy, and that ownership is created automatically when a manager account creates a new client account. A manager that links to an already existing account does not get ownership unless the client grants it. So if the agency created the account, the agency's manager account is the owner by default. It is a binary fact, and it is checkable in minutes.

The agency will not hand back access. What does the platform record show?

It shows the access history itself: which manager account holds owner status, the full record of manager links and unlinks with dates, every user with access and at what level, and when each grant or revocation happened. It also shows what happened in the final days of the engagement — mass pauses, deleted campaigns, removed conversion actions, deleted negative keyword and audience lists — with a login and a timestamp on each. Google's documentation also notes the client account still owns its data and can remove an owner manager's access by unlinking, which is sometimes the fastest practical step.

How do I test whether the agency marked up the media?

By reconciliation, and the first question is whether a platform record in the client's name exists at all. Under a disclosed model the platform cost is exportable by month and by campaign and can be set against the invoices directly; a difference is visible immediately. Under a non-disclosed or principal model the media was bought in the agency's own account, the client was never the platform's customer, and there may be nothing on the client side to reconcile. At that point margin is established from the agency's own platform records and ledgers, which is a discovery question rather than an analysis question.

Did Uber win its ad fraud case against its agency?

No. Uber sued Fetch Media in the Northern District of California in September 2017 over roughly $40 million in allegedly fraudulent clicks and installs; Fetch counterclaimed in January 2018 for about $20 million; and Uber voluntarily dismissed the case on December 22 after reassignment. There is no merits ruling, and trade coverage describing it as a victory is inaccurate. The matter Uber did prevail in was against Phunware, a mobile ad network, in California state court, and it prevailed on terminating sanctions for spoliation rather than on a merits finding about the advertising itself.

Does the finding that only 36 cents of an ad dollar reaches the consumer apply to Google Ads?

No, and transposing it is an easy way to be impeached. That figure comes from the advertiser trade body's December 2023 programmatic supply chain study, which analyzed $123 million in spend and 35.5 billion impressions across 21 advertisers and reported roughly 29% in transaction costs and 35% in media productivity loss. It describes programmatic display, with its chain of demand-side and supply-side platforms and resellers. The search auction does not have that chain. What the study does support in a search dispute is the narrower principle that log-level data access, not agency-prepared summaries, is an articulated advertiser expectation.

Are the ANA and 4A's transparency documents an industry standard the agency breached?

They are not a standard of care. The 2016 K2 study commissioned by the advertiser trade body is interview-based and anonymized: it establishes that rebates, dual rate cards and principal markups of roughly 30 to 90 percent occur and are known, and it names no agency and quantifies no prevalence. The agency trade body's Transparency Guiding Principles of Conduct, issued January 28, 2016, are voluntary guidance, released unilaterally after joint work with advertisers stalled over whether the output should be prescriptive. Their disagreement is itself the point: the industry did not agree on a standard.

How quickly does the account record have to be preserved?

Faster than most engagements move. Field-level change detail through the Google Ads API covers only the past 30 days, so a clean machine-readable export of it has to happen inside that window; a 90-day lookback shows which resources changed but not old and new values or who changed them. The interface change history holds two years and moves forward every day. Reporting data has been kept 37 months at daily granularity since June 1, 2026, per Google Ads Help read August 14, 2026. A dispute surfacing three years after the conduct may have no native change record left.

Keep reading

The guides run the sequence

A page here covers one dispute, or one kind of record. A guide covers the order the work happens in — what has to be exported before access is lost, and which analysis is worth paying for at all.

Top