Choosing an acquisition agency is mostly a question about what you can undo. Almost every claim a supplier makes about performance is unverifiable at the point of signature. What is verifiable, before you sign, is who will own the accounts, what the supplier earns on your money, and what happens on the day you leave. Those three answers predict more about the engagement than any case study.

The platforms have published the ownership rules, and they are not symmetrical. The industry has published research on its own compensation practices, and it is not flattering. Both are usable as selection criteria today.

Ask the ownership question before anything else

Where the account is created determines who controls it forever, and on one platform the word forever is close to literal.

Meta is the strict case. Meta states that if you create a new ad account in a business portfolio, “it will permanently be a part of that portfolio”, and that “the ad account can’t be deleted or transferred from the portfolio”. The same applies to datasets: “you can’t delete a dataset or transfer it to another business portfolio.”

The consequence, in practice. An agency that opens your ad account inside its own business portfolio has created an asset you can never take with you. You can be granted access to it, and that access can be revoked. There is no later remedy, because the platform documents none.

Google is the permissive case. Google states that manager account owners have full administrative access “but do not take data ownership or administrative rights away from client accounts”, and that “the client account still owns its data and has the ability to remove ownership access by unlinking.”

LinkedIn sits in between. LinkedIn states that only one Business Manager can own an ad account, and documents no transfer procedure. Its guidance where nobody with access is reachable is that “you might need to create a new ad account.”

Which gives you one instruction. Create the accounts yourself, in your own business portfolio or manager account, then grant the agency access. It costs an afternoon before the engagement and removes an entire category of dispute at the end of it.

Advertising account ownership and transfer rules compared across three platformsComparison of the published account ownership and transfer rules on the three principal advertising platforms, which determine whether a client can recover control of assets created by an agency. On the social platform, an advertisement account created inside a business portfolio permanently belongs to that portfolio, and the platform states that the account cannot be deleted from or transferred out of the portfolio, with the same rule stated for datasets, meaning an account opened by an agency inside the agency’s own portfolio can never be moved to the client and the client can only ever hold revocable access to it. On the search platform, manager account owners hold full administrative access but the platform states that this does not take data ownership or administrative rights away from client accounts, that the client account still owns its data, and that the client retains the ability to remove ownership access by unlinking, with a recovery form available on proof of ownership even where no active administrator remains, subject to the exception that unlinking while relying on the manager’s shared remarketing tag stops the audience list from populating and that a cross account conversion tag stops recording. On the professional network, only one business manager can own an advertisement account, no transfer procedure is documented, billing administration belongs to whoever created the account, and the platform’s guidance where no person with access remains reachable is that a new advertisement account may need to be created. The practical instruction that follows is to create every account in the client’s own business portfolio or manager account before the engagement begins and then grant the agency access, since this costs an afternoon in advance and removes the entire category of dispute at the end.Who owns the account, per the platforms’ own wordsMetaPermanent, no transferAccount created in a portfoliostays in it permanently.Same rule for datasets.Audiences can be shared butnever transferred.Owner deletion cascades toeveryone it was shared with.GoogleClient keeps ownershipManager access does not takeownership from the client.Client can unlink at will.Recovery form exists on proofof ownership.Exception: shared tags stopworking once unlinked.LinkedInOne owner, no procedureOnly one Business Managercan own an ad account.No transfer path documented.Billing admin is whoevercreated the account.Fallback: create a new adaccount.The instruction that covers all threeCreate every account yourself, in your own portfolio or manager account, then grant access. Doing it in theother order is irreversible on at least one platform and undocumented on another.
The three platforms document three different answers to the same question. Only one of them is reversible by default. Source : Meta Business Help Center, Google Ads Help, LinkedIn Help (2026)

Treat a performance guarantee as a disqualifier

The only supplier claim you can check at signature is whether the claim is the kind that can honestly be made at all.

Google’s position is unambiguous. “No one can guarantee a #1 ranking on Google.” Its guidance adds that you should “beware of SEOs that claim to guarantee rankings, allege a ‘special relationship’ with Google, or advertise a ‘priority submit’ to Google.”

The same logic covers paid media. Auction outcomes depend on competitor bids, on quality signals and on demand, none of which a supplier controls. A guaranteed cost per lead is either a claim about things outside the supplier’s control or a promise to buy worse leads to hit the number.

And there is a legal dimension in the US. The FTC’s substantiation doctrine requires advertisers to hold a reasonable basis for objective performance claims before making them. The Commission has acted on this in the sector: FTC v. Pointbreak Media, 2018, concerned deceptive claims made to small businesses about Google listings.

What to ask instead of accepting a guarantee. Ask what the supplier will commit to that it actually controls: response times, change logs, the definitions in reporting, the cadence of tests. Those are commitments that can be met or missed, and therefore mean something.

One nuance worth stating. A supplier declining to guarantee results is not evasion. It is the only accurate answer, and the ones who give it are usually the ones who will tell you the truth later as well.

Ask what the agency earns on your money, not what it charges you

The invoice is one number. What the supplier makes on the media can be another, and the industry’s own research says so.

The study that established this. In 2016 the Association of National Advertisers published a study conducted by K2 Intelligence into media transparency in the US market. It drew on 150 sources, 117 of them working in US media buying, and 59 reporting direct experience of non-transparent practices, 34 of them relating to rebates and 33 to principal-based buying.

The magnitudes reported. Rebates ranging from 1.67 percent to around 20 percent of spend. Principal markups, where the agency buys inventory on its own account and resells it, of roughly 30 to 90 percent.

The context that matters for reading it. Five of the six largest holding companies declined to participate, and the 4A’s, the agencies’ own trade body, called the study “anonymous, inconclusive, and one-sided”. Both facts belong in any honest citation of it.

Why it still applies to a small B2B engagement. Not because a ten-person agency runs principal-based buying, but because the disclosure question it raises is the right one and costs nothing to ask: does anything you receive from a platform, a reseller or a technology vendor depend on what I spend?

And the model most engagements now use. The ANA’s compensation survey, 18th edition, fielded in the second quarter of 2022 across 101 marketers and 336 relationships, found 82 percent of relationships fee-based, up from 68 percent in 2016, with performance incentives present in 41 percent, down from 48 percent and from 61 percent earlier.

Findings and limitations of the 2016 media transparency studyFindings and stated limitations of the media transparency study published in 2016 by the Association of National Advertisers and conducted by K2 Intelligence into the United States media buying market. On sample, the study drew on one hundred and fifty sources in total, of which one hundred and seventeen worked in United States media buying, and of which fifty nine reported direct personal experience of non transparent practices, comprising thirty four sources reporting rebates and thirty three reporting principal based buying, with some sources reporting both. On magnitude, rebates were reported ranging from one point six seven percent to approximately twenty percent of media spend, and principal markups, arising where the agency purchases inventory on its own account and resells it to the client, were reported at approximately thirty to ninety percent. On limitations that any fair citation must carry, five of the six largest agency holding companies declined to participate in the study, and the American Association of Advertising Agencies, the agencies’ own trade body, characterised the study as anonymous, inconclusive and one sided. On why the study remains relevant to a small business to business engagement, the point is not that a small agency operates principal based buying but that the disclosure question the study raises costs nothing to ask at selection, namely whether anything the supplier receives from a platform, a reseller or a technology vendor varies with the amount the client spends. On compensation structure, the separate association compensation survey, eighteenth edition, fielded in the second quarter of 2022 across one hundred and one marketers and three hundred and thirty six agency relationships, found eighty two percent of relationships to be fee based, up from sixty eight percent in 2016, with performance incentives present in forty one percent, down from forty eight percent and from sixty one percent in earlier editions.The 2016 study, with its own caveats attachedWhat it reported150 sources, 117 in US media buying.59 reported direct experience:34 on rebates, 33 on principal buying.Rebates: 1.67% to about 20% of spend.Principal markups: about 30% to 90%.What limits itFive of the six largest holdingcompanies declined to participate.The agencies’ own trade body calledit “anonymous, inconclusive, andone-sided”.How engagements are actually paid, 2022 survey of 101 marketers and 336 relationshipsFee-based: 82%was 68% in 2016Performance incentives: 41%was 48%, and 61% earlierThe question that costs nothing to askDoes anything you receive from a platform, a reseller or a technology vendor vary with what I spend?
The 2016 study's own numbers, alongside the participation gap and the trade response that any fair citation has to carry. Source : ANA and K2 Intelligence, Media Transparency Initiative, 2016 (2016)

The eight questions worth asking before you sign

None of them are about results. All of them have answers you can check.

Who creates the accounts, and in whose portfolio. The single most consequential answer, for the reasons above. Anything other than “yours” needs a written explanation.

What do you earn on my media beyond your fee. Rebates, volume arrangements, reseller margins, technology commissions. The answer “nothing” is fine and should be in writing.

What will you commit to that you control. Response times, change logs, reporting definitions, test cadence. A supplier who cannot name three is selling outcomes it cannot deliver.

What is your reporting attribution window, and can you change it. Ask for a sample report before signing and check whether it states its own window, model and counting setting. Reports that do not carry their definitions cause the argument at month six.

What access will you need, and at what level. Google’s own SEO guidance recommends granting a prospective supplier read access to Search Console only, and specifically “at this stage, don’t grant them write access”. The same restraint applies across tools during evaluation.

What happens on the day I leave. Which accounts, tags, audiences and datasets stay with me, in what format, and within how many days. Get the answer against the platform rules above, not against the supplier’s intentions.

Who actually does the work. The named person, their other accounts, and what changes if they leave. This is the question most often answered by whoever will not be doing the work.

What would make you tell me to stop spending. The answer separates a supplier from a vendor. A supplier who has never advised a client to reduce budget has either never had a client below the volume thresholds or has never said so.

Eight checkable questions to ask an acquisition agency before signingEight questions to ask a prospective business to business acquisition agency before signature, chosen because each has an answer that can be checked against a document or a platform rule rather than against a case study. First, who creates the advertising accounts and inside whose business portfolio, where any answer other than the client’s own requires a written explanation because the social platform’s rule is irreversible. Second, what the agency earns on the client’s media beyond its stated fee, covering rebates, volume arrangements, reseller margins and technology commissions, where an answer of nothing is acceptable and should be put in writing. Third, what the agency will commit to among things it actually controls, such as response times, change logs, reporting definitions and testing cadence, where an inability to name three indicates a supplier selling outcomes it cannot deliver. Fourth, what attribution window the reporting uses and whether it can change, tested by requesting a sample report before signature and checking whether it states its own window, model and conversion counting setting. Fifth, what access the agency will need and at what level, where the search engine’s own guidance for evaluating search suppliers recommends granting read access to the search console only and specifically not write access at the evaluation stage. Sixth, what happens on the day the client leaves, covering which accounts, tags, audiences and datasets remain with the client, in what format and within how many days, assessed against the platforms’ published rules rather than the supplier’s intentions. Seventh, who actually performs the work, meaning the named individual, their other accounts and what changes if they depart, a question most often answered by a person who will not be doing the work. Eighth, what circumstances would lead the agency to advise the client to stop spending, an answer that separates a supplier from a vendor.Eight questions with checkable answers1. In whose portfolio are the accounts created?Any answer but “yours” needs a writtenexplanation. On Meta it is irreversible.2. What do you earn beyond your fee?Rebates, volume deals, reseller margins,tech commissions. “Nothing” in writing is fine.3. What do you commit to that you control?Response times, change log, definitions,test cadence. Three minimum.4. What attribution window does reporting use?Ask for a sample report first. Check whetherit states window, model and counting.5. What access do you need, at what level?Google advises read access to Search Consoleonly while evaluating. No write access.6. What happens the day I leave?Accounts, tags, audiences, datasets. Formatand deadline. Checked against platform rules.7. Who actually does the work?Named person, their other accounts, and whatchanges if they leave.8. What would make you tell me to stop?The answer that separates a supplier froma vendor.
Every question has a checkable answer. None of them depends on trusting a case study. Source : Method, with Google Search Central guidance on evaluating SEO suppliers (2026)

Read the credentials for what they actually certify

Badges and scores are real, but most of them certify something narrower than the impression they create.

Platform partner status certifies spend and process, not judgement. It tells you the agency clears a management threshold and passes a certification requirement. It says nothing about whether its judgement suits an account of your size.

Optimization score is not a performance measure. Google defines it as “an estimate of how well your Google Ads account is set to perform”, scored from 0 to 100 percent, and says applying recommendations “can help improve” performance. A supplier presenting a high optimization score as a result has presented a configuration setting.

Quality Score does not aggregate the way it is usually quoted. Google states plainly: “There is no such thing as ad group-level, campaign-level or account-level Quality Score.” It is computed on exact search terms over a 90-day window. An account-average Quality Score in a pitch deck is a number the platform says does not exist.

Case studies without denominators say nothing. A percentage improvement with no base, no period and no attribution window is compatible with almost any underlying reality, including a base of four conversions.

What does carry signal. A named practitioner who can talk about an account that did not work and explain what they changed as a result. It is not a credential, and it is the most predictive thing in the room.

What agency credentials certify compared with what they are commonly taken to meanComparison between what three commonly presented agency credentials actually certify and what they are usually taken to mean by a prospective client. Platform partner status certifies that the agency clears a managed spend threshold and satisfies a certification requirement, and is commonly taken to mean that the agency’s judgement has been assessed, which it has not, and it carries no information about whether that judgement suits an account of the client’s size. Optimization score is defined by the search platform as an estimate of how well an advertising account is set to perform, scored from zero to one hundred percent, with the platform stating that applying its recommendations can help improve performance, so it describes account configuration rather than results, and a supplier presenting a high optimization score as an achievement has presented a configuration setting rather than a performance outcome. Quality Score is stated by the platform to exist only at keyword level against exact search terms over a ninety day window, with the platform stating explicitly that there is no such thing as ad group level, campaign level or account level Quality Score, so an account average Quality Score quoted in a pitch is a figure the platform says does not exist. Case studies quoting percentage improvements without a denominator, a period or an attribution window are compatible with almost any underlying reality including a base of four conversions. The credential that does carry predictive signal is not a badge at all but a named practitioner able to describe an account that did not work and explain what they changed as a result.What the credential certifies, and what it is heard asCredentialWhat it actually certifiesWhat it is heard asPlatform partnerbadgeA managed spend threshold and acertification requirement.That the judgement hasbeen assessed.Optimization score”An estimate of how well your accountis set to perform.” Configuration.A performance result.Account Quality ScoreNothing. Google: “There is no such thingas … account-level Quality Score.”An account health grade.The signal that is not a badgeA named practitioner who can describe an account that did not work, and what they changed because of it.
Three widely quoted numbers, set against the platforms' own definitions of them. Source : Google Ads Help, on optimization score and Quality Score (2026)

Design the trial so that it tests something

A short paid engagement before a long one is sensible, provided it is scoped to produce a readable answer.

What a trial cannot test. Performance. The learning periods and volume thresholds mean a 30-day trial ends before the account is in a steady state, whatever the supplier does.

What a trial can test, and test well. Whether tracking gets fixed, whether reporting arrives with definitions, whether the change log exists, whether questions get answered in the time promised, and whether the structure is explicable.

So scope it accordingly. A paid diagnostic and setup engagement with named deliverables: a tracking audit with a test conversion verified end to end, a documented account structure with rationale, a first report carrying its definitions, and a written 90-day plan.

Which produces a decision on evidence. At the end you hold four documents. Their quality is a far better predictor of the next twelve months than any cost per lead measured inside a learning phase.

And it protects the incoming work. A trial framed as a setup phase gives the stabilization window a name, so the relearning clock starts cleanly rather than being restarted repeatedly by a supplier trying to show movement inside thirty days.

How to scope a paid trial engagement with an acquisition agency so that it produces a readable answerHow to scope a short paid trial engagement with a prospective acquisition agency so that it produces an answer that can actually be read at the end of it. A trial cannot test performance, because the platforms’ published learning periods and volume thresholds mean that a thirty day engagement ends before the advertising account has reached a steady state, whatever the supplier does during it. A trial can test process thoroughly, specifically whether tracking is corrected, whether reporting arrives carrying its own definitions, whether a dated change log exists, whether questions are answered within the time promised, and whether the account structure can be explained. The trial should therefore be scoped as a paid diagnostic and setup engagement with four named deliverables. The first deliverable is a tracking audit including a test conversion verified end to end from the advertising platform through the analytics property to the customer relationship management system. The second is a documented account structure accompanied by a written rationale for why campaigns are divided as they are, given the client’s conversion volume. The third is a first report that carries its attribution window, model and conversion counting setting on the page and separates provisional days from settled ones. The fourth is a written ninety day plan. At the end the client holds four documents whose quality predicts the following twelve months far better than any cost per lead measured inside a learning phase. Framing the trial as a setup phase additionally gives the stabilization window a name, so that the relearning clock starts cleanly rather than being restarted repeatedly by a supplier attempting to demonstrate movement within thirty days.Scope the trial to four documentsWhat thirty days cannot testPerformance. The account is still inside the learning period when the trial ends.1. Tracking auditA test conversion verified end to end:platform, analytics, CRM.2. Structure with rationaleWhy campaigns are split this way,given your conversion volume.3. First report, with definitionsWindow, model, counting setting onthe page. Provisional days marked.4. Written ninety day planWhat will be changed, in what order,and what would falsify it.Naming the trial a setup phase also names the stabilization window, so the relearning clock starts once.
A thirty day trial ends before the account reaches a steady state, so scope it to produce documents instead of numbers. Source : Method, over Meta's published learning phase requirement (2026)

What a shortlist should actually compare

Once the questions are answered, the comparison is between documents, not between pitches.

The sample report. Does it state its attribution window, model and counting setting on the page? Does it separate provisional days from settled ones?

The proposed account structure, with a rationale. Not a screenshot of somebody else’s account. A written explanation of why campaigns would be split the way they are, given your volume.

The contract’s exit clause, read against the platform rules. If it promises to transfer a Meta ad account created in the agency’s portfolio, the supplier has either not read the rule or is describing something it cannot do.

The named practitioner’s availability. How many accounts, and whether your account is one where they will be doing the work or reviewing someone else’s.

And the answer on volume. A supplier who looks at your monthly conversion count and tells you the account is below the optimization thresholds is a supplier who has read your data. One who does not raise it either did not look or decided not to say.

What to do with this

Create your own ad accounts, in your own business portfolio and manager account, before any conversation moves to contract. Then ask the eight questions in writing and keep the answers.

If you are already in an engagement where the accounts sit in the supplier’s portfolio, check which platform. On Google you can unlink. On Meta, check whether the ad account and the dataset were created inside the agency’s business portfolio, because if they were, the practical remedy is to create new ones under your own portfolio and plan the migration, not to request a transfer that the platform does not support. If it is useful to read one supplier’s answers to these questions before you put them to a shortlist, ours are on the page describing how we run B2B paid acquisition, including who owns the ad account and why we will not guarantee a cost per lead.

Once the shortlist is down to two, the useful next steps are auditing your ad account before switching and setting review dates using how long before you can judge an agency.