- Establish which of the three service shapes a vendor is before comparing prices, or you are not comparing the same thing.
- The review surface differs by platform: search assesses the destination itself, short-video weighs creative-to-page consistency, image-led social looks hardest at continuity and page completeness. Depth on one platform does not imply depth on another.
- Policy accountability stays with the advertiser's account. Outsourcing does not move risk; at best it removes avoidable mistakes.
- Three questions decide most of the pain later: who owns the domains and safe-page content, whether you get raw log access, and what leaves with you when the contract ends.
- Any vendor asking for ad-account credentials, or promising a specific approval outcome, is one to walk away from.
"Cloaking agency" is three different businesses
The boundaries between these are usually blurred in sales conversations, yet their cost structure, deliverables and failure ownership are entirely different.
| Shape | What is actually delivered | Fits whom | Main cost |
|---|---|---|---|
| Managed service | Configures routing rules, writes and maintains the safe page, manages domains and DNS, reports on outcomes — typically a monthly retainer | Teams with nobody able to own the work long-term, or entering an unfamiliar platform | Highest price, and your visibility into your own pages depends on what the vendor chooses to expose |
| White-label reseller | Access to a third-party routing platform under the reseller's brand, with little operational work of its own; support depth is inherited from upstream | Buyers who want the tool rather than the labour, but prefer not to contract upstream directly | An extra margin and an extra hop of latency in information; when upstream breaks you have no direct line |
| Advisory and audit | Reviews the existing configuration, safe page, policy exposure and measurement, and returns a remediation list without executing it | Teams already running their own setup who want an outside read | Solves no staffing problem — whether the findings get implemented is entirely on you |
There is a shortcut for classification: ask what, concretely, was done for the last client each week. A managed service can name configuration changes, safe-page updates and a monitoring cadence. A reseller drifts into describing platform features. An advisory firm talks about reports and recommendations. That single question usually settles it. The build-versus-buy economics behind the same choice are laid out separately in agency versus self-serve tooling and, in cost terms, in the real cost of DIY versus SaaS.
Platform by platform: where the review surface actually differs
The habit of searching for a vendor per platform is rational. Platforms weight destination assessment differently, and a team fluent on the search side does not automatically carry that advantage into short-video. The table below covers the five platforms most often searched separately, with what each weighs most and what a vendor most needs to understand there.
| Platform | What is weighed most | What to test a vendor on |
|---|---|---|
| Google search ads | Completeness and policy fit of the destination itself, relevance to the keyword, and unobstructed crawler access | Whether it can name the specific policy clauses and their boundaries, rather than asserting outcomes |
| YouTube ads | Consistency between what the video promises and what the destination delivers, plus mobile load and usability | Whether creative is assessed alongside the page, or only half the chain is in scope |
| TikTok ads | A strongly mobile, in-app WebView environment where missing referrers and coarse fingerprints both change how a decision resolves | Whether it has hands-on knowledge of WebView, referrer and mobile fingerprint differences |
| Pinterest ads | Continuity from the image to the page, and page completeness — missing navigation, policy pages or contact details weigh heavily | Whether it can state concrete standards for safe-page content instead of reusing one generic template |
| Snapchat ads | Audience-appropriateness requirements and mobile experience, with a short, stable path from creative to destination | Whether it understands the platform's additional audience and content constraints |
The clause-level differences between platforms are itemised in the platform policy comparison, and the general mechanics on the detection side are covered in how ad platforms detect routing. The mobile column in particular is expanded in WebView and referrer differences on mobile.
What transfers to a vendor, and what never does
The most common misconception about outsourcing is that handing over execution hands over the risk. Platform terms bind the account that runs the campaign, and enforcement lands on that account. If the vendor misconfigures something, it is your account that stops.
| Item | Can be outsourced | Note |
|---|---|---|
| Routing configuration and routine monitoring | Yes | Only if you retain read access to logs and can verify independently rather than through a monthly summary |
| Safe-page content and upkeep | Yes | Ownership of the content and the domain should sit with you, or changing vendors means starting over |
| Domain registration and DNS | With caution | Convenient, but a domain held in the vendor's name is the single most common lock-in mechanism |
| Ad account operation | Not advisable | Grant access through the platform's own collaborator mechanism; never hand over credentials |
| Policy accountability | No | Terms bind the advertiser, and no contract wording changes that |
| Data compliance (privacy and cookies) | No | You remain the data controller; a vendor is at most a processor |
Labour can be outsourced; accountability cannot. So the contract conversation worth having is not about guaranteed outcomes — a promise of a specific outcome is itself a warning sign — but about how quickly you can see a problem and how quickly you can take over.
The specific obligations behind the data row are set out in GDPR and CCPA compliance, and the boundaries that should not be crossed at all are listed in the red lines.
Twelve questions to ask before signing
The list below is ordered by how expensive a vague answer turns out to be. If the first four are answered evasively, there is usually little point continuing.
- In whose name are the domains registered, who owns the safe-page content, and does all of it leave with me when the contract ends?
- Do I get read access to raw routing logs, rather than a monthly summary?
- What is the rollback procedure for a configuration change, and how fast does it take effect?
- Will you ever ask for ad-account login credentials? The correct answer is no — access should go through the platform's collaborator mechanism.
- Which platforms do you work on daily, and which do you explicitly decline?
- What are your standards for safe-page content, and do multiple clients share one template or one domain pool?
- How do you describe your own compliance stance, and can it be written into the contract?
- What is the response time and escalation path when something goes wrong — a rejection, or a sudden change in traffic structure?
- What is inside the quoted price and what is not — domains, traffic, page revisions?
- How do you define pass rate and real-page share, and how are they shown to me?
- If I migrate away later, how far do you cooperate with the handover?
- Can you provide a reference client on the same platform at a comparable scale?
If the metric definitions in question ten differ between the two sides, every later report will fail to reconcile, so it is worth aligning them first — see reading routing analytics. The migration cost behind question eleven is routinely underestimated; the full process is in the tool migration guide.
Signals worth walking away from
- A promise of a specific approval outcome. No third party controls a platform's determination, so the promise indicates either inexperience or a bet on yours.
- A request for ad-account credentials instead of the platform's own collaborator access.
- Refusal to provide raw logs, offering only a curated report.
- Shared domain pools or a single safe-page template across clients — one client's problem becomes everyone's.
- Vague pricing, or a performance share with no agreed definition of the metric.
- No exit or handover clause in the contract at all.
- Capability explained by "inside channels" or "special resources" rather than by method.
What each pricing model hides
A flat monthly retainer is the easiest to budget, but the vendor's workload and your actual need rarely track each other: in quiet months you subsidise idle capacity, and in busy months the vendor has an incentive to compress effort. Per-domain or per-link pricing follows usage more closely, but creates a pull toward opening more domains than you need. A performance share sounds fairest, and is — provided both sides can verify the metric independently. Without that, it degrades into a permanent dispute clause.
Whichever model applies, defining what counts as a change, whether a page revision is billable, and who pays for domains matters far more than the headline rate. Those three items generate most later billing disputes.
When in-house is the better answer
Outsourcing pays off in two situations. The first is bursty workload — a concentrated launch period followed by long stretches of low activity, where a dedicated hire cannot be justified. The second is entering a platform you have never run on, where what you are buying is someone else's accumulated experience. Outside those two cases, in-house is usually cheaper over time and carries one advantage no vendor can supply: routing decisions live next to your own measurement, so diagnosis never has to cross an organisational boundary.
A simple threshold works here: if the real weekly effort on this consistently exceeds half a day and someone on the team can own the configuration, the in-house arithmetic usually closes. Tool options are compared in the tools comparison, and the launch phase for a new domain is covered in warming a new domain.
One last point: this is not a one-time decision. When team size, platform mix or campaign cadence changes, the answer should be recalculated. Treat it as a choice reviewed every six months rather than as a position.
FAQ
What does a cloaking agency actually do?
The label covers three businesses. A managed service operates the configuration, safe page and often the domains for a retainer. A white-label reseller sells access to someone else's platform under its own brand and does little operational work. An advisory engagement audits and recommends without executing. Classify first, because the same monthly figure can buy an order of magnitude difference in labour.
Does using an agency reduce my policy risk?
It does not transfer. Platform terms bind the account running the campaign, and enforcement lands there. A competent vendor reduces avoidable mistakes; a poor one adds a layer of opacity between you and a page you answer for. Treat it as risk management, not risk removal.
Why do people search for a vendor by platform?
Because the emphasis genuinely differs: search weighs the destination itself, short-video weighs creative-to-page consistency, image-led social weighs continuity and completeness. A team working a platform daily may have no accumulated experience on another. Asking which platforms a vendor works on and which it declines filters faster than any feature list.
What matters most before signing?
Domain and content ownership, read access to raw logs, rollback speed, and what leaves with you at the end. Vagueness on those four costs the most later. Separately, never work with a vendor that asks for ad-account credentials — access belongs in the platform's collaborator mechanism.
When is in-house better?
When volume is steady, someone can own the configuration long-term, and you want routing decisions sitting next to your own measurement. Outsourcing suits bursty workloads or a platform you have never run on. Those conditions change, so revisit the choice rather than fixing it permanently.
See every routing decision for yourself
The ROAS365 console reports the real and safe surfaces separately, breaks pass rate down by source, and exports the underlying decision logs.