Most agency pitches show you a dashboard and a case study. Neither tells you whether the team can do the actual work your channels need. Meta and TikTok are creative jobs: turning a customer insight into an ad worth testing. Google Search is an intent job: deciding which queries to buy, which to exclude, and whether the economics work.
Hiring an agency that’s strong at one but weak at the other means half your budget gets expert attention and the other half gets managed by whoever’s available.
Ask for the right work samples before you sign. This article shows what to request and how to score it.
The short version: Judge agencies by the specific job each channel requires. On Meta and TikTok, ask how customer insight becomes an ad worth testing. On Google Search, ask which queries the team buys or excludes and why. Both work samples should connect decisions to an affordable customer acquisition cost.
Which paid media skills should buyers test first?
The skills that turn your business problem into a campaign decision. On Meta and TikTok, that means concept development and creative learning. On Google Search, it means intent selection, keyword judgment, and conversion economics.
Platform automation makes those responsibilities more consequential, not less. The system acts on the goals you give it. If the creative team feeds it weak concepts, it optimizes delivery of weak concepts. If the search team points it at the wrong queries, it buys more of them.
According to Meta’s Performance 5 guidance, creative diversification belongs alongside automation, data quality, simplified accounts, and results validation. An agency selling creative volume alone leaves several of those responsibilities unanswered.
These are different emphases, not exclusive job descriptions. Search still needs persuasive messaging, and paid social still needs sound account management and reliable customer data.
| Skill | Meta and TikTok | Google Search |
|---|---|---|
| Customer judgment | Problem, motivation, hook | Query intent and buying stage |
| Campaign input | Distinct concepts and formats | Keywords, exclusions, relevant pages |
| Learning evidence | Creative test and next brief | Search terms and qualified outcomes |
| Commercial control | CAC, quality, spending limits | CAC, CPC economics, conversion value |
Once you know which skills to look for, the next question is how to see them in action. Start with creative.
What should a Meta or TikTok creative work sample show?
A connection between a customer insight and an ad concept worth testing. Ask what distinguishes each concept from the others. Then ask what result would change the next brief.
Three edits of one video might test an opening, but they don’t test different reasons to buy your product. That distinction matters. A team producing variations of the same argument is doing production work. A team producing distinct hypotheses about buyer motivation is doing strategy work.
Give them a scenario: a scheduling product whose buyers waste hours coordinating meetings across teams. Ask for two concepts. One might show the frustration of calendar ping-pong. The other might show a sales team losing deals because onboarding calls keep slipping. Both address a real problem, but they test different buyer motivations.
Then ask what stays fixed across the test. If the audience, offer, landing page, and video all change at once, the result tells you almost nothing about which concept worked. Agree on the comparison before production starts.
According to TikTok’s creative guidance, tests benefit from substantially different creative. Its recommendations also cover early hooks, platform-native execution, and refreshing assets when results decline. That calls for judgment about when to pivot, not a fixed replacement calendar.
Request a real, anonymized example of a result changing the next brief. Our Meta creative testing guide covers that operating loop. A production schedule should leave room to act on what the campaign teaches you.
The creative sample reveals whether an agency thinks in concepts. The search sample reveals something different: whether they think in buyer intent.
What should a Google Search work sample show?
Which demand the agency wants to capture, which searches it would exclude, and why those choices reflect your product and sales process. Ask for the reasoning, not just the keyword list.
Then follow the thread from search intent to ad copy to landing page. A good agency defends those choices using lead quality and acquisition cost, not just click volume.
Take the same scheduling product. Compare “appointment scheduling software” with “free weekly schedule template.” The first searcher may be evaluating tools to buy. The second probably wants a printable PDF. Ask the agency: what does each searcher expect? Does our product fit? Which landing page meets that expectation?
Don’t exclude a query just because it contains “free.” A free product tier might make it commercially relevant. The question is whether the searcher could become a customer, not whether the word sounds cheap.
Google’s Search guidance describes AI Max expansion through broad match and keywordless matching. Ask how the team reviews that expansion, evaluates outcomes, and corrects mismatches between searches and landing pages.
Have the agency identify a search it stopped buying and explain the evidence behind that decision. Our Google Ads agency guide covers SaaS-specific selection criteria. Use the sample to test those claims against actual decisions.
Knowing what to buy and what to skip is one skill. Knowing what you can afford to pay per click is another.
Can the agency explain the CPC your economics support?
It should. Ask the agency to calculate a planning CPC using your acquisition allowance and expected click-to-customer rate. Then ask how that conversion estimate reflects your actual sales funnel.
Change the assumptions and watch what happens. If the click-to-customer rate drops by half, what does that do to the budget? The team should explain the consequences before recommending bids or promising that cheaper traffic solves the problem.
Take a business that allows $400 in advertising spend per acquired customer. At a 1% click-to-customer rate, the planning CPC is $4 ($400 times 0.01). Cut the conversion rate to 0.5% and the allowable CPC drops to $2.
Those are planning assumptions, not industry benchmarks or manual limits for Smart Bidding. Define whether your CAC allowance includes agency fees and production. If it does, deduct those costs before calculating the advertising-only allowance.
A cheaper click can still produce a more expensive customer. Ask to see the full path from click to qualified lead to sale. And account for conversion delays before judging whether recent spending has paid back.
With creative samples, search samples, and economics in hand, you need a fair way to compare agencies against each other.
How do you score agency work samples fairly?
Score the reasoning and evidence behind each sample, not the polish of the presentation. Use the same brief for every candidate. Ask for an example connecting a past campaign decision to its outcome.
Agree internally on which capabilities the role must own before you start comparing agencies. A persuasive presenter may understand the vocabulary without having done the work your account requires.
Score 0 for no evidence, 1 for a plausible explanation, and 2 for a sample showing a real decision and its outcome. These scores structure the interview. They can’t predict campaign performance, but they reveal whether the team has done this work before or is describing it for the first time.
| Work sample | Evidence worth two points | Question to ask |
|---|---|---|
| Creative test | Distinct hypotheses and follow-up brief | What changed after the result? |
| Search review | Intent decisions tied to lead quality | Which demand deserved more spend? |
| Economic model | Explicit CAC and conversion assumptions | What happens when conversion halves? |
| Measurement plan | Defined goals and quality feedback | What does the bidder count? |
| Ownership plan | Named decisions and approval boundaries | Who acts when results deteriorate? |
Scoring helps you pick a team. But the hiring decision also depends on who owns what after the contract starts.
Who owns performance when the platforms automate delivery?
The agency owns campaign decisions within agreed limits. Your team owns business constraints and sales feedback. Name the creative lead, channel operator, and measurement owner before work starts.
In a small team, one person can hold several roles. That’s fine. Each responsibility still needs an owner with the access and authority to act when the campaign needs attention.
TikTok Smart+ automates targeting, budgets, placements, and parts of creative production with advertiser controls. Google’s value-based bidding guidance ties optimization to the conversion values and funnel stages advertisers supply. Both systems depend on the quality of what goes in. Ownership of those inputs is the job.
Agree who approves new concepts, changes exclusions, adjusts spending, and returns qualified-lead data. Specify what happens when tracking fails. Those operating agreements matter as much as the person opening Ads Manager.
Our paid media with AI framework covers the broader operating model. Bring these work samples when evaluating paid media support so the discussion stays grounded in decisions the team will own.
Choosing the right skills gets easier once you know which account problem comes first. We can help you connect the hiring decision to that problem. Our Paid Media Analysis uses the information you share to identify 3 priorities for your next 90 days.
