Most brands hire their first performance creative agency at an awkward moment. Meta spend has crept past the point where the founder can keep making ads on a phone, the winning ad from spring is visibly tiring, and the in-house designer is spending Fridays resizing statics. Somebody says "we need an agency", a few pitch decks arrive, and every one of them shows the same thing: a reel of nice-looking ads and a slide promising to scale ROAS.
The reel tells you almost nothing. A brand studio can make a beautiful ad that never exits the learning phase, and a scrappy performance shop can make an ugly one that carries your account for six weeks. The difference barely shows in a pitch. It's in how they decide what to make, how fast they react to the account, and whether anyone on their side looks at your numbers after delivery. A bad pick doesn't fail loudly. It quietly burns a quarter of test budget on ads nobody learned anything from.
For the definition, see our guide to what performance creative is. This page assumes you already want a performance creative agency for Meta ads and need a way to tell the real ones from the reel merchants.
Key takeaway
Score agencies on six things you can check in a pitch: testing volume sized to your budget, how they iterate winners, whether they read your account data, how they brief, how they handle rights and disclosure, and brief-to-live speed. Ignore the reel until the scorecard says they're worth a second call.
What changed on Meta, and why it changes who you hire
Targeting used to do a lot of the work. You picked interests and lookalikes, and the ad mostly had to not be terrible. Now the algorithm does that job. In December 2024 Meta's engineering team described Andromeda, the retrieval system that narrows tens of millions of ad candidates down to a few thousand relevant ones, and said it expected the number of ad creatives in its system to "grow significantly" as generative tools spread.
In our view, that means the creative now does a big share of the targeting. Different hooks, personas and formats find different pockets of people. Your creative partner is closer to a growth hire than a supplier, and the old way of buying creative (a brief, a deck of concepts, a delivery date, goodbye) leaves money on the table. You want someone who treats each batch as a set of bets and comes back for the results.
The six criteria that separate performance partners from production shops
For each one: what good looks like, the red flag, and the question that exposes it on a pitch call.
1. Testing volume sized to your budget
Volume is the easiest thing for an agency to sell and the hardest for your account to digest. Meta's help centre says an ad set usually exits the learning phase after about 50 results in the week after its last significant edit. Meta's own page on managing ad volume is blunter: run too many ads at once and each one delivers less often, fewer exit learning, and performance gets worse.
Run the maths on a typical account before you sign anything:
| Input | Value | How it's worked out |
|---|---|---|
| Monthly Meta spend | $10,000 | Assumption for this example |
| Weekly spend | About $2,300 | $10,000 divided by 4.33 weeks |
| Target cost per purchase | $45 | Assumption for this example |
| Spend for 50 purchases in a week | $2,250 | 50 x $45 |
| Share of spend for testing | About $770 a week | One third of weekly spend |
| New ads if the agency ships 40 a month | About 9 a week | 40 divided by 4.33 weeks |
| Test spend per new ad | About $83 | $770 divided by about 9.2 ads |
| Purchases per new ad | Under 2 | $83 divided by $45 (about 1.8) |
Under two purchases per ad is a coin toss with a production bill attached. At that budget the whole account can only just carry one ad set out of learning, so an agency proposing 40 fresh ads a month is selling its own capacity. A smaller slate of distinct concepts, each with a few hook variants, will teach you more. Our post on how much to spend per creative to get a real read works through the per-ad budget properly.
Good looks like: they ask for your spend and target CPA before quoting a volume. Red flag: one ads-per-month package for every client. Ask: "Of the ads you made for a client at our spend last quarter, how many got enough budget to reach a verdict?"
2. Iteration on winners
New concepts are fun to make and fun to present, so most agencies over-index on them. In our experience the cheaper wins come from the ad that's already working: a new hook on the same body, a shorter cut, a different creator on the same script, a static pulled from the best video frame. Once your audience has seen an ad too many times, Meta shows "Creative limited" when its cost per result is above your past ads, and "Creative fatigue" when it reaches twice that. By then a strong partner already has the replacements in the edit.
Good looks like: each new batch includes variants of the current winners, labelled with what changed. Red flag: every batch is all new ideas, and nobody can say which element of last month's winner did the work. Ask: "Show me the last three iterations you made of a single winning ad, and what each one tested."
3. Accountability to account metrics
This is the line between a performance creative agency and a production company with a performance slide. The agency doesn't need to run your media. It does need to read the results of its own work, ad by ad: hook rate, hold rate, click-through, cost per purchase. Otherwise nobody on their side knows why an ad won, and the next brief is a guess.
Keep the two jobs apart in the contract. Media buying (budgets, bids, campaign structure) sits with you or your media buyer. The creative partner gets reporting access and a regular review, and owes you a next brief built on what the data said. If an agency wants full control of your ad account before it has made a single ad, ask why.
Good looks like: a standing review where they bring a short list of what to cut, what to iterate and what to test next, tied to named ads. Red flag: reports that count deliverables ("24 assets shipped") rather than results. Ask: "What did you stop making for a client last quarter because the numbers told you to?"
4. How they brief
Ask to see a real brief, anonymised. A performance brief starts from a hypothesis (who the ad is for, what they believe now, which angle should move them) and specifies the hook word for word. A production brief starts from the brand guidelines and a mood board. Both can produce good-looking work; only one produces a test you can learn from.
Watch how they separate the variables. Persona, angle, format and creator are different levers. If a brief changes all four at once, a winning result can't tell you which one mattered. Our piece on building a creative system that drives growth covers how briefs feed the testing loop.
Good looks like: one hypothesis per concept, written down before production, with what would count as a win. Red flag: concepts pitched as "vibes" with no stated audience or belief to shift. Ask: "Talk me through the hypothesis behind your last losing concept, and what you changed because of it."
5. Rights, usage and disclosure
This is the least exciting criterion, and the one that bites months later. In our experience creator content often comes with limited usage terms (a set period, set channels), and some creators charge extra to run the ad from their own handle as a partnership ad. If your winner's licence runs out in month four, you either pay again or pause the ad that's carrying the account.
Disclosure belongs here too. In the US, the FTC's guidance for influencers says a "material connection" to a brand (payment, free product, a family or employment tie) has to be disclosed. In the UK, the CAP and CMA influencer guide, now on its third edition from March 2023, sets out how ads should be labelled. An agency that shrugs at this is handing you the risk.
Good looks like: usage terms written into the contract up front, covering paid use, channels, duration and partnership ads. Red flag: "we'll sort rights per video". Ask: "If our best ad is still winning in month nine, what does it cost us to keep it running?"
6. Speed from brief to live
A slow creative pipeline costs you quietly. If fatigue shows up in week three and replacements take two months, your account spends the gap running tired ads at a rising cost. Ask for the real timeline from signed brief to ads uploaded, including casting, shipping product, revisions and your own approvals. Then ask what slows it down. A partner who knows their bottlenecks usually has a fix for them.
Good looks like: a dated plan for the first batch and a shorter cycle for iterations of existing winners. Red flag: a timeline that only holds if you approve everything the same day. Ask: "When a winner starts to fatigue, how many days until its variants are live?"
How do you score a shortlist of performance creative agencies?
Rate each agency from 1 to 5 on every criterion, multiply by the weight, divide by 5, and add it up for a score out of 100. The weights below suit a DTC brand spending steadily on Meta; shift them to fit. A brand with one hero product and a fatiguing winner might double the weight on iteration and halve briefing.
| Criterion | Weight | Scores a 1 | Scores a 5 |
|---|---|---|---|
| Testing volume sized to budget | 20 | Fixed package, no questions about spend or CPA | Volume derived from your spend, CPA and learning-phase maths |
| Iteration on winners | 20 | Every batch is new concepts only | Named variants of live winners in every batch, each testing one change |
| Accountability to account metrics | 20 | Reports count assets delivered | Reads ad-level results and turns them into the next brief |
| Briefing | 15 | Mood boards and brand guidelines | Written hypothesis per concept, hook scripted, variables separated |
| Rights, usage and disclosure | 15 | Rights "sorted per video" | Paid usage, duration and partnership ads covered in the contract |
| Brief to live speed | 10 | Vague timeline, no plan for fatigue | Dated first batch, faster cycle for iterations |
A worked example with two made-up agencies shows how the reel can mislead. Agency A is a design-led studio with a gorgeous portfolio. Agency B has a plainer reel but a sharp process.
| Criterion (weight) | Agency A rating | Agency A points | Agency B rating | Agency B points |
|---|---|---|---|---|
| Testing volume (20) | 2 | 8 | 4 | 16 |
| Iteration (20) | 2 | 8 | 5 | 20 |
| Account metrics (20) | 1 | 4 | 4 | 16 |
| Briefing (15) | 3 | 9 | 4 | 12 |
| Rights and disclosure (15) | 4 | 12 | 3 | 9 |
| Speed (10) | 3 | 6 | 4 | 8 |
| Total | 47 | 81 |
How I'd read the totals: 75 or above, move to a paid first batch. Between 55 and 74, go back with questions on the lowest-scoring criterion before deciding. Under 55, it doesn't matter how good the reel is. Agency A might be the right hire for a brand campaign. For a Meta account that lives or dies on testing, it's the wrong one.
Which kind of creative partner fits your situation?
Very different businesses call themselves a performance creative agency. The scorecard tells you whether one is good. This table tells you which kind to shortlist in the first place.
| Your situation | Best fit | Why |
|---|---|---|
| Testing paid social for the first time, small budget | A UGC marketplace or a few freelance creators | You need a handful of test videos, not a programme. Keep strategy in-house while you learn. |
| Steady Meta spend, in-house or freelance media buyer, winners fatiguing faster than you can replace them | A specialist performance creative studio | Your bottleneck is the creative pipeline and the iteration loop, which is exactly what a specialist studio is built to own. |
| No one to run the account at all | A full-service paid social agency with creative in-house | You need media buying first. Check their creative against the scorecard, because it's often the weaker half. |
| Brand launch, rebrand or TV-led campaign | A brand or design agency | Different job. Bring a performance partner in once the campaign needs testing at volume. |
| Mostly need volume of lightly edited creator clips | A UGC agency with an editing team | Cheaper per asset, though you'll own the testing plan yourself. |
If you're comparing named UGC-led agencies specifically, our rankings of the best UGC agencies for Meta ads and the wider best UGC agencies of 2026 by use case do that job. Run the scorecard on whoever makes your shortlist.
Where Spark fits, and where it doesn't
Full disclosure: Spark UGC is our studio, so weigh this section accordingly.
Spark is a specialist performance creative studio. We handle research, concepts, scripts, creator casting, production, editing and the iteration loop after launch. We don't buy media: we make the creative, and you or your media buyer run the account. We read the ad-level results with you and use them to brief the next round. We've made 5,000+ ads, more than $55M in ad spend has run on Spark creatives, and the blended ROAS across Spark client accounts is 8.7x. A typical cycle from brief to live ads takes about 21 days. How it works lays out the process step by step.
We're the wrong fit if you need a couple of cheap test videos to see whether paid social works at all, or if you want someone to run your campaigns as well as make the ads. A marketplace suits the first. A full-service agency suits the second.
Performance creative agency FAQ
What does a performance creative agency do?
A performance creative agency makes ads designed to be tested and judged on paid-media results such as cost per purchase and ROAS, rather than on brand recall alone. A good one researches angles, writes a hypothesis for each concept, produces the ads with hook variants, reads the ad-level results and iterates on the winners. It does not have to run your media buying.
How do I choose a creative agency for Facebook ads?
Score each agency on six criteria: testing volume sized to your budget, iteration on winning ads, accountability to ad-level account metrics, how they brief, how they handle usage rights and disclosure, and speed from brief to live ads. Ask for anonymised briefs and examples of iterations rather than relying on a portfolio reel.
How many new ads should I test on Meta each month?
As many as your budget can read, which is usually fewer than agencies propose. Meta says an ad set typically exits the learning phase after about 50 results in the week after its last significant edit, and that running too many ads at once means each delivers less often. Work back from your weekly spend and target cost per purchase so each new ad gets enough budget to reach a verdict.
Should my creative agency also run my Meta ads?
Not necessarily. Media buying and creative are separate skills, and many brands keep the ad account with an in-house or freelance media buyer while a specialist studio makes the creative. What matters is that the creative partner sees ad-level results and uses them to decide what to make next.
Pick the agency whose process you'd still trust after a losing month. That's the one that will find the next winner. If a specialist studio fits your account, tell us what you're spending and what's tiring, and there's more on testing and creative strategy in the resources library.