A performance creative agency for DTC is a shop whose primary deliverable is the ad itself, not a media-buying dashboard. We build the creative that scales paid social, and we accept the same scoreboard the brand does: CPA, ROAS, and contribution margin.
The reason this seat exists is structural. Once Meta's Advantage+ and Google's Performance Max took the bidding layer behind the curtain, the lever brands could still pull moved upstream into the creative.
The pitch in one line
We build the ad creative that scales DTC paid social. Judged on CPA, ROAS, and contribution margin, not awards.
That sentence is the whole offer. The rest of the page is the diagnostic that explains why it is the right offer in 2026.
Why creative is the lever DTC brands actually need in 2026
Bid and budget tuning got commoditized. Native platforms automated the mechanical work through Advantage+ and Performance Max, and the marginal performance lever moved to creative velocity and concept diversity.
Attribution got noisier at the same time. After Apple's App Tracking Transparency landed in April 2021, roughly 75% of iOS users declined tracking, deterministic matching collapsed, and platforms shifted to Aggregated Event Measurement and probabilistic models with 24 to 72 hour delays.
That noise is not an argument for buying more tracking tools. It is the reason the asset has to do more work in-platform: the algorithm reads the creative (visuals, audio transcript, hook) and matches audiences off that signal more than off any audience setting you choose.
Put bluntly, creative is the targeting. A shop that wins paid social in this environment is a shop that ships better assets faster, not one with a fancier bid strategy.
What a performance creative agency actually produces
Net-new concepts, not ten text-overlay swaps on one master video. The deliverable that moves an account is a stream of genuinely different ideas tested against different psychological angles: urgency, social proof, demonstration, founder POV, problem-agitation.
When you audit a vendor's "30 creatives per month," ask what fraction are net-new concepts versus iterations and variants. The ratio tells you whether you bought a creative engine or a render farm.
Concept, hook, iteration: the unit definitions
Volume claims are useless without unit definitions, so here are ours.
| Unit | Definition | What it costs to make | Where it lives |
|---|---|---|---|
| Concept | A distinct creative idea built around one psychological angle (e.g. demonstration vs. social proof) | Highest. A new shoot, founder cut, or fresh creator brief. | New ad set, top of the test funnel |
| Hook | The first 3 seconds, the part that earns the scroll-stop | Cheap when produced as a batch; industry rates run roughly $50 to $100 per hook variation on existing footage | Hook tests inside a winning concept |
| Iteration | A material edit to a working concept: new angle, new VO, new claim ordering | Mid. Editor time only. | Iteration sprint on a scaled winner |
| Variant | A trivial swap: text overlay, CTA card, aspect ratio | Lowest, often automated | Format coverage and creative refresh |
A "100 creatives this month" number that is 95 variants and 5 iterations is not a creative program. It is a Photoshop output dressed up as one.
Volume that actually moves an account
Volume scales with spend, not with ambition. Reference points from teardowns of DTC growth-stage accounts: 15 to 30 net-new variants per month is the typical growth-tier output, and elite enterprise programs push 100 to 300 plus per month through creator-led and AI-assisted workflows.
We brief inside that range, weighted to net-new concepts at the top of the funnel and iteration sprints on the three to five hooks per quarter that earn the spend.
For deeper work on creative strategy and platform-specific format craft, we link out to dedicated playbooks.
How the creative-led system works
Five steps, each one short, each one a thing you can audit in our weekly review.
1. Diagnose the account, not the brand book
We pull the existing ad account first. Winners and losers get segmented by hook archetype, angle, and format, and we mark what the algorithm is starved for (usually: net-new top-of-funnel concepts, more demonstration, more founder POV).
The brand book gets read second, after we know what shape of creative the account is asking for. This is the same surface a full paid media audit covers in more depth, and the lens we use to answer why ads stopped scaling.
2. Brief concepts against psychological angles
Every concept gets briefed against an explicit angle: urgency, social proof, demonstration, founder POV, problem-agitation, or category education. Briefs name the angle in the first line.
A brief that opens with a vibe ("modern, premium, aspirational") is a brief that produces a beautiful loser.
3. Produce in batches, ship on a weekly cadence
Velocity beats polish in a learning-phase account. The pattern we ship to is the one platforms reward: a hook fails on Tuesday, an edited variation is live by Wednesday.
The weekly batch is the heartbeat. If the cadence slips, the algorithm reverts to the old winners and the new spend stops scaling.
4. Test inside a real framework
Concepts launch into a structured test with explicit hypotheses, sample sizes, and read windows. Not a slush campaign with seven creatives sharing one ad set.
The mechanics live in our testing framework, and the campaign layout sits on top of a clean account structure so the test reads.
5. Read the data, kill or scale
Winners go into iteration sprints: more hooks, more cutdowns, more angles on the same proven concept. Losers die in the next batch, not the next quarter.
The 3 to 5 winning hooks per quarter pattern is real, and it is the pattern that lets you scale the winners without rebuilding the account.
What we judge ourselves on
The scoreboard is the brand's P&L, not the platform's reporting tab.
- CPA on net-new customers, not retargeted loyalists
- Blended MER, not platform-reported ROAS
- Contribution margin per order
- LTV:CAC at the 90 and 180 day marks
The reason we exclude platform ROAS as the win condition is mechanical. After ATT, Meta and Google both model and over-claim conversions through Aggregated Event Measurement and last-click bias respectively, and their combined reported revenue routinely exceeds the actual Shopify bank deposit.
An agency paid against platform ROAS will harvest cheap retargeted loyalists with deep discounts, post a 6x screenshot, and miss the fact that contribution margin is falling.
The full register lives in our metric stack, with deeper pages on MER vs ROAS, contribution margin, and LTV:CAC.
How we price (and why)
Hybrid base plus a performance bonus, with creative production included in scope. The base covers strategy and production overhead so the engine runs through learning phases. The bonus fires only above a brand-set efficiency target.
Every common alternative misaligns the incentive in a different direction.
| Pricing model | How the agency gets paid more | The hidden incentive | Who it actually fits |
|---|---|---|---|
| Flat retainer | Stays the same regardless of work done | Once secured, less work equals more margin | Brands buying a fixed scope they will police |
| % of ad spend (10-20%, often 15% median) | Bigger budget, bigger fee | Push spend up even when marginal ROAS is falling | Brands with elastic demand and a hard margin floor |
| Pure CPA / revenue share (often 5-10% of attributed revenue) | Squeeze cheap conversions | Harvest retargeted loyalists, dodge top-of-funnel | Rare; enterprise accounts with clean attribution |
| Creative-included flat retainer | Stays the same regardless of result | Production output without efficiency pressure | Brands buying throughput, not a unit-economics partner |
| Hybrid (base + efficiency bonus) | Bonus fires above a CPA or ROAS threshold | Aligned: agency wins only if margin works | DTC growth-tier brands optimizing for profitable scale |
The headline retainer is rarely the all-in cost either. Industry analysis puts hidden costs at 15 to 25 percent on top of the agency fee once attribution software (Triple Whale, Northbeam, Rockerbox at $500 to $2,000 plus per month), CRO tools, and roughly 20 to 30 percent of an internal director's time gets included.
We price the hybrid so we are accountable for the work that creates the bonus, not for the spend that creates the invoice. The full structure, and the decision against running it agency vs in-house, sits inside our performance services.
Who this is for
Self-select hard. The wrong fit costs both sides three months.
Right fit
- DTC brands at roughly $20K to $100K monthly paid social spend, the tier where hybrid plus creative-included is the correct shape
- Product-market fit on at least one SKU and a working post-purchase flow
- Operators willing to ship a weekly batch and read the data weekly with us
Right fit at scale
- $100K plus monthly spend brands that need 100 to 300 plus variants per month and creator-led pipelines
- Brands with a defined contribution margin target and a payback window they can name
Wrong fit
- Brands under roughly $5K to $20K monthly spend, where any agency engagement consumes too much of the budget to print a profitable return; better to self-execute or buy production-only
- Brands that want a media buyer to "just run the ads" with no creative refresh
- Brands chasing platform ROAS without a CAC and LTV model behind it
Why creative-led beats media-buying-led
When the bid and budget layer is automated, the honest seat for an agency to occupy is the seat that produces the asset the algorithm rewards. A media-buying shop without an in-house creative engine has outsourced its own performance ceiling to whichever editor it can find that month.
The creative team is not a vendor of the media buyer. In a 2026 account, the media buyer is a vendor of the creative team, deploying budget into concepts the creative engine has already proven.
That inversion is what we are. For more on the platform mechanism, creative is the targeting covers it in depth; for the brief-level craft, see creative strategy.
What we will not do
- Fabricate case study numbers or invent client logos
- Take a percentage of spend that pays us more when your margin gets worse
- Promise a fixed CPA on a brand we have not audited
- Run media without owning the creative
- Cite platform-reported ROAS as the win condition
How an engagement starts
Two doors. Pick the one that matches your appetite for commitment.
Door 1: Paid media audit
A diagnostic, no commitment to a retainer. We read the live account, surface the creative gaps and the test design gaps, and hand back a written plan. Start with the audit if you want a second read before you change agency relationships.
Door 2: Full creative engagement
Hybrid base plus performance bonus, creative production included, weekly batch cadence, weekly read. Start here if you already know the diagnostic and want the engine.
Where to go from here
Book a paid media audit if you want the diagnostic first. Or see the full scope of our performance services if you already know what you need.