How much should we spend on creative production, and which sourcing model gives the cheapest winners?
Our answer
Budget production by new ads needed, not share of spend. Relaunch past winners first, then lean on statics and image-with-text ads. Run creator video as partnership ads, and use AI for speed and variations, not to replace people on camera.
- Confidence
- Medium
- Strongest proof
- 2 controlled tests
- 8 datasets, 8 practitioners, 7 opinions or claims
- Sources
- 13 sources
- 25 statements
CP
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentMost new ads never matter, and a few carry the spend. Across 504 stores, about 1 in 100 new ads became a breakout, yet breakouts took 41% of new-ad spend (Common Thread Collective). So it relaunches past winners first.
Simple image formats win most often. Text-only and product-image-with-text ads had the highest hit rates in Motion's 2026 benchmark and drew more spend than their share of ads.
People on camera still pay. Human-made ads scored 14% higher short term than fully AI-made versions (Ipsos and Syracuse University). Meta's own figure from its lift studies ties partnership ads to 45% lower cost per incremental purchase.
Our judgment
We plan production from Common Thread Collective's ad-count formula rather than a share of spend, because its figures come from 504 stores while 1800DTC's split (about 30% of spend under $100K a month, about 20% above) and Cody Plofker's 15% target are stated practice, and Plofker himself says total efficiency matters more than the share. Relaunching comes first because it is Common Thread Collective's first lever and needs no new production, and only after a rest because its list is paused winners and the prior year's leaders. We count statics as cheap winners because Taylor Sicard puts one at 30 to 60 minutes against four to eight hours for a video. Using AI for variations and scale, not to replace people on camera, follows the Ipsos and Syracuse University advice to lean on AI there while keeping humans at the center of brand storytelling.
Our judgment, built from Common Thread Collective, 1800DTC, Cody Plofker, Motion, Taylor Sicard and Ipsos and Syracuse University: our reading of these sources, not a finding from any one of them.
What the sources say
25 statements from 13 sourcesCody Plofker says his brand has about 60 creator videos waiting for review while spending about $600 a day on Meta, and that it has too much content relative to spend. He gives no ideal ratio.
“To be honest, we probably have too much content relative to what we're spending.”
In four preregistered field studies for a major music label, AI-made video ads beat human-made ones on TikTok by 26% in click-through, with no significant difference on Meta, at 98% lower cost.
“did not detect statistically significant performance differences”
Cody Plofker says Winks gets 60 videos a month from five retained creators, and an AI agent writes copy, picks landing pages and uploads ads to Meta paused for his review. Its first drafts usually need tweaks.
“Tbh, V1 of ads usually need tweaking.”
Meta's own paper on 20 EMEA lift studies says cells with at least 30% partnership ad assets had 45% lower median cost per incremental purchase. PPC Land notes the claims are Meta's own.
“advertisers mixing at least 30% partnership ads into business-as-usual campaigns recorded a 45% lower median cost per incremental purchase across 20 EMEA conversion lift studies”
Cody Plofker says to budget at least 15% of media spend on creative production.
“I like to at least budget 15%.”
An Operators podcast guest says to reuse creative from 2020 to 2022 campaigns for peak shopping periods, because audience recall is low and there are likely strong ad ideas in there.
“go back to the creative from 2020, 2021, 2022, you probably have banger ad ideas in there.”
Connor Rolain says not to reinvent the wheel for holiday creative: reuse what worked in previous years, such as CTV or linear TV assets, if still relevant.
“don't reinvent the wheel if you've did something in the previous years that you can like rinse and repeat this year.”
Cody Plofker says his new brand Winks is AI-first, with most ads AI-made, statics and videos, and a rule to prove a task cannot be done with AI before hiring for it.
“Most of our ads are AI too, statics and videos”
Common Thread Collective's podcast says they start new offers with static ads because statics are the cheapest to produce and the best way to validate an idea.
“They're the cheapest thing to produce best thing to validate the idea”
Connor MacDonald says creator partnerships can't be justified by the creator's own reach. The content has to resonate enough that you can pay to promote it further.
“you can't justify it via the distribution of the creator themselves.”
Common Thread Collective analyzed 504 stores and about $3.35B of Meta spend: 7.2% of new ads passed $1,000. About 1% of new ads, the whales, took 41% of new-ad spend.
CTC says about 70-75% of Meta spend carries over on its own, so 25-30% must come from new creative; two brands at $200K a month can need 40 or 200 new ads. Relaunch past winners first.
Common Thread Collective says about 70 to 75% of current ad spend carries forward on its own, so 25 to 30% has to come from new creative.
“roughly 70% to 75% of your current spend carries forward on its own”
CTC's data on 504 stores ($3.35B Meta spend) shows only 7.2% of ads created in 2026 reached $1K of spend and about 1 in 100 reached $10K. A new ad gets about $192 to $397 of spend.
“minimum three production sources active”
Taylor Sicard, building on Motion and other datasets, says only about 5% of creatives become winners and suggests one new ad per $3,000 of monthly spend. A static takes 30 to 60 minutes, a video four to eight hours.
“only around 5% of creatives become winners”
A media buyer at the Japanese agency Opt says Meta treats a duplicated creative as carrying its past wear, so it fades a day sooner when relaunched; add new ads that differ in meaning instead.
In an Ipsos test of 3,000 US people, human-made ads scored 14% higher short term and 17% higher long term than fully AI-made versions. It is survey-based ad testing, not live ad delivery.
“14% stronger on short-term effectiveness”
Motion, a creative analytics vendor, says text-only ads, product images with overlays and GIFs produce more winners than high-production videos or fancy motion graphics.
“Text only ads, product images with overlays and GIFs, they all lead to more winners.”
Motion's 2026 benchmark of 578,750 ads finds text-only and product-image-with-text ads lead on hit rate and spend share, UGC ranks 4th, and carousels and animation rank last.
Motion's 2026 benchmarks of 578,750 ads: offer-first banners had an 8.6% hit rate and took 29.3% of spend on 21.9% of ads; unboxing hit 9.8%, demos 8.1%, testimonials 6.5%.
“Offer-First Banner is strong on both.”
Motion's 2026 benchmarks: accounts spending $200K to $1M a month launch 11.2 new creatives a week on average (top quartile 31.1); over $1M a month, 18.8 (top quartile 54.6).
“To move into the top quartile, the evidence suggests you need to approximately double your weekly testing cadence”
A DTC agency founder with no data of his own suggests spending about 30% of budget on creative below $100K a month and about 20% above it, with 4 to 6 new concepts a month at $30K and 15 to 20 at $150K.
Cody Plofker says partnership ads are now 35% of Jones Road's Meta spend.
MHI Growth Engine, an agency selling creative production, splits budget by annual revenue: 35% creative under $500K, 30% at $500K to $3M, 20 to 25% above $3M. No data shown.
“Creative budget should represent 25-35% of total marketing spend, increasing at earlier stages for testing velocity”
Motion's 2026 benchmarks find winning formats differ by category: health and wellness favors founder, testimonial, transformation and letter formats, while fashion favors Post-it, meme and stylized product shots.
“Fashion's shorter, more impulsive cycle rewards formats that create instant recognition (Meme, Post-It).”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer
Read next
More in Team, tools and account risk- Should we let AI agents or Meta's agentic tools (Ads MCP, API agents) manage campaigns?
- How do we avoid disapprovals, account restrictions and hacked Business Managers?
- How often should we review and change our campaigns?
- What should we pay an agency at our spend level, and how should the fee be structured?