What Actually Works on Meta in 2026: A $100M+ Playbook

Creative is your targeting now

Meta in 2026 is a different platform than Meta in 2022. The October 2025 Andromeda update finished a transition that started two years earlier: targeting is no longer where the leverage is. Audiences are mostly broad. Detailed interest stacks are mostly noise. Lookalikes have been quietly demoted by the algorithm itself.

What replaced all of that is creative. The ad itself is the targeting signal. Meta reads the content of your ad, the visual style, the format, the implicit audience your creator embodies, and routes it to people most likely to engage. The creative is the audience.

This sounds obvious until you realize most brands are still running their accounts like it's 2020. They are spending hours optimizing ad sets, layering interests, building lookalikes, and producing the same UGC talking-head video they have been producing for three years.

That playbook is dead. This is the new one.

Across more than $100M in personal Meta spend and over $250M annually at Impremis, here is what is actually working in 2026, what to produce, in what ratio, at what volume, with what testing structure, and how to know when something has crossed into scale.

The four formats that account for 80% of profitable spend

In 2026, four creative formats consume the majority of profitable Meta spend across the accounts I run. Everything else is testing, refresh, or experimentation around the edges.

1. UGC testimonial video

Still the workhorse. A real person, smartphone-quality production, narrating their experience with the product. Lo-fi production beats studio production on Stories and Reels placement about 85% of the time in the data I see.

Four sub-formats inside this category that are working in 2026:

• Talking-head problem-solution. Person to camera, problem first, then product, then result. Still effective if the creator is genuinely diverse from your last shoot.
• Unboxing reveal. First-time experience captured in real time. Works best for products with a strong physical component (skincare, supplements, tech).
• Day-in-the-life integration. The product appears as part of a routine, not as the focus. Subtle and increasingly powerful.
• Authority partnership. Doctors, dermatologists, athletes, credentialed experts on camera. We see CAC drop 40-55% on partnership ads versus comparable UGC, consistent across categories.

2. Static images

Still underrated. On most Impremis accounts, static images deliver between 50% and 70% of conversions despite being a smaller share of the creative library. The reason: they occupy a different lane in the auction than video, they fatigue more slowly at low-to-mid spend, and they communicate quickly in feeds where attention is fragmenting.

Four sub-formats that are scaling:

• Product hero with overlay claim. Single product, single dominant claim, minimal copy.
• Comparison grid. Your product vs. category alternative, side by side.
• Social-proof card. Verbatim review with star rating, treated as design.
• Mechanism diagram. How the product works, simplified, sometimes with a chemical or anatomical visual.

3. Founder / expert long-form video

Mid-form video, 60-90 seconds, hosted by a founder or credentialed expert, often shot in an authentic environment (manufacturing facility, podcast studio, clinic). These ads have the longest creative lifespans in our portfolio. Some have run profitably for 6+ months without major refreshes.

The format is hard to copy because it requires a brand asset most competitors don't have: a charismatic founder or a real expert relationship. That difficulty is the moat.

4. Carousel

Underused. For products with comparison, education, or step-by-step value, carousels routinely outperform video on cost-per-acquisition. Each card carries one beat of a narrative: problem, mechanism, social proof, offer. Five to seven cards is the sweet spot.

How much of each to produce

A reasonable allocation, calibrated against accounts spending $100K-$500K/month:

| Format | Share of Production | Share of Profitable Spend | |---|---|---| | UGC testimonial video | 50% | 35-45% | | Static images | 25% | 25-35% | | Founder/expert long-form | 10% | 15-25% | | Carousel | 10% | 5-15% | | Experimental (mini-doc, podcast, rap, street interview) | 5% | 0-20% |

The experimental row is the one most brands skip. It is also where the asymmetric upside lives. The single highest-LTV ad I shipped at Impremis last year was a documentary-style mini-doc shot on the manufacturing floor of a CPG brand. It absorbed over $900K in spend at a 4.1x ROAS while the rest of the account was averaging 2.5x. That ad never would have existed if the brand had spent 100% of its production budget on UGC.

Production volume: the only formula that matters

The single most useful heuristic for production planning at scale: one new ad concept per $10,000 in monthly spend, one new variation per $3,000. Below that throughput, you fatigue faster than you replenish. The accounts that scale past $1M/month in profitable spend are the ones that solve the production-volume problem first.

A quick math check by spend tier:

| Monthly Spend | Concepts Required | Variations Required | |---|---|---| | $50K | 5 | 17 | | $150K | 15 | 50 | | $300K | 30 | 100 | | $500K | 50 | 167 | | $1M | 100 | 333 |

Most brands spending $500K/month are producing fewer than 30 ads per month. Their CAC is climbing because their supply chain is mathematically too small to feed the spend.

This is the same supply-side argument that drives creative fatigue. Volume is not a vanity metric. It is the rate at which your account can stay alive.

The 3-3-3 testing structure

For every new concept, we run a 3-3-3 framework: three hooks, three visual variants, three CTAs. That produces 27 combinations per concept, but Andromeda will collapse most of them into 4-6 retrieval clusters, which means you're really testing across that many distinct experiences.

Decision rules:

• Give each concept a minimum of 7 days and 10-12 conversions before judging.
• Cut underperformers at 2x target CPA with no conversions.
• Scale winners by duplicating into fresh placements, not by raising budget on the existing ad set.
• Move budget changes in 20% increments to avoid resetting learning.

The 7-day minimum is non-negotiable. Andromeda needs that long to find an audience pocket for new creative. Cutting at day 4 because the ROAS is soft is the most common reason brands kill ads that would have become winners.

Hook frameworks that are still working

The first 2-3 seconds determine whether an ad gets a chance to work. Across thousands of tested ads, the hooks that consistently break through in 2026:

• Direct problem callout. "If you're sweating through your shirts every day, this is for you."
• Counter-narrative. "Everything you've been told about [category] is wrong."
• Cost reveal. "Here's what we actually pay to make this product."
• Time-bound result. "After 30 days, here's what happened."
• Audience qualification. "This isn't for everyone. If you're [X], keep watching."
• Insider admission. "I worked at [category leader] for five years. Here's what nobody tells you."
• Mistake reframe. "I was using [product type] wrong. So is everyone else."
• Visual pattern interrupt. Something genuinely unexpected on screen in the first second. Not a logo. Not a product hero shot.

The targets I shoot for:

• Hook rate (3-second views / impressions): 35%+
• Hold rate (15-second views / 3-second views): 25%+

Ads below these thresholds rarely scale. Ads above both routinely cross $100K in lifetime spend. (For the structural anatomy of the rare ads that cross $1M, see the anatomy of a $1M ad.)

Format specifications and placements

The spec game has consolidated. The default is now:

• Master format: 9:16 vertical, 1080x1920. Reels and Stories are the highest-volume placements. Build for them first.
• Secondary: 4:5, 1080x1350. Best for Feed; takes 15% more screen real estate than 1:1.
• Tertiary: 1:1, 1080x1080. Multi-platform safety, lowest impact per impression.

In 2026 I rarely produce 16:9 horizontal anymore unless we have a specific YouTube cross-post in mind. The Reels-first economy is just too dominant.

What changed with Andromeda, and what to do about it

A few practical implications of the post-Andromeda world that operators are still adjusting to:

Creative diversity matters more than account architecture

In the pre-Andromeda era, the smart play was tight ad set architecture, heavy interest layering, manual placement control. In the post-Andromeda era, the algorithm groups ads by content similarity across the entire system. That means the only architecture that consistently buys you incremental impressions is creative-level diversity. Two genuinely different ads will outperform two ad sets running similar ads, every time.

Broad targeting wins by default

Advantage+ Audience or simple broad with light geographic targeting is the right starting position for most accounts. The algorithm has more signal from the creative than from any audience layer you could specify. The exception: brands with non-obvious customer fits where the model needs a hint. Even then, lean broad first, narrow only when broad confirms it cannot find the audience.

Old creative dies faster

The Andromeda clustering means once your library is mapped, new ads with similar features inherit the fatigue of their predecessors. This is why the production-volume formula matters more in 2026 than it did in 2023. The penalty for under-producing is steeper.

Learning phase resets are expensive

More than 30% budget changes restart learning. Significant audience changes restart learning. Even structural changes to the campaign objective can trigger it. The discipline of 20% increments and surgical changes pays off in stable delivery.

Production budgets by spend tier

These ranges hold across the brands I work with at Impremis and the ones we advise at AdFuse:

| Monthly Ad Spend | Monthly Creative Budget | Concepts Per Month | |---|---|---| | $25K-75K (founder-led, lean) | $1,500-$4,000 | 8-15 | | $75K-200K (small team) | $5,000-$15,000 | 15-30 | | $200K-500K (scaled) | $15,000-$40,000 | 30-60 | | $500K+ (mature) | $40,000-$100,000+ | 60-120+ |

Brands below this ratio for their tier are creative-starved and almost always have rising CAC. Brands above this ratio for their tier are typically over-investing in production polish at the expense of variety.

Detection: knowing when an ad is fatigued

The early warning signals I use:

• CTR drops below 1.0% on a previously-strong ad.
• Week-over-week CTR decline of 20%+.
• Frequency on prospecting ad sets above 3.0.
• CPA rising while CPM is flat (the fatigue signature; if CPM is rising too, it's an auction-wide issue, not creative).
• Cohort decay chart showing your top-3 ads delivering more than 60% of spend.

When two or more of these fire on a single ad, it's done. Replace it before it costs you another two weeks of inflated CAC.

FAQ

Is UGC dead in 2026?

No. It's the largest single category of profitable spend. What's dead is undifferentiated UGC. The same talking-head format from the same kind of creator with the same 5-beat script. Diverse UGC, with varied creators, formats, and angles, still works at scale.

Should I use Advantage+ Shopping (ASC) or manual campaigns?

Both. ASC for proven creative and broad audiences where you trust the optimization. Manual for new tests, specialized audiences, and any segment where blended numbers diverge from platform numbers. See the three-lever Meta playbook.

How long should I let a new ad run before judging it?

7 days minimum, 10-12 conversions ideally. Cutting earlier means cutting on noise. Andromeda needs the time to find audience pockets.

What's the right ratio of static to video?

Start at 70/30 video/static for most categories. Track which is delivering profitable conversions and rebalance every 30 days. Many of the most efficient accounts I see are closer to 60/40.

Are AI-generated ads working yet?

As hybrid components, yes. AI-generated backgrounds, voiceover variations, and scripted variants paired with human creators are producing competitive results. Pure-AI creative without human touchpoints still underperforms hybrid by a wide margin.

How do I produce 50+ ads a month without burning out my team?

Production-flow systems. Weekly creator check-ins, scheduled shoot days, modular asset libraries that can be remixed. Do not run creative as a project model where each batch is a separate sprint. Run it as a manufacturing line.

What about TikTok and YouTube?

Different formats, same principles. TikTok rewards even more native production and faster pacing. YouTube tolerates longer-form storytelling. The four-format framework holds; the calibration shifts. We test cross-platform but treat each platform as its own creative system.

Should I copy what big brands like AG1, Rhode, and Gymshark are doing?

Learn from them, don't copy them. They have brand equity that allows formats you can't replicate. Adapt their structural decisions (creative diversity, expert partnerships, volume) and ignore the surface-level aesthetics.

The 2026 operating model in one paragraph

Produce one concept per $10K of monthly spend, in a 50/25/10/10/5 mix of UGC video, statics, founder long-form, carousels, and experimental. Run them through a 3-3-3 testing structure with 7-day minimums. Cut at 90% confidence below target. Scale by duplicating winners into fresh placements, not by raising budget on the existing ad set. Stay in 9:16 master format. Trust the algorithm to handle audiences. Mistrust the algorithm on attribution. Measure with a stacked attribution model, not with platform-reported ROAS alone.

Creative is the targeting. Production is the moat. Discipline is the multiplier.

Brands that internalize those three sentences will outperform brands that don't, by a margin that compounds with every quarter the algorithm gets smarter.

Keep reading

Pieces I've written on related topics that pair well with this one:

• The Creative Fatigue Playbook: Predict When a Meta Ad Is Dying Before It Kills Your ROAS — Meta ad creative fatigue is predictable — if you know which signals to watch.
• The Creative Learning Phase: What Meta Is Actually Doing During Those First Seven Days — Most teams panic or ignore the Meta learning phase. Here's what the algorithm is actually calibrating — and how to structure creative launches around…
• Why Your Best Ad Will Fail in 30 Days (And How to Stay Ahead of It) — Creative fatigue follows a predictable 30-day decay curve. Here's the early warning signals to watch and the 5-step system to stay ahead of it.
• The Meta Manual Controls That Still Matter in 2026 — Automation is Meta's default, but it has specific failure modes.
• Long-Form Ads Are Working on Meta. Volume Is Still a Trap — Why 5-minute and 14-minute ads are outperforming on Meta, and why producing 100 ads a month is the wrong response to it.