The Anatomy of a $1M Ad: Why Most Die at $10K Spend

A handful of ads carry the entire account

Across every account I have ever managed at Impremis, the distribution of ad performance is the same. Out of every 100 ads we ship, roughly 60 die before $5K spend. Another 30 plateau between $5K and $50K. Eight or nine cross $100K and earn a real seat at the table. One, maybe, crosses $1M in lifetime spend on a single creative.

Not per campaign. Not per account. One ad. One creative. Seven figures of spend, profitable, sustained for months.

When founders see those numbers they ask the wrong question. They ask what the hook was, what the offer was, what the targeting looked like.

Those are the wrong questions because they assume the $1M ad is a content question. It is not.

It is a structural question. The $1M ad is built differently from the $10K ad, and the difference is not creative talent. It is engineering.

What actually breaks at $10K

The reason most ads die at $10K is mechanical, not creative. By the time you have spent $10K on a single ad, three things have happened:

1. You have served roughly 1-3M impressions, depending on placement and CPM.
2. The fraction of your addressable audience that's seen the ad has crossed the threshold where frequency starts hurting CTR.
3. Meta's Andromeda retrieval system has clustered your ad with the rest of your creative library, capping its share of fresh impressions.

That ad is not bad. It is just running into the structural ceiling that every "good" ad runs into. To break past $10K, the ad has to be built such that those three failure modes do not apply.

That is what we are going to unpack. Five structural traits, each independently difficult, that together produce ads capable of absorbing seven-figure spend.

Trait 1: Distinct from the rest of your own creative

In late 2024, Meta rolled out Andromeda, which fundamentally changed how ads get retrieved and served. The most consequential change for advertisers: Meta now clusters ads it considers similar into the same retrieval node. Those ads compete with each other for impressions instead of getting separate auction eligibility.

Which means if you upload 40 ads with the same creator, in the same setting, with slight script variations, you do not have 40 ads. You have 4-6 ads, and Meta is going to ration impressions across them.

A $1M ad has to be structurally different from your other ads. Not different in copy. Different in format, creator, environment, and value angle, with at least three of those four dimensions changed simultaneously from your existing creative library.

| Dimension | Sample Variants | |---|---| | Format | UGC video, founder-on-camera, static, carousel, podcast clip, mini-doc | | Creator | Customer, founder, expert, athlete, comedian, employee | | Environment | Apartment, manufacturing floor, retail aisle, podcast studio, outdoor | | Value angle | Problem-led, transformation, mechanism, social proof, contrarian |

A $10K ad usually changes one dimension from its siblings. A $1M ad changes three or four. The reason brands struggle to produce these is operational, not creative. Producing genuinely diverse creative requires multiple shoots, multiple creators, multiple locations, multiple production styles. It is a logistics problem more than a copywriting problem.

Which is exactly why it works. The ones who solve the logistics get the moat.

Trait 2: Distinct from every other advertiser in the auction

Being different from your own ads is not enough. The Andromeda system also clusters your ads against every other advertiser's ads. If your creative looks structurally similar to thousands of others in the same category, you are competing in the most crowded lane of the auction.

Think about what most DTC ads look like in 2026. A creator in their kitchen, lo-fi iPhone shot, problem-discovery-skepticism-result-CTA arc. This format works. It is also so saturated that no individual ad in this format can break through past mid-five-figure spend.

The $1M ads I have shipped at Impremis tend to occupy formats that almost no competitor in the category was running. A few examples, anonymized:

• For a supplement brand, a single founder-led static image with a chemical-diagram overlay. Plain. Visually unusual in a feed full of UGC video. Spent over $1.4M lifetime.
• For a telehealth product, a partnership ad with a credentialed clinician filmed inside an actual clinic, not a studio recreation. Crossed $2M.
• For a CPG brand, a documentary-style ad shot inside the manufacturing facility, with the founder narrating the production process. The format was structurally unique because almost no competitor could film at their own factory. Spent past $900K and still climbing.

The pattern is consistent: the format itself, not just the message, is the differentiator. The thing that makes the ad scale is that no competitor can copy it inside 30 days.

I think of this as the Innovation Spectrum:

| Type | Hit Rate | Spend Ceiling | Allocation | |---|---|---|---| | Small iterations (new hook, new copy) | 18-22% | $25K-$75K | 40% of pipeline | | Medium swings (new creator or new setting) | 12-15% | $75K-$300K | 50% of pipeline | | Big swings (entirely new format) | 5-8% | $300K-$2M+ | 10% of pipeline |

Most brands allocate 90%+ of creative budget to small iterations and never produce the swings that could become $1M ads. The hit rate looks scary. The expected value, accounting for spend ceiling, is dramatically higher.

Trait 3: Entertaining enough to earn an algorithmic discount

Meta has been public about the fact that creative quality accounts for the majority of campaign outcomes. The auction reflects this. Ads with higher engagement (hook rate, hold rate, shares, saves, comments) win more auctions at lower CPMs. Ads with lower engagement pay a creative tax.

This is not a soft factor. It is a hard discount on every impression. A $1M ad almost always sits in the top decile of engagement metrics for the account. The numbers I look for:

• Hook rate (3-second views / impressions): 35%+
• Hold rate (15-second views / 3-second views): 25%+
• Engagement rate (likes + comments + shares + saves / impressions): 1.5%+

Ads that hit these thresholds compound. They get more impressions per dollar. They generate organic shares that are functionally free distribution. On one of the supplement brand $1M ads I mentioned, we estimated 18% of total reach came from organic sharing. That is roughly $180K of distribution Meta did not charge for.

The formats that earn this kind of engagement share three traits:

1. Pattern interrupt in the first 2 seconds. Not a logo. Not a product hero shot. Something visually unexpected that interrupts the scroll.
2. Genuine entertainment value. The ad has to work as content even if you took the CTA out. Watchable for its own sake.
3. Replay tolerance. Some formats absorb repeat exposure better than others. A funny street interview, a musical doctor explainer, a documentary segment. People will watch them multiple times. The ad's effective frequency is lower than its reported frequency.

A mediocre script in an unusual format outperforms a perfect script in the same talking-head setup everyone else is running. Format eats script. Always.

Trait 4: Wide TAM, not narrow product focus

This is the trait that most cleanly separates the $100K ad from the $1M ad.

The Eugene Schwartz awareness ladder maps almost perfectly to Meta's audience math. Most of your potential market is unaware they have the problem your product solves. A smaller fraction is problem-aware but solution-unaware. An even smaller fraction is solution-aware but doesn't know your brand. Only a tiny sliver is product-aware and ready to buy.

Product-focused ads ("Use code SAVE20 on our starter kit") only address the bottom of that ladder. That audience is small. It saturates fast. The ad cannot absorb large spend without burning through everyone who might possibly care.

Problem-aware ads ("Are you sweating through your shirts every day? Here's why this happens") address the top of the ladder. That audience is 8-10x larger.

Larger addressable audience means:

• Lower frequency buildup at the same spend level
• More room for Meta to find cheap inventory pockets
• Slower fatigue curves
• Longer creative lifespan
• A higher spend ceiling before saturation

The $1M ads I have shipped all led with the problem, the world the customer is living in, the pain point. The product was the resolution, not the hook. That structural choice is what allowed Meta to deliver them to a wide audience without saturating.

This does not mean make your messaging vague. It means make it specific to a problem rather than specific to a product. The specificity moves up the awareness ladder, and the addressable audience grows by an order of magnitude.

Trait 5: Intentional over-scale beyond what last-click suggests

The final trait is operational, not creative. A $1M ad cannot scale to $1M without an operator who is willing to fund it past what last-click attribution says is profitable.

Last-click systematically underweights top-of-funnel ads. The whole job of a wide-TAM, problem-led, entertaining ad is to create awareness. The conversion happens later, on a different surface, often through a retargeting ad or a Google search or a direct visit. Last-click gives the credit to whoever fired closest to the purchase.

If your target ROAS is 2.5x and your $1M-candidate ad is running at 2.0x last-click, the wrong move is to kill it. The right move is to fund it 20-30% above what attribution says is profitable, because the halo effect is real and measurable.

The rule I use: if an ad checks four of these five boxes, over-scale it.

• Hook rate > 35% and hold rate > 25%
• CPM consistently below account average
• Wide-TAM, problem-led messaging
• Format that is structurally unique to the auction
• Last-click ROAS within 25% of target (not necessarily above)

If you have access to incrementality testing through geo-lift or holdout cells, validate the over-scale empirically. We do this on every Impremis account spending more than $100K/month. The true ROAS of high-engagement TOF ads is consistently 30-50% above last-click. Your retargeting CPA falls. Your branded search rises. Your blended margin lifts even though the individual ad's last-click number looks soft.

This is one of the reasons I keep coming back to the question of why ROAS is not the goal. If you optimize for the last-click number, you starve the ads that are doing the actual heavy lifting.

The five traits combined

A $1M ad is not five separate things. It is one structurally engineered asset that satisfies all five conditions at the same time. The reason they're rare isn't talent. It's that each requirement is independently hard, and they have to compound.

| Trait | Failure Mode If Missing | |---|---| | Distinct from your own creative | Andromeda clusters it; impressions get rationed | | Distinct in the auction | Competes in the saturated lane; CPM stays high | | Entertaining | No engagement subsidy; pays full CPM tax | | Wide TAM | Audience saturates inside $50K of spend | | Over-scaled past last-click | Killed before the halo effect compounds |

Miss any one of the five and the ad caps out somewhere between $50K and $300K. Hit all five and the ad has the structural ability to keep absorbing budget for months.

This is also why production matters more than ideation. The ideation is the easy part. The hard part is building the operational machine that produces structurally diverse, format-distinct, problem-led, entertaining creative on a weekly cadence. Most brands cannot. The ones that can build the moat. (See why ads stop working: creative fatigue for the supply-chain side of this argument.)

FAQ

How long does it take to actually produce a $1M ad?

The ad itself usually takes a few weeks of production. The system that consistently produces ads with $1M potential takes 6-12 months to build. The bottleneck is operational maturity, not any single shoot.

Can a static image really cross $1M?

Yes. I have shipped multiple. Statics work especially well when the visual is itself the differentiator and when the brand has a tangible USP that is hard to explain in video. The key is that the static occupies a unique format lane in the auction, not that it's better designed than the next static.

What's the realistic batting average for big swings?

5-8% become genuine winners. Of those winners, maybe one in ten crosses $1M. So roughly 0.5-0.8% of big swings produce $1M ads. The arithmetic is brutal until you compare it to the ceiling of small iterations, which is much, much lower.

How do I balance big swings against keeping the lights on?

40% of creative pipeline goes to small iterations and reliable refreshes. 50% to medium swings. 10% to genuine big swings, even though they cost more per unit. This split keeps daily performance steady while still producing the asymmetric upside.

Does this apply to TikTok and YouTube too?

The principles transfer. The thresholds shift. TikTok rewards even more format novelty and fatigues faster. YouTube tolerates longer-form storytelling and replays better. The five traits still hold; the calibration changes.

What's the single biggest mistake brands make trying to build $1M ads?

Producing 30 "different" variations of the same ad and calling it a portfolio. If a competitor cannot tell two of your ads apart at a glance, neither can Andromeda, and neither can the audience. Variation is not differentiation.

How do I know when to over-scale vs. let the ad die?

Run the five-trait checklist. Four out of five with last-click ROAS within 25% of target is the over-scale signal. Three out of five or worse, let it run normally. Two or fewer, the ad is not a $1M candidate and over-scaling will not save it.

The moat is the production system

The brands that build $1M ads repeatedly, not as luck but as a process, share one thing in common. They have a creative production system that is engineered to produce structurally distinct, format-unique, entertaining, wide-TAM creative on a weekly cadence.

The ad is the output. The system is the asset.

Most competitors will never build the system because it is operationally expensive, slow to mature, and doesn't show up in any quarterly review. That is the moat.

Difficulty is the moat.

If you are willing to build for it, the algorithm will reward you with ads that print money for months while everyone else is wondering why their CAC keeps creeping up.

Keep reading

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

• Creative Fatigue Is the Most Expensive Problem Nobody Measures — Your ads aren't broken. They're tired. Here's how to diagnose, quantify, and fix creative fatigue before it eats your CAC.
• The Power Law of Ad Creative: Why 5% of Ads Do 95% of the Work — Meta isn't a normal distribution. It's a power law. Stop spreading budget evenly and start treating outliers like the only thing that matters.
• What Actually Works on Meta in 2026: A $100M+ Playbook — Across $100M+ in personal Meta spend and $250M+ at Impremis, here's the creative format playbook that's working post-Andromeda.
• 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 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.