The demo is table stakes
“Watch it generate an image” stopped being impressive around 2024. Capability demos now prove you exist, not that you’re worth paying for. Everyone’s reel looks identical.
Work Case 03 Consumer AI
148.6 million views
at 29 cents per thousand.
Consumer AI is the most crowded category on the internet. Every app has the same models, the same demo and the same promise. We stopped selling AI entirely, and sold the thing people actually feel instead: the $110 a month they’re already bleeding on five subscriptions.
Case 03 · Consumer AI · 96-day campaign
62Creators cast
4,180Videos shipped
2Creator pools
61 / 39Volume split
01 · Context
A genuinely good product in an undifferentiable market.
AI4Chat bundles 40+ frontier models (GPT, Claude, Gemini, Sora, Midjourney and the rest) into one app with one subscription. Chat, image, video and music generation in a single place. Its own line puts it plainly: “Every AI you’ve heard of. One app.”
The product works. The positioning was the problem.
By 2026, “access to all the AI models” describes about four hundred products. Every competitor has the same demo reel, the same model logos on the landing page and the same promise. When every player in a category says the same true thing, the message stops being information and becomes noise.
And this is a consumer product with a $1 trial converting to $40/month. That price point tolerates almost no CAC. Awareness wasn’t the constraint: differentiation at scale was.
02 · The problem
Feature marketing had hit a wall.
The category has no villain
Great creative needs something to push against. “AI is amazing” has no tension, no stakes and nothing for a viewer to feel. Without a villain there’s no story, and without a story there’s no watch time.
The price point punishes mistakes
At a $1 trial converting to $40, you need volume at very low cost per view. One bad month of $12 CPMs and the whole channel is underwater before retention even enters the conversation.
03 · The hypothesis
Give the category a villain.
The insight didn’t come from AI4Chat’s feature list. It came from its pricing page, where the comparison table quietly noted what the equivalent separate subscriptions cost: around $110 a month.
That number is not a feature. It is a wound. And most of the target audience is currently paying it without ever having added it up.
“I’m not sure what I’m paying for AI each month. I know it’s more than it should be.” The feeling we built the entire campaign on
Campaign hypothesis
Stop selling capability. Start selling the invoice. If a creator opens their own billing page on camera, adds up five AI subscriptions in real time, and cancels four of them, the video is not an advert for AI4Chat. It is a viewer discovering they are being overcharged. The product becomes the resolution to a problem the audience didn’t know they had thirty seconds earlier.
This is loss framing, and it is the most reliably underused mechanism in consumer marketing. People will move faster to stop a loss than to pursue an equivalent gain. The category was saturated with gain messages. We ran the only loss message in it.
04 · Loss framing
We A/B tested the framing before scaling it.
Two message architectures, 320 videos each, identical creators and identical budget. Then we put 84% of remaining production behind the winner.
Gain framing vs. loss framing
Indexed to gain framing = 100 · 640-video head-to-head, weeks 1–3
Gain framing: “everything in one place”
Loss framing: “you’re paying $110 for this”
Retention lifted 87%, but signups lifted 240%. The gap between those two numbers is the whole argument for measuring creative on conversion instead of engagement, loss framing doesn’t just hold attention, it creates urgency to act on it.
05 · Geographic arbitrage
Why we run 675 US and 425 international creators.
This campaign is the reason the network is built the way it is. The two pools do different jobs, and the mistake almost everyone makes is treating them as interchangeable.
Share of volume, cost and conversions by creator pool
All three series are percentages of the campaign total, one common scale
International pool (24 creators)
US pool (38 creators)
View as table
| Pool | Creators | Videos | Cost | Paid conv. | Cost/conv. |
|---|---|---|---|---|---|
| International | 24 | 2,550 | $13,359 | 2,353 | $5.68 |
| United States | 38 | 1,630 | $29,735 | 4,999 | $5.95 |
| Blended | 62 | 4,180 | $43,094 | 7,352 | $5.86 |
The critical detail: cost per conversion is almost identical between the two pools ($5.68 vs $5.95). International creators are not “cheaper and worse”. They are cheaper and lower-converting, in almost exactly offsetting proportion. What the split actually buys you is reach you could not afford at US rates, which then feeds the top of a funnel that US creators close.
Failure mode 01
All-international
Blended CPM drops to about $0.14 and the campaign looks spectacular on a views dashboard. Paid conversions collapse, because purchase intent concentrates in markets with card-on-file habits and the right pricing power. We have watched agencies sell this as a win.
Failure mode 02
All-US
Conversion rate holds, CPM triples to roughly $0.79, and total reach falls by about 58% for the same budget. The funnel is efficient but starved. You end up with excellent unit economics on a volume too small to matter.
The actual rule
International creators are a reach instrument. US creators are a conversion instrument. Set the ratio by what the campaign needs this month, not by what the budget will tolerate. On AI4Chat the winning ratio was 61:39 by volume, on the fintech campaign it was 12:88, because a trading audience will not take a market opinion from outside its own market.
06 · Results
The volume record.
Full-funnel conversion, 96 days
$1 trial converting to $40/month
Paying figures are counted after the $1 trial converts to a full $40 month, trial starts are deliberately excluded, because counting them would roughly quadruple the headline number and mean nothing.
Blended CPM by month as the creator mix optimised
Cost per 1,000 views · lower is better
CPM fell 67% across the campaign without a drop in conversion rate, as losing creators were cut, the international ratio was tuned upward, and winning hooks were redistributed to more creators. Campaigns get cheaper the longer they run. This is the single strongest argument against three-month pilots.
07 · Unit economics
The board slide.
Campaign economics, 96-day window
Creator payouts plus production only
| Total creative spend | $43,094 |
| Videos produced | 4,180 |
| Effective cost per video | $10.31 |
| Views delivered | 148,600,000 |
| Cost per 1,000 views (CPM) | $0.29 |
| Accounts created | 350,100 |
| Paying customers acquired | 7,352 |
| Blended CAC | $5.86 |
| Subscription price | $40 / mo |
| CAC payback period | 5 days |
| Net-new MRR | $294,080 |
| Net-new ARR | $3,528,960 |
| Return on creative spend | 81.9× |
An 81× return is not a number we expect to repeat, and we say so to every prospect who points at it. It is the product of an unusually low CPM, an unusually strong offer, and a category where the loss-framing angle was completely unoccupied. Treat it as the ceiling of what this channel can do, not the expectation.
08 · What it proves
Three transferable conclusions.
1. In a saturated category, reframe rather than out-feature.
AI4Chat’s competitors were all making true, accurate, identical claims. The win came from changing what the video was about, not from finding a better feature to lead with. If everyone in your category says the same true thing, that thing has stopped being a message.
2. Loss beats gain, reliably.
Same product, same creators, same budget: 3.4× on signups purely from framing. Before you spend anything on production, check whether your message is asking someone to chase a gain when it could be asking them to stop a loss.
3. The US/international split is a lever, not a discount.
Used as a cost-cutting measure it destroys conversion. Used as a deliberate reach-versus-revenue allocation it cut blended CPM by 62% at effectively flat cost per conversion. The ratio should change per campaign, and on our fintech work it was almost inverted.
Where this transfers
Consumer subscription products in crowded categories, AI tools, productivity apps, streaming, fitness, finance apps, and anything where the customer is already paying for two or three partial solutions to the same problem.
Next case study
A $9 product that paid back in 15 days.
Airbrush: where paid ads were mathematically impossible, so we made the product the content.
Free distribution audit
What’s the loss your customer is already feeling?
Most consumer products are marketed on the gain. We’ll look at your positioning, your competitors’ creative and your funnel, and tell you where the reframe is.