Earlier this year, we scaled a DTC client from $100K/month to $1M/month in ad spend over six months. It was a masterclass in what breaks at scale.
Here's what we learned—the parts that weren't in any playbook.
Context
The client was a consumable product with strong unit economics: 60% gross margin, 3-month average payback, and a subscription model that drove LTV.
At $100K/month, they were printing money. ROAS was 4x. CAC was stable. Creative was working.
The task seemed simple: spend more money.
It wasn't.
Lesson 1: Efficiency Drops Before You Expect It
At $100K/month, we were capturing demand—people who already wanted the product. As we scaled, we had to create demand—convincing people who weren't in-market.
The curve: CAC increased roughly 35% from $100K to $500K/month. It increased another 25% from $500K to $1M/month.
What we learned: Budget the efficiency drop in advance. If your model assumes linear CAC, your model is wrong. We built scenarios with 30%, 50%, and 70% CAC increases at scale. Reality landed at ~50%.
Lesson 2: Creative Exhaustion Accelerates
At $100K/month, a winning ad could run for 8-12 weeks. At $1M/month, winners exhausted in 2-3 weeks.
Why this happens: Higher spend means more impressions. More impressions means faster audience saturation. The same people see your ad more times, and fatigue sets in.
The solution: Creative volume had to increase 4-5x. We went from 10 new concepts per month to 50+. Without this, performance would have collapsed within 8 weeks.
The infrastructure: We couldn't produce 50 concepts with our original process. We built modular production systems, brought on additional creators, and developed template libraries. The production system became as important as the strategy.
Lesson 3: Platform Diversification Isn't Optional
At $100K/month, Meta handled nearly everything. At $1M/month, we were hitting audience ceiling.
The signal: CPMs started rising faster than seasonal norms. Frequency crept up. Retargeting pools shrank relative to spend. The algorithm was running out of people to show ads to.
The solution: We diversified:
- TikTok: 25% of spend (different creative formats, younger skew)
- YouTube: 15% of spend (longer consideration journey)
- Connected TV: 10% of spend (brand lift, audience expansion)
- Meta: 50% of spend (still the workhorse)
Each platform required its own creative, its own strategy, and its own learning curve. Platform expansion is a multiplicative investment.
Lesson 4: Attribution Broke Completely
At $100K/month, last-touch attribution was roughly accurate. At $1M/month, it was fiction.
The problem: With multi-channel spend and brand marketing effects, no single platform could claim accurate credit. Meta claimed 4x ROAS. Google claimed 4x ROAS. Our blended ROAS was 3x. The math didn't add up.
The solution: We shifted to blended metrics—total revenue / total spend—and used platform data directionally, not literally. We also implemented incrementality testing (holdout regions) to measure true lift.
The uncomfortable truth: At scale, you're making decisions with incomplete information. Accepting this and building systems for it beats pretending your data is clean.
Lesson 5: The Brand Started Mattering
At $100K/month, we could win with direct response alone. At $1M/month, brand awareness became a competitive necessity.
Why this happens: When you're reaching millions of people monthly, some percentage are seeing your ads without clicking. Those impressions build (or fail to build) brand associations.
The shift: We invested 15% of spend in brand-building content that didn't optimize for direct conversions. This felt like waste by DR standards. But branded search volume increased 40%, and overall CAC actually improved because warmer audiences converted better.
Lesson 6: Team Structure Changed
At $100K/month, one senior buyer could handle everything. At $1M/month, we needed:
The management overhead increased. Coordination became a job in itself.
Lesson 7: Profitability Required New Mental Models
At $100K/month, simple ROAS calculations worked. At $1M/month, we needed:
The answer to the last question was sobering: LTV of customers acquired at $1M/month was 15% lower than at $100K/month. They were less loyal, less likely to subscribe, and had higher support costs.
This doesn't mean scaling was wrong—it means the economics were different than linear extrapolation suggested.
The Summary
What works at $100K/month:
- 1-2 platforms
- 10 creative concepts/month
- Simple attribution
- Lean team
- Capturing existing demand
What's required at $1M/month:
- 4+ platforms
- 50+ creative concepts/month
- Blended attribution + incrementality testing
- Specialized team
- Creating demand + capturing it
- Brand investment
- Different economic expectations
Would We Do It Again?
Yes—but we'd build the infrastructure earlier.
The scale-up was harder than it needed to be because we kept trying to stretch $100K/month processes. Starting with scalable systems would have made the ramp smoother.
The opportunity was real. The revenue impact was substantial. But the effort required was 3x what we initially estimated.
Scaling isn't just spending more. It's building a different machine.
Every stage of growth requires different capabilities. What got you here won't get you there. Build for where you're going, not where you are.