Ramp's 70/30 marketing split and why AI won't kill creativity

Apr 13, 2026 · Exit Five
🎧 PodShort 31 min squeezed to 2 ARMARM Marketing Automation New
Episode artwork
Drew Pinta
Head of Growth Data at Ramp
Exit Five
31 min squeezed to 2
Full episode from Exit Five
Quotable Moments

The fastest way to kill creative marketing ideas is to try to measure them like direct response.

Drew PintaHead of Growth Data at Ramp
10:10

I strongly believe that the measurement should meet the marketing, not vice versa. I think too often people try to force the marketing to fit into the measurement.

Drew PintaHead of Growth Data at Ramp
10:30

AI and LLMs are trained to give you the median output. That's how these things are built. They're statistical machines and they're spitting out the average or the median. And in marketing, the median is death.

Drew PintaHead of Growth Data at Ramp
25:00
Key Insights
  • The fastest way to kill creative marketing ideas is to try to measure them like direct response, as measurement should meet the marketing where it is, not the other way around.
  • Ramp's marketing budget is split 70/30, with 70% allocated to proven core channels and 30% to experimental channels for future growth, acknowledging that new channels may take 6-12 months to mature.
  • AI and LLMs are built to provide median outputs, which is detrimental in marketing where uniqueness and creativity are key for standing out, suggesting a continued need for human creativity.
  • 80% of what marketers currently do is likely to be automated by AI, necessitating a shift from running campaigns to building AI agents that execute marketing tasks at scale.
  • Ramp's marketing team has developed a 'vertical machine' using AI to automate the process of spinning up marketing efforts for new industry verticals, including research, SEO, and sales enablement.
  • To enable AI adoption, leadership must explicitly support employees in learning new tools and workflows, even if it means a temporary dip in performance, fostering a 'builder' mindset.
  • Incrementality testing, while expensive and slow, is the gold standard for understanding the causal impact of marketing spend, especially for channels where direct attribution is difficult.
  • Marketers are in a better position than engineers in the age of AI because human creativity and taste remain crucial for marketing success, whereas for engineers, average code is often sufficient.
Metrics Mentioned
  • 211 days (Average buying cycle in B2B.)
  • 22 people (Average number of people involved in a B2B buying decision.)
  • 70% sales-led (Ramp's go-to-market motion.)
  • 3x more (LinkedIn was mentioned 3x more than Meta in sales calls, despite MTA models showing Meta as more effective.)
  • 70/30 split (Ramp's marketing budget allocation between core channels and experimental bets.)
  • 6-12 months (Typical timeframe for a new marketing channel to mature at Ramp.)
  • 80% (Estimated percentage of current marketing tasks that will be automated by AI.)
  • 1 million a month (Amount a company was spending on Google Ads before their account was locked down due to AI interaction issues.)

RevBots.ai View:

  • The 70/30 budget split is classic AI Sprinkler behavior: testing automation while preserving human-led initiatives.
  • Vertical machine automation shows ARM potential but still requires manual oversight.
  • Tab Hoppers take note: incrementality testing is expensive but necessary for accurate attribution.
  • SaaS Hoarders will recognize the tool sprawl challenge in reconciling MTA models with real sales data.
🎧Full Episode:Exit Five →