Podcast Summary: Perpetual Traffic

Episode: The Future of Media Buying: John Moran on cAPI Imports + Edge Tagging (Part 2)

Date: October 3, 2025
Host: Ralph Burns (Tier 11)
Guest: John Moran

Episode Overview

This episode dives deep into the future of media buying with a focus on how artificial intelligence (AI), modeled data, Meta’s Andromeda engine, cAPI imports, edge tagging, and content diversification are radically reshaping how performance marketers optimize paid media. John Moran shares actionable insight on attribution accuracy, creative strategy, and why marketers now need to build ads for AI first—rather than humans.

Key Discussion Points & Insights

1. From Human-First to AI-First Creative

• “We don't create ads for people anymore. We have to create ads for AI first.”
  (John Moran, 00:00; 25:25)
  The digital ads ecosystem now requires marketers to prioritize how AI systems (especially Meta’s Andromeda engine) perceive and process creative, which then guides delivery and results.

2. Understanding Modeled Data and Attribution Shortcomings

• Definition and Importance:
  Modeled data refers to conversions "filled in" by AI algorithms using probabilistic matching, not actual user action, leading to potentially inaccurate stats within Meta and Google Ads dashboards.
• Optimistic vs. Pessimistic Views:
  • Optimistic: Platforms try to fill in missed conversions to help advertisers.
  • Pessimistic: Platforms over-credit themselves to capture more ad spend.
• Real-World Example:
  • A large promotional company ran a Meta campaign, spending $120 and supposedly generating 4,000 sales—a result driven largely by modeled/over-attribution, not real performance.
  • Quote: “We launched a meta campaign. We spent $120. We had 4,000 sales... we have like a 62 million robots.”
    (John Moran, 03:57)
• Attribution Issues Across Channels:
  In platforms like Shopify, email is often incorrectly credited as the first click—demonstrating how “every channel claims every sale.”
• Quote: “Every campaign is like TikTok, I have 10 sales. Meta, I have 10 sales. And Google, I have 10 sales. And you only have 10 sales.”
  (John Moran, 11:12)
• Summary: True business growth is only reflected in actual backend metrics (like Shopify sales)—not just platform-attributed conversions.

3. How Much of Ad Platform Reporting is Modeled?

• Estimated Proportion:
  • About 40% of Meta/Google ad platform reported conversions are modeled data; up to 60% for video-heavy campaigns (07:38–08:50).
  • Quote: “Forty to sixty percent of what you see inside the ad platforms is modeled. It’s not actual data.”
    (Host + John Moran, 08:50)
• Implication: Media buyers must be cautious trusting platform data at face value.

4. Why ROAS Alone Is Misleading

• ROAS Critique:
  Relying solely on in-platform Return on Ad Spend (ROAS) is dangerous—it's essentially a "directional" metric, not truth.
  • “ROAS sucks... because it’s an in-app metric, it’s directional” (John Moran, 11:26)
• cAPI Import / Tier11 Data Suite:
  The team’s own method pulls first-click, new customer events via edge tagging and cAPI, getting closer to real, de-duplicated attribution.

5. The Challenger: Content Diversification (Creative Iteration)

• Major Theme at Meta’s 2025 Agency Conference:
  True scaling depends on creative diversification—testing different types, angles, and stories in coordinated sequences.
• Quote: “Every single speaker talked about diversity of creative…have lots of creator content, obviously video content. You name it.”
  (Host, 12:03)

How to Build Effective Creative:

• Andromeda Engine’s Role:

  • Meta’s new Andromeda update (rolled out by May 2025) uses a “creative retrieval” process to decide which ads to serve and how to sequence them.
  • Ad accounts should consolidate campaigns and test many variants within a single ad set—not spread them across multiple campaigns.

• The Ad Sequence:

  • Use founder stories, product highlights, UGC, testimonials, competitive comparisons, and more.
  • The AI will select and sequence these based on what it "thinks" will convert each user.

• Quote: “You’re basically building a logical sequence to a sale inside of an ad set that the Andromeda engine can start to show this ad first, second, third, fourth. Or maybe it needs to show the second ad first...”
  (John Moran, 17:05)

Creative Fatigue & Iteration:

• Meta auto-allocates spend to top-performing ads, lowering spend on underperformers.
• Rule of thumb:
  When an ad gains most of the spend, iterate new hooks on that concept rather than pausing it—capitalize on algorithmic “pockets of gold.”
  • Quote: “If you know, you might—a lot of people just say, oh it's spending too much money, let's pause it. No, you want to make five new versions of that.”
    (John Moran, 22:32)

6. Ad Spend, Attribution, and “Follow the Money”

• Attribution Problems:
  With last-click models, only the last touched ad gets credit, even if the sales process was multi-step and cumulative.
  Example—The final, trivial detail (e.g. “it also comes in gray”) could get attribution, ignoring prior impactful messages.
  • Quote: “Because honestly the last thing may not really have much of an impact at all on the decision making process of that person.”
    (John Moran, 24:16)
• Solution:
  Focus on creative concepts, hooks, and iterations—their ability to win more ad spend is a better indicator than CPA alone.

7. Building for the Algorithm, Not the User

• Shift in Mindset:
  Ads must be explicit and clear for the AI, not just appealing to people. “We have to convince Meta that this is what product we have. And that Meta says, okay. Understand, I will go find the user now.”
  (John Moran, 25:32)

Notable Quotes & Memorable Moments

| Timestamp | Speaker | Quote | |-----------|---------|-------| | 00:00, 25:25 | John Moran | “We don't create ads for people anymore. We have to create ads for AI first.” | | 03:57 | John Moran | “We spent $120. We had 4,000 sales... we have like a 62 million robots.” | | 08:50 | Host + John Moran | “Forty to sixty percent of what you see inside the ad platforms is modeled. It’s not actual data.” | | 11:12 | John Moran | “Every campaign is like TikTok, I have 10 sales. Meta, I have 10 sales. And Google, I have 10 sales. And you only have 10 sales.” | | 11:26 | John Moran | “ROAS sucks... because it’s an in-app metric, it’s directional.” | | 15:09 | Host | “We'll leave links in the show notes for this one. It's a great article. It's actually a great image.” | | 17:05 | John Moran | “You’re basically building a logical sequence to a sale inside of an ad set that the Andromeda engine can start to show this ad first, second, third, fourth. Or maybe it needs to show the second ad first...” | | 22:32 | John Moran | “If you know, you might—a lot of people just say, oh it's spending too much money, let's pause it. No, you want to make five new versions of that.” | | 24:16 | John Moran | “...the last thing may not really have much of an impact at all on the decision making process of that person.” |

Timestamps for Key Segments

• 00:00–08:50: Modeled Data, Attribution Challenges, How Platforms “Overcredit”
• 08:50–11:30: True Incrementality vs. Platform Reporting
• 11:30–14:50: ROAS Limitations and the Rise of cAPI/Edge-Tagging Attribution
• 14:50–18:40: Creative Diversification and Meta’s Andromeda Engine
• 18:40–25:47: Sequencing Creative, Iteration, and Interpreting CPA in Context
• 25:25–27:00: Building Ads for AI, not People
• 27:00–31:20: Next Steps, Where to Learn More, Wrap-up

Action Items & Takeaways

• Trust—but verify—platform-reported conversions. Check backend sales data to measure real impact.
• Prioritize creative diversity. Build ad sets with founder stories, product features, UGC, and more to enable the AI to find angles that work.
• Focus on iteration. When spend accrues to a particular ad, test variations on that theme rather than pausing it.
• Leverage consolidated campaign structures. The Meta Andromeda engine works best with single-campaign setups for sequenced creative testing.
• Always build ads for the algorithm. Clearly communicate benefits and product details so AI can identify and target the right users.

Where to Learn More

• John Moran LinkedIn: Best way to contact John.
• Tier 11 YouTube – The Ad Lab (Fridays, 2:30pm Eastern): Case studies and advanced campaign breakdowns.
• Tier 11 Services: For hands-on help, visit tier11.com/apply.

For digital marketers, this playbook—adopted by agencies spending millions a month—offers ground truth on campaign measurement, creative best practices, and evolving platform AI. The future of scaling paid media is not just about better targeting or bigger budgets, but about harnessing AI-driven attribution, precision creative iteration, and true business-impact measurement.