Marketecture Podcast Episode 168 Summary: A Guide for Using AI Agents in Media Analytics, with Newton Research’s John Hoctor

Overview In this episode, host Ari Paparo and co-host Eric Franchi sit down with John Hoctor, Co-Founder and CEO of Newton Research (and former CEO of Data + Math). The discussion dives deeply into the evolution of AI agents in media analytics, exploring the practical realities of deploying agentic AI for brands, agencies, and publishers. Hoctor offers in-depth technical and strategic insights into the use of AI agents for analytics automation, incrementality testing, and the future of marketing measurement. The conversation is rounded out with analysis of major industry news, including OpenAI’s acquisition of TBPN, MediaOcean/Disney direct buying, MIQ’s latest acquisitions, and AI breakthroughs from Meta and Anthropic.

1. Introducing John Hoctor & Newton Research

• Who is John Hoctor?
  • Hoctor is lauded by both hosts as “scary smart” and “the smartest guy in ad tech.” (03:08)
  • Ari: "It’s almost universal. So what sort of PR do you have, John?" (05:34)
• Company Origins & Name
  • Newton Research is named both for Isaac Newton and Newton, MA, where the founders live.” (06:44)
  • Hoctor jokes: "We are, I think, historically bad at naming companies. Data plus Math was our last one, so I kind of see a trend." (06:44)
• Founding Story:
  • Newton’s founders had an early start: “We were very lucky to be on the sidelines when ChatGPT was unveiled… we started from scratch.” (07:50)
  • Inspired by the REACT framework for Reason and Action agents—Hoctor: “Matt read an academic paper on this REACT framework… we started building agents that could do media analytics, because that’s the analytics that we know.” (08:11)
  • First shown at RampUp three years ago, progression of the industry from confusion to universal agent adoption: "First year nobody knew what we were talking about. Second year, everyone was agent curious. And this one, every single ad tech company was talking about their agents." (09:41)

2. How Do AI Agents Work in Media Analytics? (Technical Dive)

• Deployment and Data Access
  • Newton agents are “containerized,” run in the cloud, and access data where it resides (e.g., AWS, GCP, Azure, Databricks, Snowflake).
  • Hoctor: “Our agents act like data scientists opening up the Jupyter Notebook. That’s a really good analogy.” (10:57)
• Setup & Context
  • Agents require understanding of client data schemas and files: “It can understand files… when you’re having a conversation with Newton…” (12:07)
• Agent Expertise
  • Newton builds agents with domain knowledge bases:
    • Hoctor: “We try to give these agents master’s degrees in marketing science. We have an agent that’s really good at MMM, incrementality, lookalike modeling…” (12:45)
    • Emphasis on reproducibility: “They (clients) don’t like curious choices when they’re doing analytics… you want it to be enterprise grade, you want to trust it, you want to get the same answer if you ask the same question tomorrow.” (13:37)
• Prompt Engineering & Methodology
  • Success depends on well-designed prompts and methodology encoding:
    • “If you provide sample code, if you provide all the way down to like, ‘No, this is how I want you to do it,’ and then you leverage the LLM for writing code… you get a much better result.” (15:10)
• Automation & Multi-Agent Workflows
  • Newton supports repeatable, automated multi-step analytics:
    • “Not like starting from scratch and typing in what you want every single day… that’s why we have buttons.”
    • Multi-agent workflows speed up tasks that take humans hours or days:
      “…when you set up these multi agent workflows, the agents know where the data lives, they know what the data means, they can do a lot of that work and then the human can spend the time doing what humans do best: being insightful and figuring out, ‘OK, interesting, what does this mean?’” (16:06–17:38)

3. AI Agents & Incrementality Testing (19:47–20:47)

• Practical Benefits
  • Newton automates the tedious parts of incrementality test setup and measurement:
    “Now it’s easy to set them up and easy to measure them… you’re entering a world with unlimited analytics.” (17:52)
• AI Is Not “Set and Forget”
  • Limitations remain:
    “You don’t just go into the machine and walk away and go do other stuff… there’s still heavy lifting involved. But we try to make the job easier, more fun, try to take a lot of the steps that were cumbersome and make them easy.” (20:10)

4. Cutting Edge: Causal Models and New Data Sets (21:03–23:37)

• Moving Beyond MMM
  • Hoctor explains causal modeling: combining more data and neural nets to analyze real causality, not just correlation.
    • “There’s a whole new generation of models coming out, causal just being one category. If you could analyze your marketing differently because of what technology has brought us, how would you do it?” (22:38)
• Example Data Sets
  • Blending martech and adtech data; e.g., customer journey, call center info, brand awareness.
  • “You can really make this primordial soup and see what life comes out of that soup.” (23:11)

5. Practical Advice for Marketers Dabbling in AI (24:43–26:35)

• Prompting isn’t Enough for Enterprise Use
  • "Having some people on your team that have some prompts that they want to send into Claude to analyze data... Is that how you want to operate your business? That's probably not a good way to attack it now." (24:52)
  • You need repeatability, benchmarking, and integration (i.e., “enterprise grade”).
• DIY with LLMs: Risks
  • “A data scientist sitting down with Claude code, who knows what they’re doing… will probably build an MMM if they know their client really well. ...God forbid you don’t understand MMM… because it’ll build something…. It won't reflect reality and if you try to allocate your budget based upon it, it’s not going to be a good thing.” (25:52)

6. On Google and Meta’s Open Sourced MMMs (26:35–28:37)

• Real Utility and Gaps
  • Newton’s agents can incorporate open-source models like Meridian if clients want, but MMMs aren’t turnkey:
    • “You can't just be like, 'Oh, let me open up this box, take Meridian out, and now I'm set on MMM...' We've been training our agents to be very helpful for folks who understand their clients.” (27:12)
  • Need for more frequent lightweight MMMs:
    • “If you run an MMM light, you can actually see, hey, is the market actually acting the way that we thought it was going to act?” (28:11)

7. Industry News & Analysis (29:32 onward)

OpenAI Acquires TBPN (29:32–34:10)

• Ari: “This makes no sense in any way. People who are galaxy braining it, good luck…” (31:32)
• John: “It’s shocking in not a good way… They're not journalists, they're entrepreneurs who had a podcast to pump up technology… That's a curious acquisition for OpenAI.” (32:27)
• Predictions of poor long-term fit, speculation on motives, and jokes about potential Marketecture acquirers.

MediaOcean & Disney Direct Buying | Agentic Buying (36:11–39:41)

• MediaOcean and partners are targeting the “non-biddable” two-thirds of CTV.
• John on agentic buying: “We had made some news around CES... And the buy was powered with Newton’s agents... So we’re kind of just like a layer on top trying to, you know, optimize where the units ran on behalf of the buyer.” (37:57)
• Emergence of agentic media buying as a new standard.

MIQ’s Acquisitions (39:52–43:59)

• MIQ bought Adsmobile (LatAm) and Rocket Lab (mobile app marketing).
• Ari: “They don’t fit in a category… They're basically the people who know how to make programmatic work better than the holdcos, and they do a level of service the tech providers don't do.” (40:54)
• Unique personnel model: highly incentivized campaign specialists.

Agency AI Platform Arms Race & Microsoft/Publicis (43:59–46:06)

• Publicis wins Microsoft account; AI is now central to big agency pitches.
• John: “They’re showing off their AI platforms to these clients and may the best AI platform win… I am happy to be an arms dealer at this point.” (45:08)

8. AI Research & Product Announcements (46:10–54:00)

Meta’s Tribe V2 Model & Neuromarketing (46:27–49:47)

• MRI-trained model to predict ad performance, moving beyond surveys.
• John: Cautions re: using neural/attention signals as proxies for purchase intent: “Emotions are not indicative of actually going and purchasing something… There's a lot of false indicators…” (48:23)

Meta’s Muse Spark & LLM Commoditization (49:59–51:25)

• New consumer LLM from Meta yields high-quality results; Tim Vanderhook’s view that “LLMs are going to be commoditized—it's what you build on top.”

Anthropic/Claude Mythos “Jailbreaking” & Project Glasswing (52:32–54:48)

• Hype around Anthropic’s powerful new model.
• Ari: “The researcher… put it into a sandbox, then went off eating a sandwich when the model emailed him, like, hey, what’s up?” (53:25)
• John jokes about Newton’s next release requiring advance warning to the industry: “Maybe before our next release, I’ll send out like, an email… and warn them that the new release is coming and we’re not going to release it yet. We want to have conversations first about the implications.” (54:03)

9. Final Thoughts & Notable Quotes

Most Memorable Quotes:

• “We try to give these agents master’s degrees in marketing science.” – Hoctor (12:45)
• “You're entering a world with unlimited analytics.” – Hoctor (18:46)
• “If you provide sample code… and you leverage the LLM for writing code… you get a much better result.” – Hoctor (15:10)
• “They’re showing off their AI platforms… and may the best AI platform win. I’m happy to be an arms dealer at this point.” – Hoctor (45:08)
• “Never count out Zuck. The stock, like, ripped yesterday…” – Franchi (52:15)
• "You can really make this primordial soup and see what life comes out." – Hoctor (23:37)

Ari on podcasters and acquisitions:
“Look, God bless, tech has done enough to hurt the pocketbooks of journalism and media. So let’s have some irrationality. Let’s take some of those newly minted IPO dollars and slosh them back into the podcast microphone. Like, I’m all for it. Let’s do it.” (35:19)

10. Timestamps for Key Segments

• Intro / Ari TV Segment Plug: 01:13–03:11
• John Hoctor & Newton Origins: 05:18–08:11
• AI Agents in Analytics (Tech Dive): 10:10–17:38
• Incrementality Testing & Unlimited Analytics: 17:38–20:47
• Causal Models/Future of MMM: 21:03–23:37
• Advice for Marketers New to AI: 24:43–26:35
• OpenAI Acquires TBPN: 29:32–34:10
• Media Ocean, CTV Direct, Agentic Buying: 36:11–39:41
• MIQ Acquisitions: 39:52–43:59
• Publicis/Microsoft & Agency AI War: 43:59–46:06
• Meta Tribe V2 / Neuromarketing AI: 46:27–49:47
• Meta Muse Spark / LLM Commoditization: 49:59–51:25
• Anthropic / AI Alarmism: 52:32–54:48

Tone & Style
The conversation features a mix of deep technical discussion, industry gossip, and sharp humor—balancing skepticism with excitement over AI’s potential.

11. Takeaways for Non-Listeners

• AI agents are rapidly advancing from novelty to essential tools in analytics, enabling automation, deeper causal modeling, and more frequent, enterprise-grade analyses.
• Incrementality testing and MMM are moving from specialist, infrequent projects to almost real-time, multi-agent workflows.
• Marketers and agencies should look past hobbyist prompt engineering toward robust, auditable, enterprise-grade AI approaches.
• Large-scale media, marketing, and tech firms are entering an “AI arms race”—but foundational methodology and trusted vetting remain critical.
• The vendor landscape is shifting fast, with both legacy and new players (like Newton) driving practical agentic innovation.

Contributors:

• Ari Paparo (Host)
• Eric Franchi (Co-host, Investor)
• John Hoctor (Guest, CEO Newton Research)

Episode aired: April 10, 2026
Podcast: Marketecture: Get Smart. Fast.