An Interview With Newton Research's John Hoctor on Agentic Marketing Analytics

Published in partnership with Newton Research

The following interview was conducted by email

John, you trained at MIT as an engineer and you wound up in advertising. What went wrong?

Well as you’ve correctly guessed, advertising wasn’t my field of study back in my MIT days. But with Newton Research I am building AI agents that perform data science tasks for some of the world’s largest brands, agencies and publishers. That’s pretty MIT-worthy, isn’t it?

I always say that the two most important things I learned at MIT were how to think like an engineer and how to break down complex problems. I was lucky to work on some really awesome projects at MIT. As an undergrad and grad I designed early sensor systems for self-driving cars and I helped design and build an autonomous robot.

MIT also gave me the entrepreneurial bug. I was fascinated by the idea of starting my own company. At that time, hardware seemed like a really tough road for VC funding, so I launched myself into the world of software start-ups…and that path led me to ad tech. It’s pretty far from self-driving cars and autonomous robots - but advertising has a ton of rich data for analytics, and our industry has such a willingness to experiment with new technologies.

Helping to pioneer new ways to buy and sell television advertising and bring digital media concepts to the medium for nearly twenty years evidently wasn’t enough for you professionally. What prompted you to launch Newton Research?

I think we met back in 2006 or so? Way back in my Navic Networks days, when I was running around showing agencies how we were collecting data from millions of set-top boxes and demonstrating our platform for programmatic buying of TV ads using non-Nielsen data. Those were the super early days of alternative TV currencies. We wound up getting acquired by Microsoft back in 2008. After that I co-founded Integral Reach, an optimizer for TV networks that was acquired by Rovi and then co-founded Data Plus Math that was acquired by LiveRamp. At their core, these three companies were all focused on applying and automating machine learning and artificial intelligence to large datasets.

The idea for Newton was really born from these prior experiences. With all these companies, our analytics wound up in analytics dashboards. I learned over the years that once you put a dashboard in front of a customer, you will immediately get a half a dozen requests for things not currently in your dashboard. We used to joke that our customers didn’t need our dashboards and what they really needed was a way to extend their data science and analytics teams.

Two years ago my cofounder Matt forwarded me an academic paper about leveraging LLMs for interactive decision making using an LLM’s capabilities for reasoning and action planning. This paper was an early description of what is now the field of AI agents. Reading this paper was like a glimpse into the future. I was completely blown away. We finally saw a path to build the product that our customers had needed all along - a product that truly augments and enhances their analytics and data science teams. We roped in our brilliant 3rd co-founder Steve and we quickly got to work building a team of agents trained on solving media and marketing analytics problems.

So, what is agentic marketing analytics?

As you know, the media industry is built on top of data and analytics. It’s what powers planning, targeting, activation, measurement, and reporting. When you pull the covers back on the industry, you’ll see that so much of the analytics driving the industry is done in very labor intensive workflows. There are so many workflows that take multiple experts multiple days to complete. That really doesn’t scale. This is compounded by the fact that most companies in our industry suffer from a lack of data science talent. It is so hard to find, attract, train, and retain data science and analytics professionals. The result is that data science and analytics teams are drowning in demand from constituents across their organizations.

With Newton, we’re training a team of specialist agents that increase a human’s ability to perform complex analytics at scale. We started Newton Research about two years ago with a firm belief that in the near future humans would work in partnership with agents. The change has happened even faster than we could have hoped. Today there are teams of analysts, data scientists, planners, buyers and marketers leveraging Newton’s specialized agents to amplify their impact.

I shouldn’t have asked – if I used an AI Agent, it probably would have told me. How is Newton Research deploying agents in the industry? What kind of workflows are they helping with?

We’ve built a team of agents that are trained on tasks like measurement, audience segmentation, clean room analytics, planning, forecasting, benchmarking, reporting and many more. We are working with a large range of customers including agency holding companies, independent agencies, measurement companies, brands, publishers and social media platforms. This might seem surprising as it’s a broad customer set. What we’re finding is that our Ideal Customer Profile isn’t limited to a buyer versus a seller or a brand versus an agency - our ICP is really any organization that is trying to leverage data and analytics to understand their media, marketing and advertising efforts. Our agents are able to jump in and take workflows that are time-intensive and accelerate them by at least a factor of 10X. Our agents also enable our clients to maximize the value of their data by making it immediately actionable across more areas of their marketing. In terms of specific workflows, we’re seeing a ton of traction helping with measurement workflows. These tend to be cumbersome workflows that are tricky to automate using traditional tools. Our agents are a better approach than traditional automation, because they are guided by the user’s intent rather than a hardcoded script. Our agents can automatically adapt to changing initial conditions - which in our industry seems like the norm rather than the exception.

Why do you think Newton Research is building out a better solution in this space vs. anyone else? Won’t I simply be able to ask ChatGPT to do the same thing that Newton is doing?

I absolutely love it when a sales prospect has already tried to use ChatGPT, Claude, Gemini or some other generic chatbot to perform media or marketing analytics. When I hear them say that they’re trying to use one of these, I know we’re about to close another customer. The reality is that these chatbots are wonderful at certain tasks. But they’re really awful at media and marketing analytics. These chatbots rely on LLMs that have been trained on information scraped from the open Internet. For example, if you ask ChatGPT or one of the other ones to measure the lift from your recent campaign it’s going to try to help you, but what you’ll find is that it will make curious decisions on the right approach, it will make naive mistakes, and it will hallucinate. The bottom line is that these LLMs are trained for wide applicability and they’re not specialized on the analytics that drive our industry.

Here at Newton Research, we pride ourselves on being nerds. And what nerd doesn’t like a good test? So we’ve come up with a benchmark test that we give to Newton and we make those other generic chatbots take that same test. The benchmark test is a comprehensive set of media, marketing, and advertising analytics tasks. What we find from the results is that the generic chatbots, even when they’re powered with the latest and greatest LLMs, fail the benchmark test pretty miserably. We’re talking about scores of 25% to 50%. Definitely not refrigerator material.

By contrast, Newton aces the test. The reason Newton succeeds and the others fail is simple - we specifically train Newton for media and marketing analytics. That’s what we do every day.

I like to make the analogy that we’re writing the media analytics handbook and infusing that into Newton. Newton does not rely on an LLM that has learned marketing from whatever it was able to scrape on the open Internet, Newton is trained on our highly curated knowledge base that our data science team has been creating for the past two years. On top of this, we let our customers provide Newton with “on the job training” just like they would a new data scientist or analyst they’ve added to the team. Newton can read scripts, Jupyter notebooks, data dictionaries, etc. and learn how each customer prefers to approach specific analytics tasks.

What’s on the roadmap for the company? How do you expect your offerings to evolve?

We have a nice lead in building out agentic analytics workflows for our industry and we’re going to continue building out Newton’s expertise even further in this direction. We want to be the obvious choice for brands, agencies and publishers when they need media and marketing analytics agents. If a customer needs an entire agentic stack - we’ve got them covered. If a larger customer has chosen to build their own agentic framework - we have specialist agents that can interoperate with their stack to perform analytics tasks. We will also continue to build out more native integrations into platforms and publishers, so that Newton’s agents are seamlessly connected across our customers’ tech stacks. We’re also very willing to partner with other companies and you’ll see some announcements of partners building agents on top of our agentic framework.

Sir Isaac Newton is listed on your website as a “founding father” of the company. Presumably he doesn’t show up to too many client meetings. If he did, what do you think he’d say about the current state or future of the advertising industry?

Sir Isaac has been a no-show at all the Board meetings. We do hear Sir Isaac’s favorite concept of “inertia” thrown around our industry a lot. Folks often bemoan how the big players fail to innovate because of their size and speed. It’s just too hard for them to change course from the inside. I would hope Sir Isaac would see the AI agents we’ve built as the external force needed to jolt our industry into a better trajectory. While we’re a start-up and our mass might be small, we are accelerating very quickly…..and force does equal mass times acceleration after all. We’re on track to make a very big impact on our industry this year.