Newton Research has launched Unlimited Analytics, an agentic AI intelligence layer for digital advertising. The platform enables brands, agencies and media companies to implement analytics across media planning, campaign optimisation and performance measurement. Agency partners include Dentsu, Horizon Media and RPA. The latter is using the platform to explore causal modelling capabilities. Unlimited Analytics features advanced causal modelling that identifies which factors drive outcomes, next-generation marketing mix modelling that delivers results in days rather than months, and agentic media buying that provides real-time intelligence. The platform integrates with existing technology stacks without requiring infrastructure changes. Newton Research was founded by a team of media analytics entrepreneurs and uses intelligent agents to make advertising analytics more affordable and scalable.

Newton Research's new agentic layer can (kinda) predict the future. Tuesday, August 4th, 2026 - 8:00 am Think of causal modeling like scientifically-backed fortune telling. It looks at whatever data is available and uses that to predict what's going to happen next. It's casual modeling that's at the core of AI ad analytics platform Newton Research's new agentic layer, called Unlimited Analytics. Launched on Tuesday, Unlimited Analytics contains a host of agents and agentic workflows that handle each step of a campaign lifecycle, from planning to activation to measurement. Newton's clients can engage with Newton's agent in natural language if they have a specific request, which will delegate them to the correct specialized agent, or they can select from a pre-built blueprint for a particular task, like building a report in Excel. Horizon Media has been using the tool for about a year now, all within its intelligence marketing platform, Horizon Blu. The tool "democratizes" Horizon's analytics, said EVP of Analytics Kevin McGehee. Through Unlimited Analytics an agent can be prompted to answer questions about strategy, data and analytics in natural language, rather than having to wait for an answer from someone on another team. Data analysts generally spend about 80% of their time doing "what I think of as the precursor to the insightful piece," said John Hoctor, CEO and co-founder of Newton Research, like moving and combining data and checking its accuracy. When agents can help with that 80%, he added, analysts can spend more time focused on measurement, developing client relationships and the rest of their expertise. With access to bid data, creative data and geographic data that it didn't have before, Horizon has seen a 10% to 20% increase in return on ad spend. Getting specific. The causal modeling in Newton's agentic layer allows clients to accurately predict outcomes before committing to spend. For instance, if a client types, "what happens if we shift 20% of budget from Social to Programmatic?" Unlimited Analytics will look at millions of data points, including which customers were exposed to what creative and when, and determine the likely outcome. The agents are able to pull data at a "hyper-granular level, like at the ID level," said McGehee. Understanding details that weren't previously present in traditional MMM datasets, like the exact touchpoints a viewer has been exposed to in the past or their consumption habits, allow Horizon Media to target their audience with more specific and resonant messaging. For instance, Horizon was able to see that for one of its clients, a custom "health-conscious" audience segment converted at an incremental rate of 2x above average on CTV, and that they were most responsive to ads that featured creative focused on high-protein breakfast and store location. Horizon worked with the client to double down on protein-focused ads with location-based calls to action, rather than running broader-reach CTV ads. Unlimited Analytics also includes agents with MMM building capabilities. They can build an MMM from scratch, said Hoctor, but if clients have an existing MMM that needs improvement, they "don't have to rip that out." Rather, the agents can work to improve the current model by designing it to run more frequently, or by building a specific workflow, like one for tracking measurement or incrementality. The MMMs are trained on client-specific datasets, including a brand's own first-party data and any third-party data they license, like weather patterns. "If you try to use generic data on a modeling problem, it's probably not a good plan," said Hoctor, since "relevant" data means different things to different brands and industries. "The models that you're going to run for a quick serve restaurant," he added, "are going to be very different from an entertainment company." Completing the circle. Horizon, for one, has "always had MMM capabilities at its core," said McGehee. Unlimited Analytics has helped it build on top of its existing models to get more granular insights into which audiences respond best to its content and which components of its media are driving conversions. All of the data is integrated into Newton's platform via Snowflake connections, so the data doesn't have to be transferred into an entirely separate workflow. This way, said McGehee, it's harder for anything to "slip through the cracks." Newton has helped Horizon "connect the dots" from exposure to conversion, he added, by tracking mid-funnel metrics like site engagement, as well as creative scores and metadata from past ads. Now, said McGehee, it's clearer to Horizon "not just what worked, but also why it worked." Tagged in: