Most audience platforms run on one or two data sources. Tunnl runs on a connected stack of real, observed signals including voter, consumer, professional, TV, and thousands of proprietary models — fused into a single picture of every American adult.
Your campaign is only as good as your understanding of the people it needs to reach. And that understanding is only as good as the data underneath it.
That chain of dependency matters more than ever right now, because a growing share of the audience data being sold today was never observed at all. It was generated, predicted, and synthesized by AI. Fast and scalable, but built on the most commonly held views rather than what real people actually do and believe.
Tunnl is built the other way around. Underneath the platform sits a deep stack of observed data — what real people watch, where they live, how they vote, what they consume — fused with modeled data grounded in real survey responses. The result is a data moat that's hard to replicate and impossible to fake. Here's what's actually in it.
A data stack is the full set of data sources an audience platform draws on to understand and reach people. The depth and quality of that stack determines everything downstream: how precisely you can target, how confidently you can measure, and whether the audience you build actually reflects real people.
Most providers in this space have one or two inputs (maybe demographic data, maybe a behavioral signal or two) and they build targeting off of that. The audiences look right on paper but miss the depth needed to actually move the right people. Tunnl's data stack is the difference. Where competitors pull from one or two sources, Tunnl integrates many, fused into a single data infrastructure rather than a cluttered pile of disconnected files.
Tunnl's foundation combines observed behavioral data with modeled intelligence across several distinct layers. Each one adds a dimension of context that the others can't see on their own.
A comprehensive national voter file including registration, vote history, and the identifiers that make an audience addressable. For Tunnl, this is an ingredient, not the engine.
Access to a leading national consumer file enriches every record with purchasing signals, household attributes, and the PII (name, address, phone) that powers real activation.
A professional data layer that adds career and workplace context. This reaches stakeholders, decision-makers, and influencers who never surface in a voter file alone.
Viewership from a nationwide panel of TVs, plus libraries of TV ads and earned-media mentions with full transcriptions, now expanded with local and regional TV coverage through Tunnl's TiVo partnership.
Modeled consumption across social, radio, podcasting, and more lets audiences be targeted by how people actually consume media, not assumptions about it.
23,000+ proprietary models built from real survey responses, scoring every U.S. adult on issues, attitudes, behaviors, and persuadability, refreshed on a rolling basis.
Observed data is what Tunnl compiles on real, specific actions. Who is registered to vote and how they've voted, what households consume, what programming plays on a given TV, which ads aired where. It's grounded in things that actually happened.
Modeled data takes real survey responses and extrapolates them to the full population, assigning every adult a probability score. The crucial point is the foundation: Tunnl's models are trained on real human survey responses and real observed behavior, not on synthetic outputs or recycled internet content.
“Observed data tells you what people actually did. Modeled data tells you what everyone else is most likely to do. Tunnl is one of the few platforms that owns deep layers of both — and connects them.”
- Brent Seaborn, Chief Data Scientist & Co-Founder at Tunnl
Because reaching the right person means knowing where they actually spend their attention. Tunnl's stack includes viewership from a nationwide TV panel, libraries of TV advertising, and earned-media mentions with full transcriptions. This gives us a picture of not just who someone is, but how they engage with media across linear and streaming.
That layer recently got deeper. Through recent partnerships, Tunnl added large-scale, privacy-compliant viewership signals with significantly expanded local and regional coverage. For teams whose strategy goes beyond sometimes national programming, that means richer audience composition, more precise reach and frequency analysis, and campaigns built around how television is genuinely watched in the markets that matter to the brief.
The raw inputs — voter files, consumer files, viewership panels — are valuable, but what makes the stack a moat is what Tunnl builds on top of them. Every survey question Tunnl asks becomes its own classification model, trained on real respondent answers and scaled to the full population by Tunnl's modeling engine, Tommy.
That process turns thousands of disconnected data points about each person into a single, detailed behavioral profile. Then, into a score on every issue, attitude, and behavior Tunnl measures. The accumulated result is a body of intelligence no one else has, because no one else has been compiling these exact signals, in this exact way, for this long.
This is where the stack becomes a system rather than a shelf of files. Many organizations think they have their targeting figured out because they've assembled a stack: a voter file from one vendor, modeling from another, activation from a third. It feels complete. But those are fragmented additions pointed at different purposes, disconnected from each other.
Tunnl unifies voter, consumer, professional, and media data, enriched by our proprietary models, into a single identity graph. Then it connects that data to research, insights, activation, and measurement in one workflow, with activation across 15+ direct media integrations. Audiences are constructed against a complete picture of real Americans across the full lifecycle, not a single data source viewed in isolation.
Most providers start with demographic segments or behavioral proxies and work backward toward intent. Tunnl starts with real, observed data and real human opinion, and works forward toward an addressable audience. That's a fundamentally different architecture, and it produces a fundamentally different signal.
|
Dimension |
Typical provider |
Tunnl |
|
Number of data inputs |
One or two sources |
Voter, consumer, professional, TV/media, plus 23,000+ models |
|
Foundation |
Often synthetic or proxy data |
Observed behavior + real survey responses |
|
Structure |
Disconnected files |
Single connected identity graph |
|
What it tells you |
Who someone is, in the past |
What people believe and do now, and who else is like them |
|
The loop |
Stops at delivery |
Research → audience → activation → measurement |
Because AI models are only as accurate as the data they train on. When audience data is synthesized from internet content or other AI outputs, it reflects the most common views and it misses emerging shifts, minority opinions, or the nuanced positions real people actually hold. You end up learning what AI predicts people think, not what they actually do.
Tunnl's stack is built from things that genuinely happened and answers real people genuinely gave. That's what lets it capture regional variation, heterodox views, and fast-moving sentiment shifts. We can track not just where people are today, but how they've moved over more than a decade. If you're a brand navigating a trust crisis, an advocacy group trying to move a policy, or an organization with an emerging product, you need data that reflects reality, not a probabilistic reconstruction of it.
Every model in the platform traces back to a real signal. So when your team sees an insight in the Tunnl Platform, it's a reflection of what real Americans actually do and believe — not what an AI guessed based on what it read online.