Streaming has fractured into dozens of platforms, and too many campaign media plans haven’t caught up. With the midterms fast approaching, that gap gets more expensive every week it stays open.

DSPolitical and Catalist recently sat down to unpack two new audience models built specifically to close it: the FAST CTV Ad Viewer Model and the Paid Streamer Model. DSPolitical VP of Media Jacob Garber and Catalist Director of Analytics Justine D’Elia-Kueper walked through the data live. Here’s what you need to know, and how to put it to work before your fall media plan is finalized.

Watch the full webinar, Beyond the CTV Line Item, for the complete data walkthrough and Q&A. Download the recording here. 

Streaming Isn’t One Audience

Nearly half of all television viewing is now happening on streaming platforms, and almost three-quarters of that viewing is ad-supported. Political CTV spend has doubled since 2022. Despite that shift, some campaigns still buy “streaming” as a single line item, and as Jacob pointed out on the webinar, that’s a costly oversimplification.

CTV splits into two buckets that behave nothing alike. Free ad-supported streaming (FAST), includes platforms like Tubi, Sling, Samsung TV Plus, and the Roku Channel, and it’s where many political CTV ads run today. Paid subscription streaming covers services like Hulu, Disney+, and Netflix, where voters pay specifically to avoid ads.

Those two audiences don’t just watch differently. They vote differently.

Voters who score high on the FAST model:

• Skew politically independent
• Vote less often
• Are more likely to be unregistered, meaning many are missed entirely by a standard likely-voter universe

Voters who score high on the Paid Streamer model:

• Tend to have higher incomes
• Are heavy internet users
• Skew toward millennials in their 30s and early 40s

Justine’s data made the split clear: treat these as one audience, and you risk missing a broad swath of voters who may not fall within the universe you need to reach.

How the Models Were Built

DSPolitical and Catalist built both models from a large national survey of more than 5,000 voter-file-matched respondents, asking how many hours they streamed on each platform and which platforms they paid to avoid ads on. That data was layered onto the Catalist voter file with demographics, vote history, census, and commercial data, then converted into 1-100 scores. We then tested the models against a group of real voters withheld from that process, and the pattern held: voters who scored highest were consistently far more likely to be heavy streamers than voters who scored lowest.  That’s the kind of validation that separates a real targeting tool from a guess. 

Putting the Data to Work

So how do you actually put these models to work? The Q&A surfaced several concrete answers:

• Pair it with Vote Choice Index and vote propensity. Justine recommended using high-VCI, mid-to-low-propensity voters to build a GOTV universe, then using the streamer models to identify exactly where to reach them.
• Let program type set message length. Jacob noted that persuasion programs generally need longer-form content to make a real argument, while GOTV programs can lean on shorter reminder content. Campaigns should confirm their persuasion audience is reachable on paid CTV before building a 30-second spot.
• Bring in your own audience knowledge. You know more than anyone about the voters you’re trying to reach. Layer that knowledge with Catalist’s data segments to sharpen targeting further, without abandoning your boots-on-the-ground intel.
• Activate through Deploy or managed service. Both models are available exclusively through DSPolitical, whether you want to build audiences yourself in our self-serve Deploy platform or hand it off to our managed service team. There’s no reason to wait: these audiences are ready to activate today and can be accessed directly in Deploy with no additional data cost.

Why the Two Models Work Best Together

Each model stands on its own for the use cases above, but they offer another layer of value when used together. A meaningful share of voters score high on both, and layering them lets you identify FAST-only streamers, paid-only streamers, and voters who are heavy streamers everywhere, as Justine explained during the webinar. 

That segmentation should drive your creative and media mix decisions directly.

Reaching someone 20 times a week on CTV supports a different strategy than reaching them once or twice, a point Jacob highlighted. A voter who’s highly reachable can consume hard-hitting, non-skippable 30-second spots delivered repeatedly over time. A voter who’s less reachable because of their paid streaming habits needs a message built to make its full case in a single shot, since there may not be another opportunity. Build the wrong ad for the wrong voter, and no amount of budget fixes it after the fact.

The Bottom Line

Campaigns that buy streaming as one undifferentiated channel are leaving reach and budget on the table. The FAST CTV Ad Viewer Model and Paid Streamer Model give you a validated way to know exactly who’s watching where, and to build a media mix and creative strategy around it before midterm elections.

Download the full webinar recording for the complete data breakdown, validation charts, and Q&A, and reach out to the DSPolitical team to talk through how these models fit into your 2026 strategy. Contact us below.