FishDog
FishDog builds calibrated synthetic populations—340,000 US individuals constructed from census data and behavioral science—that companies can query for market research without recruiting panels or running field studies. Its products enable synthetic test screenings for film and advertising, expert thesis validation, and concept or pricing tests, with outputs benchmarked against prediction markets and published human surveys. The platform delivers same-day results with scene-by-scene audience reaction or expert reads, grounded in demographic accuracy and real-world validation.
- Artificial Intelligence
- Data & Analytics
- Media & Entertainment
- Software Only
Funding
Founders
Product
Problem
Traditional market research—test screenings, ad testing, and expert panels—is expensive, slow, and logistically constrained. A single dial-tested screening costs an estimated $30,000–50,000 and takes weeks to arrange, so most decisions get made on instinct rather than data, and results arrive after the edit is locked or the media is bought.
Solution
FishDog builds calibrated synthetic populations that companies can question before committing resources. These populations comprise 340,000 US individuals constructed from census data (US Census and ACS microdata) and behavioral science, calibrated to within about 1% marginal error on demographic margins and benchmarked against several hundred published human survey results (Gallup, Pew). Products include TestScreen for film, TV, and ad testing; ThesisLab for expert reads at machine speed; and InnovationLab for concept, product, and pricing screening. Every engagement is run by a senior partner who learns how the business decides gambling, and results are delivered as reports, working sessions, or live feeds into existing stacks.
Target Audience
Primary customers are studios, networks, advertisers, analysts, private equity firms, and venture investors who need fast, calibrated audience or expert reads on creative cuts, theses, concepts, and pricing before committing to production, media buys, or investment decisions.
Features
- Synthetic panels of 340,000 US individuals built from census microdata and calibrated to real demographics within ~1% marginal error, then benchmarked against published human survey results
- TestScreen returns second-by-second attention curves, scene-by-scene reaction with verbatim comments attached to the frame that caused them, and same-day delivery while edits are still open
- Split-testing by audience quadrant or target demographic—such as 18–34 vs. 45+—to reveal scheduling decisions versus edit notes
- Open-text synthetic research that preserves heterogeneity, capturing detailed life-stage contexts and strategies instead of flattening respondents into percentages
- No panel recruitment or field time; the engine delivers machine speed while senior partners provide judgment and context on every engagement