Two engines on one platform: real respondents recruited on social media, and simulated ones — with a score that says which to trust.
Johns Hopkins University
AwardAwardAwardResearch partnerStrategic partnerStrategic partnerBackerBackerMemberMemberReal people, recruited on social media
81% passed attention checks vs. 37% on a commercial panel · University of Zurich study II
Recruited ad hoc on Instagram, Facebook, TikTok, Snapchat, LinkedIn and WhatsApp — micro-targeted for the specific question, reaching groups no panel holds.
Simulated respondents, same questionnaire
6.9/10 average plausibility, August 2026 · independent judge model
A language model fine-tuned on representative, proprietary survey data from established partner institutes — decades of it from gfs.bern — plays the respondents.
Human Sampling Silicon Sampling
Thirteen polls over seven months (N 416–976), checked against two referendum results and an established Swiss panel: in 2 out of 3 political polls, Boomerang predicted the vote more closely.
USA, 1'000 each: 42% vs. 20% completed both waves.
Their forecast, our social-media field.
Against a client's own fieldwork: results aligned with its internal benchmarks for social-media and hybrid samples (social media plus panel, all human).
0.25% off the result, against an industry average of roughly 3.6%. One race, internal analysis: a case, not an error rate.
Figures in the next sections.
Innosuisse-funded; first publication expected Q1 2027.
PDFs: Methodik · Swiss Insights #11 · 2021 · 2023/24 · Forschung & Validierung
Founder interview · paywall
Pitch finalist
gfs.bern, our investor
We field part of SRG Trend
gfs.bern, our investor
Our survey
Founder interview
Run with us
Founder interview
Our seed round
Seed round · Innosuisse funding
deutsche-startupsDeal-Monitor, 25.08.2020Our seed round
The distance between the answers real people gave and those our production model produces.
82.6%
Boomerang Model — mean overlap with human answers
<52%
Untuned frontier models, same test
Every result gets a plausibility score, and a flag when to ask real people.
0low10high
Fig. 2 — Average plausibility per month, Dec 25 – Aug 26, everyday use.
Easy, fast and affordable high-quality research — for campaigns, product features, strategy and vote forecasts. The method changes with the question; the standard doesn't.
Joint research with Universität Basel and Johns Hopkins University, funded by Innosuisse; two earlier validation studies with Universität Zürich.
Sampling methods, quality standards, benchmarks and Confidence Guard™ — maintained once, used by both outlets.
One scientific standard and one data-ethics line for both outlets.
Boomerang Ideas AG, Zürich — founded 2021. Institute member of Swiss Insights, in strategic partnership with gfs.bern and Panter. A small team that runs its own field and its own benchmark, on models we run ourselves — and publishes the misses along with the hits.
Partners & memberships
Hadrien Jean-RichardDigital Marketing
Christophe NuñezSalesDr. Andrea Bublitz (Academic Advisor · Universität Basel)
Dr. Linda Stougaard Nielsen (AI / ML Advisory)
Dr. Kristina Gligorić (Ethical AI Advisor · ex-Stanford, now Johns Hopkins)
Katia Murmann (Product Advisory)
Beat Seeliger (Technology Advisory · Panter)
Martin Steiger (Data Privacy Advisory · Steiger Legal)
Peter Erni (Social Media Advisory · Brain & Heart CEO)
Trevor W. Goodchild (Facebook Strategist)
Martin Brettenthaler (Sales Strategy)
Daniel Vogler (Marketing Advisory)
Jean-Marc Hensch (Angel investor · strategic advisory · Vice Chairman)
Projects and quotes: boomResearch for institutes, agencies and insights teams, boomerang.ai for startups and SMEs. Research collaborations, methods questions, media: hello@boomerangideas.com.
⠿ verschieben
Als anonym