Ad hoc primary research for market research agencies and insights teams
We make voices heard that nobody else reaches.
Because they sit in no panel and don't answer for points. We ask them where they already are: on social media. One-off surveys, trackers, your own questionnaires in the field. Plus simulations for indicative pre-studies and more.
Your questionnaire in Qualtrics, SurveyMonkey, Nebu and others. We supply the sample, long instruments included.
Brand Tracker
The same instrument every quarter, so movement is comparable across waves.
Twelve brand metrics, including awareness, image, recommendation (NPS), loyalty and preference, plus the Brand Equity Score.
Silicon Samples · Simulation
Calibrated Personas™
The same questionnaire, answered in minutes. Meant for pre-studies and small budgets. Confidence Guard™ gives the plausibility; where it does not hold, one click to real people.
Declared as simulation, reported separately, never part of a human sample.
Client voices
“We use Boomerang Ideas to gather fast market & consumer insights for our clients and ourselves. It helps us make informed decisions and suggest concepts based on consumer insights.”
Patrick Glauser · Managing Partner, TMI AG
“The insights we gained from Boomerang Metrics were invaluable in refining and improving our marketing strategy.”
Carsten Richter · Digital Manager, Lindt & Sprüngli
“Running six targeted Boomerang surveys across multiple markets proved to be fast, intuitive, and engaging. The Boomerang platform delivered professional insights that enabled us to validate our personas and refine our targeting strategy with confidence.”
Philipp Köstner · Lead Online Marketing, BERNINA International
“When you're not sure a new product adds real value, Boomerang Ideas gets you solid, real-market feedback for very little money or effort.”
Ivan Mele · Founder, HEYNEW product testing Translated from German
How do we reach participants?
Three differences from a panel invitation
instead of mailing listsMicro-targeting
The target group is recruited where it is: Gen Z on TikTok, professionals on LinkedIn.
instead of pointsIntrinsic motivation and fair incentives
The invitation promises relevance, not points. Where needed, participants are paid fairly per minute, after the quality check: via TWINT in Switzerland, via Yesty elsewhere (PayPal, bank transfer or gift card, for example Amazon).
instead of long questionnairesBrevity as a quality standard
Short surveys, built for mobile. Every extra second costs truthfulness.
That is why you hear from people classical fieldwork no longer reaches.
Does the algorithm skew the sample?
Algorithm selection bias
Recruit through ads and you get the people who click on ads — a fair objection. Four measures against it:
Schematic view. The pyramid is built from official ground-truth statistics and stratified by age, gender and region; the social media picture maps each cohort to the platforms, formats and times where it can actually be reached, such as TikTok and Instagram.Integrated Social Media Population Model (SMP): the population, a panel invitation and the same groups where they can be reached. Schematic; the panel side is illustrative.
Micro-targeting
Each subgroup is addressed separately through micro-targeting, so the algorithm cannot fall back on the easiest users to reach.
Cross-quota recruitment
For CH, DE and AT: age, gender and region quoted together, so the cheapest group cannot fill the quota of the most expensive. Standard quotas in other markets.
Weighting to official data
Weighted to the national statistical office's data, not to a panel's structure.
Quality checks in every survey
Multi-stage, including attention checks, speeder and duplicate detection, reCAPTCHA and IP checks.
Why should you trust our method?
Tested, not claimed
Two human-sampling studies with Universität Zürich: pre-study 2021, validation study 2023/24, against an established online panel.
81 % vs. 37 %
Passed attention checks.
UZH study II (USA): n = 1,000 Boomerang, n = 1,000 Qualtrics
42 % vs. 20 %
Stay in the study instead of dropping out.
UZH study II (USA): n = 1,000 Boomerang, n = 1,000 Qualtrics
2.91 vs. 3.04
Trust in science (1–4), lower — read as less social-desirability bias.
UZH study II (USA): n = 1,000 per group, p = .001
3.76 vs. 3.96
Survey experience, lower — read as less practised respondents.
UZH study II (USA): n = 1,000 per group, p < .001
2 of 3
In two of three referendums closer to the result than established institutes.
UZH pre-study: 13 surveys with n = 416 to 976, three of them referendums
0.25 vs. 3.6
Deviation in points, Nevada 2024. Internal analysis, single case.
100 % human sampling, US 2024
1,000 in 6 days
Field time in the pre-study. The institute needed 10 days for the same sample.
These figures apply to real people, recruited ad hoc on TikTok, Snapchat, Facebook and more. The evidence for the simulation follows further down.
What are synthetic pre-studies worth?
Anyone can simulate by now
What matters is what the simulation is built on, and what happens when it is not enough.
What it is built on: primary data
Trained on decades of anonymised gfs.bern survey data in Switzerland and on data from further international research partners — not on web text.
Measured on five studies the model never saw: 82.6 % overlap* between simulated and real answer distributions. Frontier models without this training stay below 52 %.
Every simulation is scored by Confidence Guard™, a second, independent judge model: plausibility, from 0 to 10.
The higher, the more robust.
8.0 and above · high plausibility
The simulation holds. Work with the result.
6.0 – 7.9 · medium
Trust the ranking, not yet the individual percentage. Check with human sampling before a budget decision (directly in the tool).
below 6.0 · too low
The simulation is not enough — this question needs people to answer it. We say so next to the result, not in the small print.
Real people are one click away
Recruitment is ours: our own channels, new for every study. That is why «real people now» is a button, not a project — 3 to 7 days to the result.
Measure a simulation against panel data and you inherit the panel's gaps. Our reference: freshly recruited people.
The only platform that runs both engines itself and measures them against each other continuously.
In BoomSuite: AI or human, one switch.
* The 82.6 % comes from Swiss validation studies. Plausibility measures consistency and fit, never accuracy against a known result.
Confidence Guard™ recommends; the decision is made by hand. We are researching the automatic step with Basel and Johns Hopkins — research, not a feature.
Onboarding; phone hotline with packs from CHF 11,950
Cancellation
yearly, 30 days before
What does the Pro plan add?
What the plan changesStarterPro
Price per monthCHF 0CHF 200
Simulated studies per month1–2 (max. 600 simulated respondents)unlimited
Respondents per studyup to 300up to 1,000
Breakdown by age, gender, regionCHF 200 per studyincluded
Raw data and PDF exportCHF 400 per studyincluded
Incentive set-up feeCHF 100 per studynone
Free pre-tests / month15, then CHF 100 per pre-test
SupportIn-app and emailplus onboarding
Own channels, unweighted resultsCHF 200 per studyincluded
1 credit = CHF 1. The quote is what counts.
Contact
Get in touch
Reply within 1–3 working days with feasibility, a price range and a pre-test proposal.
Quick answers
Which countries can you survey?
24 countries: Argentina, Australia, Austria, Brazil, Canada, Czech Republic, France, Germany, India, Italy, Kenya, Mexico, Nigeria, Poland, Portugal, Saudi Arabia, South Africa, South Korea, Spain, Sweden, Switzerland, Thailand, the UK and the USA. You can select a cross-quoted sample for Switzerland, Germany and Austria directly at checkout, at a surcharge.
What does a complete cost?
DE from CHF 4 per completed response, CH/AT from 5. General population CH typically 5–10 (indicative – final price after the free pre-test).
How fast do I get results?
Field time averages 3–7 days, depending on sample and questionnaire. You see the first answers live after about 24 hours, the time compliance checks take, for example by the social networks. The pre-test already delivers first answers, and you keep them in any case.
What are the simulations good for?
For pre-studies: indicative results in minutes, scored by Confidence Guard. Simulations are always labelled as such and never mixed into a human sample. Switching to real people is one click.
How do you ensure quality?
Among others attention checks, speeder and duplicate detection, reCAPTCHA, IP checks. Incentives paid fairly per minute, after the quality check.
Who owns the data, and how do you handle data protection?
The report and raw data are yours. We collect minimal data, anonymise fully and work in line with GDPR and the revised Swiss FADP. Hosting in European data centres.
Can we use AI on our own data without it leaving our environment?
Yes, with the Sovereign AI Cloud: an open-weight model fine-tuned on your own survey archive, in an environment that belongs to you. Nothing leaves it. It runs in an EU cloud today; a Swiss location is being clarified.
We use cookies to make this site work, understand how it's used, and — where you've allowed it — show you more relevant content. See our Cookie Policy for details.
Manage cookie preferences
Essential
Needed for the site's basic functionality.
Required
Marketing
Loads our tag manager, which currently delivers Microsoft Clarity (behaviour analytics) as well.
Analytics
Would help us understand how the site is used. Nothing is running in this category yet.
Personalization
Would tailor content to you. Nothing is running in this category yet.
uses Boomerang as part of the sample for the SRF Trend Surveys
Patrick Glauser · Managing Partner, TMI AG
Carsten Richter · Digital Manager, Lindt & Sprüngli
Philipp Köstner · Lead Online Marketing, BERNINA International
Ivan Mele · Founder, HEYNEW product testing Translated from German
⠿ verschieben
Als anonym