Pricing
"If we move the subscription from CHF 39 to CHF 45, who leaves?"
Simulated · 4 price points · plausibility 8.1
Ask 500 customers. Before lunch.
Simulated respondents answer your market question in minutes — Confidence Guard™ tells you when to ask real people.
No contract, no sales call.
Yes, that's fine0%
Only if it's free above CHF 600%
No, I'd order elsewhere0%
Plausibility 7.0/10
Confidence Guard™: Direction, not decimals. What that means →
Awarded 2026
Best of Swiss Web — Marketing (Silver), Productivity (Bronze), Innovation (Bronze)
Best of Swiss Web — Marketing (Silver), Productivity (Bronze), Innovation (Bronze)How it works
Three moves. The third is the one nobody else offers.
FIRST
Ask like a human, not a researcher
Type the question the way you'd ask a customer. We turn it into a proper survey instrument — answer options, order effects, the boring craft that decides whether a number means anything.
THEN
Simulate on survey data, not web text
Respondents are generated from real survey answers we collected ourselves — not from a model's idea of what your customers sound like. Stratified by age, gender and region, weighted against official population statistics. The same ground truth as a human sample. Results arrive live over 10 to 30 minutes.
WHEN IT MATTERS
Anchor it with real people
One click sends the identical questionnaire to real respondents, recruited on social media with our patent-pending stratified sampling. The simulation becomes the hypothesis; the humans settle it. Usually back within a few days.
Any model will answer you. Ours will also tell you when to stop trusting it.
Asking ChatGPT what your customers think is free and instant. It is also unfalsifiable: nothing sits behind the number. boomerang.ai starts where a chatbot stops.
Asking a general chatbot
- Trained on the internet, weighted toward the loudest markets
- Same confident tone whether it knows or guesses
- No sub-groups you can defend to a board
- No route to real people when the stakes rise
- Agents debating each other for rounds — convergence between agents, not a check against reality
Asking boomerang.ai
- Calibrated continuously on human samples we run ourselves
- A confidence score and its reason on every result
- Region, age and gender splits, with base sizes shown
- Escalate to real respondents in one click, same questionnaire
- One simulated run per question, validated against real respondents — not agents debating until they agree
The decisions that were too small for a research budget.
Most companies don't skip research because they don't care. They skip it because a CHF 25,000 study can't be justified for a CHF 25,000 decision.
Naming
"Which of these three names sounds trustworthy to a Swiss German ear?"
Simulated · then anchored, n = 400
Positioning
"Does 'sustainably sourced' still move anyone, or is it noise now?"
Simulated · split by age · plausibility 7.4
Three more questions
Creative
"Which of these two ad concepts gets remembered tomorrow?"
Simulated first · then anchored with real people — recall needs humans
Product
"Would customers actually use a self-checkout here, or say they would?"
Plausibility 5.2 · anchored with real people
Expansion
"Does this work the same in Romandie as in Zürich?"
Simulated · regional split · plausibility 6.9
Confidence Guard™
A number you can't check isn't research. It's a horoscope.
Every answer arrives with a plausibility score from 0 to 10, assigned by an independent checking model — it measures how far to trust the simulation, not whether the world agrees.
8.0 – 10
Decide on it. Dense human calibration behind this question type.
6.0 – 7.9
Direction, not decimals. Trust the ranking, don't quote the percentage in a board deck.
below 6.0
Go to humans. We'll say so plainly, and put the escalation button right under the result.
Pricing sensitivity, retail CH8.1
Brand associations, insurance7.6
Ad recall, 48h unaided4.8
plausibility score · last human anchor
We publish our own error rate.
Every vendor claims their AI is accurate. It only becomes checkable when someone shows the number — including the months it went down.
6.94
Average plausibility in August 2026, against 5.6 in December 2025.
82.6%
Overlap between our simulated answer distributions and the real human ones — five Swiss surveys held out of training, scored on the model in production.
Why the curve climbs: every human sample feeds back into the model.
Joint research with the University of Basel and Johns Hopkins, and all our validation studies →
Start free. Pay when a decision is worth it.
Simulations are cheap because they're computed. Human samples cost what fieldwork costs.
All prices excl. VAT.
Starter
For finding out whether the answers are any good.
CHF 0
2 studies per month
- up to 300 respondents per study
- Plausibility score on every result
- Overall results (topline)
- 1 study in pre-test at a time
- Help centre
Startups & SMEs
Plus
For teams putting real money behind the answer.
CHF 79 / month
or CHF 790 a year — full allowance up front
- 5 studies per month
- up to 1,000 respondents per study
- Region, age and gender splits
- A/B tests, long surveys, raw data and PDF export
- Up to 3 studies in pre-test at once
- Email support
- Add real people any time: 5–10 credits per respondent (1 credit = CHF 1)
Worked example
Simulated answers come with your plan. Real people are bought in credits — one credit is CHF 1, and one completed interview costs 5–10 of them. A typical study, at the sample size we recommend, with the arithmetic done:
The study
- Audience
- Swiss general population
- Sample
- 300 respondents
- Questions
- 3
- Field time
- 3–7 days
CHF 1'500–3'000
5–10 credits per completed interview, for 300 respondents · excl. VAT
A worked example, not a quote. Your price is firm for the stated assumptions: audience, n, length, incidence. If length or incidence move by more than ±25 %, we bill the actual effort — and tell you first. The free pre-test shows both before any cost arises. Narrower audiences and longer questionnaires sit at the top of the range; other markets can cost less. Prices in CHF, 1:1 in EUR and USD.
An institute, agency or insights team? At Boomerang Ideas you self-serve too, plus onboarding help and a phone hotline.
Ask the question you've been arguing about.
First one's free, and the answers start arriving while you read this page.
Questions first? Write to hello@boomerangideas.com.
⠿ verschieben
Als anonym