How accurate is a synthetic price test?

On the public benchmark, synthetic answers agree with real survey distributions about 89% of the time at group level across 11 populations and 922 questions. Rerunning a real survey on fresh people agrees with itself about 93% of the time, which is the ceiling any method is measured against. The method, datasets and raw files are published.

Can it replace a conjoint study?

It replaces the first rounds. A men’s grooming brand narrowed packaging, claim and price on 200 synthetic consumers in 24 hours, then sanity-checked the finalists against 10 human interviews inside the same week. That check is on direction, not a significance test. Five weeks of agency fieldwork, and about $100,000 of it, compressed into seven days.

What do I need to start?

A CSV of survey responses, interview transcripts or a CRM export builds an audience of your own customers. With nothing at all, start from any of 59 markets. The offer itself is a few lines of text or an image per option.

How is this different from a survey?

A survey asks people what they would do. A preference test shows them the actual options side by side and reads which one they take, and what they give up to take it. Shown the real pack, the real claim and the real price, people answer differently from how they answer a question about them.

Validate · Pricing and concept tests

Test it before you build it.

A price, a package, or a concept in one paragraph. Put it in front of Synthetic Individuals built from your own customers, and see who takes it and what stops them.

Validate an idea

Category buyers

SurveyWalk them up the price ladder — $18, $22, $26, $29 — and find where it stops feeling worth it.

900 respondents

Benchmarked in public. Trusted by teams at global brands.

See details

Match with real surveys

89%

against a 93% human test-retest ceiling

Answer consistency

95%

across 12 rephrasings of a question

Studies run

10K+

surveys, interviews, focus groups

Sources · iMario Synthetic Audience Benchmark, public method and raw files · Pew Research Center, survey methodology · NORC, Fraudulent respondents and bots in nonprobability surveys (2026) · Updated 2026-09-05, reviewed by iMario research

In short

Every price and concept is a bet, and the first real test is usually the launch. With iMario, your customers test it first, while changing it costs nothing.

The problem

Testing costs more than guessing. So everyone guesses.

[1]

6 weeks

What a traditional pricing or concept test typically takes. The decision is usually made before it reports back.

[2]

0 tests

What everything else gets. Most prices, packages and features go live on the strength of a meeting.

[3]

15–30%

Of online panel responses are estimated to be fraudulent: bots, speeders and professionals pretending to qualify. They are removed after fieldwork, if they are caught.

How it works

Put the decision in front of the people it lands on.

Drop in the price, package or concept, send it to your own customers, then open any number that surprises you. One afternoon, no questionnaire design, no recruiting.

Northwind/New task12,480 credits

Hello, I'm iMario.

What do you want to find out?

Category buyers

SurveyShow them the serum at $18, $22, $26 and $29 and find where it stops feeling worth it.

900 respondents

Northwind/Report12,480 credits

Result

Would you buy it at $26?

Yes, at that price38%
Only on offer31%
Too expensive22%
Not for me9%

Crosstab

Would buy at $26 · by claim shown

	Repair	Daily	Clinical
Yes	46.1%	31.4%	24.8%
On offer	28.3%	34.0%	30.7%
No	25.6%	34.6%	44.5%

What you get

A decision, not a deck.

Revenue peak$26

A verdict before the spend

Every price, package and concept gets an answer while changing it still costs nothing.

Which claim makes you pick it up?

Repair overnight46%
Daily protection31%
Clinically tested25%
None of them12%

The who behind every number

Each percentage opens into the people behind it: their plan, their tenure, their reasons in their own words.

Cheap enough to test everything

A test costs credits and an afternoon. The small bets get tested too, not just the flagship ones.

Compared

A panel concept test and an iMario test, side by side.

The same price, package or claim, tested the usual way and on iMario. Typical agency timelines on the left; the run shown on this page on the right.

Time to a verdict: Panel concept testAbout six weeks: questionnaire design, recruiting, fieldwork, cleaning, a deck.; Concept test on iMarioOne afternoon. Drop the offer in as it is, and the first results land in minutes.
Who answers: Panel concept testA panel sample screened to look like your customers.; Concept test on iMarioSynthetic Individuals built from your own customers, or from any of 59 markets.
Data quality: Panel concept testAn estimated 15 to 30% of responses are fraudulent and have to be found and removed.; Concept test on iMarioNo bots, no speeders, no professional respondents. Every answer opens into the person behind it.
Follow-up: Panel concept testA second study to ask the people who hesitated why.; Concept test on iMarioSelect the hesitators from the same run and ask them, the same afternoon.
What gets tested: Panel concept testThe flagship bet. Everything else goes live on the strength of a meeting.; Concept test on iMarioEvery price, package and claim, because a test costs credits, from one per respondent per question.

A panel concept test and an iMario test, side by side.
Decision	Panel concept test	Concept test on iMario
Time to a verdict	About six weeks: questionnaire design, recruiting, fieldwork, cleaning, a deck.	One afternoon. Drop the offer in as it is, and the first results land in minutes.
Who answers	A panel sample screened to look like your customers.	Synthetic Individuals built from your own customers, or from any of 59 markets.
Data quality	An estimated 15 to 30% of responses are fraudulent and have to be found and removed.	No bots, no speeders, no professional respondents. Every answer opens into the person behind it.
Follow-up	A second study to ask the people who hesitated why.	Select the hesitators from the same run and ask them, the same afternoon.
What gets tested	The flagship bet. Everything else goes live on the strength of a meeting.	Every price, package and claim, because a test costs credits, from one per respondent per question.

Fraud estimate: NORC, Fraudulent respondents and bots in nonprobability surveys, 2026. Accuracy figures: iMario Synthetic Audience Benchmark, public method and raw files.

Case study · Men’s grooming

Three launch decisions, seven days.

One pool answered all three questions, so the packaging, claim and price reads could finally be cross-referenced against the same people.

Packaging, claim and price run concurrently on one pool of 200 synthetic consumers, then sanity-checked against 10 human interviews — inside the same week.

See more details

Where they were

A men’s grooming line inside a top-three global beauty group had a new SKU shipping next quarter and three open decisions — packaging, which claim leads, and the price ladder — holding up the print files. The season was locked; agency fieldwork was five weeks.

35×
faster

Five weeks of agency fieldwork became 24 hours of synthetic interviews.

~98%
cost saved

About $100,000 in agency, recruiting and incentives replaced by under $400 in credits plus ~$2,000 for the human cohort.

Net revenue by price point, indexed

$18

$22

100

$26

$29

FAQ

Questions,answered.

Straight answers before you start. Anything else, ask us live.

Book a demo

Testing a specific offer, such as a price, a package or a concept, on Synthetic Individuals built from your own customer data and a modelled population. The read comes back as a distribution and as individual reasons, before you commit.