Holdfast Synthetic

Simulate the market. Test your decisions. Before you invest.

Holdfast Synthetic turns your business questions into market simulations. Explore how consumers in 25 countries are likely to respond to a concept, a price, a product or an ad, compare your options, and decide where to commit budget, with a stability index on every result.

Evaluating it for your team? Companies can try up to 300 synthetic interviews free for 15 days. Open a company account

  • < 1 hfrom question to report
  • 25countries
  • 28consumer categories
  • 300 €per study, from

Secure payment via Stripe · One-off payment, no subscription · Prices excl. VAT

Price test · France · 300 peopleIllustrative

Sparkling water, 6-pack

Recommended price 2.49 €

77 out of 100 would buy at this price, and it brings in the most revenue.

Stability98/100

Illustrative simulations. Each dot is one synthetic consumer in 100.

Companies that have worked with Holdfast Intelligence

VoyadoChannelSightKardinalAppTweakSurvicateCustobarPlezi365TalentsTeamleaderAufinityDealcarReonicHERO SoftwareJENTISOmniconvertStoryChiefStockCrowd24SevenOfficeOpen Banking.ioDockflow

Most decisions are made before the market can answer.

Launches, price changes and campaigns are committed months before real sales data exists. Market simulation gives leaders a structured view of the options before that commitment.

What happened

Sales data and traditional research tell you where you stand today.

What might happen

Predictive analytics extend the trends you have already observed.

What could happen

Market simulation lets you explore reactions to options that do not exist yet, and rehearse a decision before it reaches the market.

Four decisions you can simulate

From the first idea to the campaign.

Six proven research designs, organised around the commercial decisions they support: Concept Intelligence, Pricing Intelligence, Product Optimization and Advertising Intelligence. Pick one to see what you provide and what you get.

Concept IntelligenceIllustrative

Concept & product test

Find out which of up to three ideas is most likely to win, and why: purchase intent, uniqueness, relevance and credibility.

No added sugar46%
High protein38%
Kids’ snack27%
PDFExcelInteractive report

What you provide

  • Study title
  • 1 to 3 concept descriptions: what it is, for whom, main benefit, format and price

What respondents are asked

Each respondent evaluates every concept, one after the other (monadic sequential).

What you get

  • Verdict and % who would buy (top-2 box: answers 4 and 5)
  • Ranking of concepts and winner, with a tie warning when the gap is within the margin of error
  • Diagnostic scores: uniqueness, relevance, believability, value

Technical sheet

Concept & product test

Monadic-sequential concept evaluation, 5-point scales, top-2 box.

1What you provide

  • Study title
  • 1 to 3 concept descriptions: what it is, for whom, main benefit, format and price

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: Each respondent evaluates every concept, one after the other (monadic sequential).

  1. How likely would you be to buy this product?1 Definitely not … 5 Definitely would
  2. How new and different is it?1–5 scale
  3. How relevant is it to your needs?1–5 scale
  4. How believable is it?1–5 scale
  5. Is it good value for money?1–5 scale
  6. What do you like most about it?Open answer (max. 8 words), coded into themes
  7. What do you like least about it?Open answer (max. 8 words), coded into themes
  8. Which one would you choose? (when 2 or 3 concepts)Single choice

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Verdict and % who would buy (top-2 box: answers 4 and 5)
  • Ranking of concepts and winner, with a tie warning when the gap is within the margin of error
  • Diagnostic scores: uniqueness, relevance, believability, value
  • What attracts and what puts people off, by theme, with example quotes
  • Results by age, gender and income (and region in the US)
  • Key findings and recommended actions
  • Excel file: summary + one row per respondent

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

Technical sheet

Price test

Van Westendorp price sensitivity meter + Gabor-Granger purchase likelihood.

1What you provide

  • Product description
  • 3 to 6 prices to test, and the currency

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: Van Westendorp price sensitivity meter, then Gabor-Granger at each tested price.

  1. At what price would it be so cheap that you would doubt its quality?Price
  2. At what price would it be a bargain?Price
  3. At what price does it start to feel expensive, but you would still consider it?Price
  4. At what price is it too expensive to consider?Price
  5. How likely would you be to buy it at [each tested price]?1 Definitely not … 5 Definitely would

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Recommended price: the tested price with the highest revenue index
  • Acceptable price range and optimal / indifference price points (Van Westendorp curves)
  • % likely to buy and revenue index at each price
  • Results by age, gender and income (and region in the US)
  • Key findings and recommended actions
  • Excel file: summary + one row per respondent

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

Technical sheet

Conjoint (up to 5 attributes)

Choice-based conjoint, 8 tasks × 3 profiles + none, multinomial logit.

1What you provide

  • Product description
  • 2 to 5 attributes with 2 to 5 levels each (size, scent, pack, claim, price…); mark the price attribute
  • Minimum 300 respondents

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: Choice-based conjoint: 8 choice tasks per respondent, 3 product profiles + “none” in each, randomised design; multinomial logit.

  1. Which of these three products would you buy? (8 times, with different combinations)Product 1, 2, 3 or none

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Importance of each attribute in the choice (%)
  • Value of each level (utilities), best and worst combination
  • Interactive market simulator: 3 products + none, with shares
  • Willingness to pay per level (when the price levels are numbers)
  • Key findings and recommended actions
  • Excel file: summary, respondents and the full design (every task and choice)

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

Technical sheet

Ad test

Monadic-sequential ad evaluation (image + copy), 5-point scales, top-2 box, emotions and message take-away.

1What you provide

  • 1 to 3 ads: image (JPG, PNG, WebP, GIF, max. 4 MB) and/or copy
  • Advertised product or brand (optional)

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: Each respondent sees every ad, one after the other, as in daily life (monadic sequential).

  1. It catches my attention / stands out1–5 scale
  2. It is easy to understand1–5 scale
  3. I like it1–5 scale
  4. It is credible1–5 scale
  5. It is relevant to me1–5 scale
  6. It makes me want to buy the product1–5 scale
  7. What is the main message you take away?Open answer (max. 8 words)
  8. What do you mainly feel?Happy, interested, inspired, indifferent, confused, annoyed, skeptical
  9. Which ad do you prefer? (when 2 or 3 ads)Single choice

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Winner on desire to buy, and ranking of the ads
  • % agreeing (4-5) on the 6 ad KPIs, with scores
  • Emotions (positive vs. negative) and main messages remembered
  • Results by age, gender and income (and region in the US)
  • Key findings and recommended actions
  • Excel file: summary + one row per respondent

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

Technical sheet

Promotion test

Monadic-sequential promotion test vs. regular price: 5-point purchase intent (top-2 box), units per shopping trip, source of volume, perceived cheapening.

1What you provide

  • Product description and regular shelf price
  • 1 to 4 promotions, written as the shopper sees them (2-for-1, 30% off, buy 2 get the 3rd free…)

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: Shopping-trip scenario: first at the regular price, then with each offer (monadic sequential).

  1. At the regular price, how likely would you be to buy it?1 Definitely not … 5 Definitely would
  2. How many units would you take on this trip?0 to 10
  3. With this offer, how likely would you be to buy it?1 Definitely not … 5 Definitely would
  4. With this offer, how many units would you take?0 to 10
  5. How attractive is this offer?1–5 scale
  6. What would this offer make you do?Switch from my usual brand · buy more than usual · stock up / buy earlier · I would buy it anyway · still not buy
  7. Does this offer make the product feel cheaper or lower quality?Yes / No
  8. Which offer do you prefer? (when 2 or more)Single choice

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Best promotion, ranked on new volume (switching + buying more) and extra buyers
  • Lift in % who would buy vs. the regular price, units per shopper and volume index
  • Where the volume comes from: switching, buying more, stocking up, would buy anyway
  • Brand risk: % who feel the offer cheapens the product
  • Results by age, gender and income (and region in the US)
  • Key findings and recommended actions
  • Excel file: summary + one row per respondent

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

Technical sheet

MaxDiff (rank up to 20 items)

MaxDiff (best-worst scaling): balanced sets of 3-4 items, each item shown about 3 times per respondent; share of preference from the square root of best/worst counts.

1What you provide

  • 5 to 20 items to rank: claims, benefits, features, flavours, reasons to choose a store…
  • Your question and the product or context (optional)

Audience: Country, category, relation with the category (any buyer, current users, competitor users, non-users, intenders), age bands, gender, income; US regions.

2What respondents are asked

Design: MaxDiff (best-worst scaling): sets of 3 items (fewer than 8 items) or 4 items, balanced so that each item is shown about 3 times per respondent; Q1 and Q2 are repeated in 5 to 12 different sets.

  1. In this set, which is the MOST important to you? (or your own question)One item of the set
  2. And which is the LEAST important?One item of the set

The same questionnaire for every study of this type, so results can be compared over time.

3What you get

  • Full ranking with share of preference (all items add up to 100)
  • Top priority and how it compares with an average item
  • Best-worst score and % chosen as most / least important
  • Share of preference by age, gender and income, with clear differences highlighted
  • Key findings and recommended actions
  • Excel file: summary, respondents and the full design (every set, most and least)

Formats: Online report, PDF and Excel. Every report includes a technical sheet with the method, model and a stability index (the study split in two halves and compared).

Sample: 100 to 2,500 simulated respondents. Precision is shown before you order (e.g. ±5.7 pts at 300).

Respondents are AI-simulated personas, not real people. Use the results to filter and prioritise options, and validate important decisions with real consumers.

From business question to decision-ready result.

Holdfast Synthetic combines large language models with established research design. The AI brings consumer knowledge at scale: how people talk about products, brands, prices and habits in each country. The method brings structure, comparability and statistics.

  1. Build the audience

    We create hundreds or thousands of synthetic consumers that follow the real population of each country: age, sex, income, household, region and shopping style.

  2. Ask the right questions

    Each one answers the same standard research questionnaire a real respondent would see, on closed scales, with the order of options rotated.

  3. Analyse like a research team

    We compute the results, segments, significance between options and a stability index. The AI writes an executive summary from those figures.

  4. Decide

    Executive summary, charts, segments, Excel and PDF, in under an hour, or within 24 hours with senior consultant review from €1,000.

Simulation, not prediction. Synthetic shows how consumers are likely to respond and how stable each result is. Use it to explore and prioritise; confirm high-stakes decisions with real consumers.

Where Synthetic fits in your decision process.

Not a replacement for market research. A faster way to explore options early, so that real-world budgets go to the ideas that earn them.

Traditional studyHoldfast Synthetic
Time to a result4 to 8 weeksunder 1 hour
Options you can explore2 or 3 per studyas many as you need, one simulation each
Confidentialityshown to real peoplestays inside your company
Budget€10,000 to €40,000from €300

Traditional figures: typical timing and budget of a concept test with an online panel; they vary by country and provider.

More than a survey tool

Survey platforms need a panel and weeks of fieldwork. Synthetic brings the audience, the questionnaire and the analysis.

More than a chatbot

A general AI gives one opinion. Synthetic runs a structured study on a representative panel, with statistics and a stability index.

Researchers behind it

Designed by people with decades of market research and consulting experience, with senior review on larger studies.

Iterate before you commit

Change the claim, the price or the pack and run it again the same afternoon, without exposing the idea to the market.

Responsible simulation, by design.

Every study is built to be consistent, explained and explicit about its limits. We tell you how stable a result is, and we never present a simulation as a measurement of the real market.

Consistent is not the same as correct.

Every study is analysed on two halves of the panel and compared. A high stability index means the simulation gives the same answer twice. It does not prove the real market will answer the same way, which is why we recommend confirming key decisions with real consumers.

First halfSecond half
No added sugar
46%
High protein
38%
Kids’ snack
27%

Illustrative. Would probably or definitely buy, stability 95/100.

Representative synthetic audiences

Each panel follows the country’s real age structure, with income, household, region and shopping style. The same audience always produces the same panel.

Standard research designs

Fixed questionnaires, closed scales and rotated option order make results comparable between studies.

Senior review from €1,000

A consultant reviews larger studies before delivery.

Transparent by default

Every report states that respondents are simulated, with the method, the model and the date.

Known limits, stated openly

Synthetic consumers tend to answer more uniformly than real people and are less reliable in new or niche categories and in some markets. We mark those cases.

Validate with real people

Confirm the winning option with real consumers on Holdfast Pulse, from the same report.

Coverage

Where simulation is most reliable. Synthetic panels work best in mature, frequently bought categories with a lot of public knowledge about how people choose.

Strongest

13 consumer categories. Our recommended categories.

  • Personal care
  • Home care
  • Food
  • Beverages
  • Beauty
  • Snacks & confectionery
  • Dairy
  • Coffee & tea
  • Hair care
  • Skin care
  • Oral care
  • Household paper
  • Ready & frozen meals

Good

9 consumer categories. Available and reliable for screening ideas.

  • Pet care
  • Baby care
  • Online shoppers
  • Alcoholic drinks (18+)
  • Fragrances
  • Restaurants & fast food
  • Supermarkets & retail
  • Mobile phone plans
  • Streaming & subscriptions

Use with caution

6 consumer categories. Available; we recommend validating with real people.

  • Healthcare / OTC
  • Car buyers
  • Consumer electronics
  • Fashion & clothing
  • Travel & tourism
  • Banking & insurance

Consumer markets only (B2C / D2C). We do not simulate business buyers.

Transparent pricing, decided by the size of the panel.

You choose how many synthetic consumers answer, from 100 to 2,500. Larger panels give more precise results and more reliable segments. One-off payment per study, no subscription.

500

Prices excl. VAT. From €1000, a senior consultant reviews the report before delivery. Secure card payment via Stripe.

Total1 390 €
Per respondent2,78 €
Simulation precision±4.4 pts
Delivery24 h, expert-reviewed
Compare sample sizes and precision
SamplePer respondentTotalSimulation precisionDelivery
1003 €300 €±9.8 ptsUnder 1 hour
2002,94 €588 €±6.9 ptsUnder 1 hour
3002,89 €867 €±5.7 ptsUnder 1 hour
4002,83 €1 132 €±4.9 pts24 h, expert-reviewed
5002,78 €1 390 €±4.4 pts24 h, expert-reviewed
1,0002,53 €2 530 €±3.1 pts24 h, expert-reviewed
1,5002,30 €3 450 €±2.5 pts24 h, expert-reviewed
2,0002,09 €4 180 €±2.2 pts24 h, expert-reviewed
2,5001,90 €4 750 €±2.0 pts24 h, expert-reviewed

What “simulation precision” means

It tells you how much a result can vary if you run the same simulation again (95% confidence, for a result of 50%). With 100 respondents, a 40% purchase intent can come out between 30% and 50%; with 1,000, between 37% and 43%.

It does not measure distance to the real market: synthetic consumers are not a random sample of real people. For comparisons you will act on, we recommend 400 synthetic consumers or more, and confirming high-stakes decisions with real consumers.

From question to report in four steps.

  1. Frame the question

    Choose the test, enter your concepts, prices or attributes, and define the audience: country, age, sex, income and category.

  2. See the price and precision

    Pick your sample from 100 to 2,500. Price and precision update instantly.

  3. Order securely

    Pay securely. Your synthetic panel answers the questionnaire right away.

  4. Receive the report

    Executive summary, charts, segments, Excel and PDF — in under an hour, or 24 h with expert review from €1,000.

Jacques de Raucourt

AI technology, built on decades of market research.

“I built Synthetic so that every brand can test a decision before it invests in it.”

Holdfast Synthetic was created by Jacques de Raucourt, founder of Holdfast Intelligence. He led commercial strategy and growth for EMEA at Qualtrics and spent more than a decade in strategy and market intelligence leadership at Ipsos and GfK, working with consumer brands across Europe, Latin America and North America. Synthetic puts that experience into every questionnaire, every panel and every report.

Jacques de RaucourtFounder, Holdfast Intelligence

Book a demo

Questions

What is market simulation, and what are synthetic consumers?

Synthetic consumers are AI personas built to reflect the real population of a country: age, income, household, habits and shopping style. A market simulation asks them a standard research questionnaire about your concept, price, product or ad, and analyses their answers like a real study. Holdfast’s AI engine is built on a large language model trained on a vast amount of public text.

How is this different from a survey tool or a general AI chatbot?

A survey tool needs you to recruit and pay real respondents, then wait for fieldwork. A general chatbot gives one opinion, with no audience, no method and no statistics. Synthetic runs a structured study on a representative synthetic panel and returns results, segments and a stability index.

Does it predict how the real market will behave?

No, and we do not claim it. Simulations reproduce general patterns well (which attribute matters most, how sensitive people are to price, which concept is clearer), but synthetic consumers answer more uniformly than real people and are weaker in new or niche categories. Use Synthetic to explore and prioritise; confirm the decisions that matter with real consumers.

If I run the same simulation twice, will I get the same result?

Very close. The same audience always uses the same panel, questionnaires are fixed and the engine runs with minimal randomness. Differences come mainly from panel size: from 400 synthetic consumers, key results typically stay within a few points. Every report shows a stability index.

Which markets are available?

France, Spain, United Kingdom, Germany, Italy, United States, Mexico, Chile, Colombia, Argentina, Brazil, Portugal, Netherlands, Belgium, Switzerland, Austria, Ireland, Sweden, Poland, Canada, Australia, Peru, Japan, United Arab Emirates, Saudi Arabia.

How long does it take?

Usually under an hour after payment. Studies from €1,000 are reviewed by a senior consultant and delivered within 24 hours.

What happens to my concepts and data?

Your material is used only to run your study. It is processed by our AI provider to generate the answers, stored in your account and never shared with other clients. No personal data of real people is involved.

Can I confirm the results with real consumers?

Yes. Every report links to running the winning option with real people on Holdfast Pulse, or to talking to a Holdfast consultant about a full study.

Ask. Test. Decide.

The intelligence to simulate what comes next.

Frame your question in five minutes, see the price before you pay, and get a decision-ready result today.

Secure payment via Stripe · One-off payment, no subscription · Prices excl. VAT

Want to ask real people instead? Holdfast Pulse

See a study from start to finish