Frequently asked questions.

Kesmet is a synthetic market research platform. We've built around 500 AI-generated personas, each grounded in real demographic, economic, and psychological data from Flanders. You query them for market insights — adoption probability, willingness-to-pay, objections, segment fit — in minutes rather than months.

You design a survey — a few questions about what you want to learn. Kesmet runs it past a panel of 15–25 synthetic personas drawn from our population model. Each persona answers based on their demographics, life stage, values, personality, and channel behavior. You get a structured result back in under a minute.

One credit equals one persona answering one question. A 15-persona panel × 3 questions = 45 credits per survey. Subscription credits reset on your billing date. Top-up pack credits never expire.

Under a minute for most runs. The bottleneck is panel size and depth of behavioral inference, not waiting for human respondents.

Yes. Each result includes the panel composition with persona profiles you can browse. We're transparent about what's direct from data and what's inferred.

Subscribers (Starter, Pro, Team) get permanent storage and re-run capability. Free tier conversations expire after 7 days.

No. Personas are statistical reflections built from authoritative public data — not real individuals. No one whose data we use is identifiable in our model.

Public sources. Statbel (Belgium), Statistiek Vlaanderen, Eurostat, the European Values Study, the European Social Survey, Eurobarometer, Hofstede cultural dimensions, Big Five personality literature, DataReportal, and others. We list our sources transparently in the product and on our methodology page.

Honest answer: Kesmet is a directional instrument, not a substitute for polling. We've validated the population against a public Flemish reference survey. Where it's strong: demographic and household composition mirror Flanders closely, and the panel tracks real responses well on practical, material questions — financial security, contact and exposure, broad concept and pricing reactions. Means and rank-ordering usually point the right way, which is what most early decisions need. Where it's weak: on emotionally or politically charged attitudes — migration, identity, political satisfaction, perceived safety, subjective well-being — synthetic personas skew more moderate and uniform than real people, so we do not claim the panel is representative on those. We model in clusters because that's how purchasing behavior moves, and we surface the panel composition so you can see exactly who answered. Read results as a strong steer, not a precise measurement.

Continuously, on a rolling cadence. National statistics refresh annually. Eurobarometer and ESS roughly biannually. Eurostat HBS roughly every five years. We re-ingest as sources update, so the model reflects the world today.

For dimensions where direct data is thin — impulse buying patterns, decision style, feature priorities — we infer using consumer-psychology frameworks (Big Five, Schwartz Maximization Scale, Verplanken impulse-buying tendency, Rogers Diffusion of Innovations, Hofstede). Inferred signals are tagged in the result so you can tell them apart from direct data.

Tell us. There's a feedback button on every result page. We use feedback to refine the model — and during our bootstrapped phase, we reward feedback with extra credits.

No, and we won't pretend otherwise. When stakes are existential, talk to real people. Kesmet is for the dozen smaller decisions you can't afford to validate that way: prices, positioning, audience fit, message variants, feature priorities. The decisions that today get made on a hunch.

Most “would they / wouldn't they” questions. Common ones: would my audience pay €X for this; what's my pricing ceiling; who's my strongest segment; what would make this product fail; would they switch from a competitor; which message lands hardest. The clearer your question, the sharper your answer.

English, Dutch, and French at launch. The persona model reflects the regional and dialectal texture of Flanders — from Antwerp to West-Flanders.

Custom audiences are on the near-term roadmap, prioritized for creators with niche followings. Default Flemish panels at launch.

Not yet. We'll build it when there's clear demand.

Yes — copy, image, link, and PDF on every result. Team plans include branded exports for client deliverables.

Free: 20 credits, no card. Starter: €29/month, 400 credits. Pro: €59/month, 2,000 credits. Team: €199/month, 6,000 credits, 5 seats. Top-up packs (€19 / €39) on any plan.

Subscription credits reset on your billing date. Top-up pack credits never expire.

Yes. Monthly plans cancel at end of period. Annual plans cancel at year-end. No retention games.

Buy a top-up pack, or upgrade to a higher tier. Credits from packs never expire.

We refund unused subscription time on a pro-rata basis, no questions asked. Email us.

Annual billing saves 15% on every tier. Team is our agency / accelerator plan at €199/month. For larger or custom needs, contact us.

Yes. We don't train on your queries. We don't share them with anyone.

Account data for billing and product features. Query data for delivering your survey results. That's it. We don't sell, don't train on, don't share.

Yes. We're based in Belgium and built under GDPR from day one.

The bare minimum: a session cookie for keeping you logged in (essential, always on), and — only if you accept our cookie banner — privacy-friendly analytics. Decline and we load no analytics at all. No tracking pixels, no advertising cookies.

Anthony De Smet and Luana Keszeg, in Belgium, 2026.

We live here. We can validate against reality every day. It's the right size — small enough to model deeply, large enough to matter. Depth is the moat. Each new region we add will hold up to the same standard.

We don't have a public roadmap yet. Likely sequence: the rest of Belgium next, then neighbouring markets, then broader EU. We open new regions only when each holds up to the Flanders quality bar.

Not yet. We're focused on getting v1 right. If anything changes, we'll say so on the homepage.

anthony@kesmet.org · luana@kesmet.org — we answer.