Simulated citizens that answer questions like real people

Surveying people is difficult, slow, and expensive. That's why survey data is used too infrequently to support decision-making.

Kansa.io solves this problem: the entire Finnish population is always at your fingertips, ready to answer your questions. Answers from our synthetic population reliably reflect how real Finns respond on a variety of topics. With us, surveys become an easy part of everyday business.

Why guess when you can ask?

Synthetic Population
n=200

Question

How important is it that a company communicates its environmental impact transparently?

Distribution

12345678910

Mean

7.1

Median

7.0

Std Dev

1.27

Min-max

3–10

Actual answers by AI agents in our Finnish population model

Instant insights in 4 easy steps

1

Write your question

Input the survey items you want answered. Single questions or full questionnaires.

2

Choose your audience

Select demographics, sample size, and any specific criteria for your respondents.

3

Generate results

Receive simulated responses in under a minute. No waiting for recruitment or data collection.

4

Analyze patterns

Explore how different groups answered. Filter and segment by any demographic variable.

What you can do with Kansa

Our product can be utilized for a wide variety of use cases. Some common ones include the following.

Test before you launch

Validate new products, services, and pricing with synthetic consumers. See how attitudes, interest and perceived value vary across demographics. Kill weak ideas early and double down on winners.

Simulate policy reactions

Preview how different citizen groups respond to policy options — from bike lanes to tax changes. Compare support across regions and demographics to identify paths that minimize backlash before any real consultation.

Stress-test your surveys

Iterate on questionnaire design before fieldwork. Identify confusing items, check variance, and see how sampling choices affect your results. Walk into real data collection with a stronger instrument.

How our model works

01

Population data model

We have built a detailed statistical model of the Finnish population, representing its demographics, behavior, and attitudes. This model is based on extensive survey data and research, and reflects not only demographic distributions but also complex interdependencies between various population features.

02

Persona sampling

The data model is sampled for personas to create an agent panel. These agents are like a real random sample of Finns: they have rich profiles of demographic and psychological features that influence their responses.

03

LLM prompting and collation

The agent panel is prompted with survey questions using large language models, and their answers are collated to produce synthetic survey results. Based on our training and validation, these results reliably reflect how real Finns would respond to the same questions.

Early Access

Be among the first to harness synthetic populations

We're at an exciting stage: our model has been validated against real survey data and consistently delivers reliable results. Now we're ready to work with forward-thinking organizations.

Pilot Partner Benefits

  • ✓Direct influence on platform development and features
  • ✓Preferential pricing for early adopters
  • ✓Dedicated support and onboarding from our research team
  • ✓Co-marketing opportunities as an innovation leader

Interested in becoming a pilot partner?

We're looking for organizations that want to be at the forefront of survey methodology. Join us in revolutionizing how businesses gather insights.

Who we are

We're a small team of Finnish psychologists and IT professionals who've spent the better part of a decade in applied behavioral science research. We've seen firsthand how hard it is to get quality survey data when you actually need it.

Research roots

Our background is in academic and applied research, where we learned to be skeptical of easy answers and rigorous about methodology. We bring that same standard to everything we build.

AI pragmatists

We've been working with AI and LLMs since the early days and see through the slop. We're focused on building something that actually works.

Scratching our own itch

Kansa.io is something we would've wanted to exist for years.

Frequently Asked Questions

How accurate is the synthetic population model?
Our model has been rigorously validated against real Finnish survey data on many topics, especially attitudes and opinions. In these validated areas, the synthetic population's responses consistently match actual survey results. However, we're not claiming to have a magical oracle: the model is validated for specific use cases, not every single topic. We're transparent about where the model performs well and where it has limitations, and we can discuss the suitability for your specific questions.
How is this different from traditional surveys?
Traditional surveys require recruiting participants, waiting for responses, and dealing with messy data from real humans. It is a process that takes weeks and costs thousands. Our synthetic population is always available and provides results in minutes at a fraction of the cost. You can iterate on questions, test different phrasings, and explore multiple scenarios without the constraints of traditional survey methods.
What topics can the model reliably answer about?
The model excels at scaling relatively simple inferences about opinions and attitudes that can be reasonably derived from demographic data. Good examples are things like consumer preferences, policy attitudes, and general opinions. However, when questions become very specific or detailed, especially regarding real-world behaviors, there are inherent limits to what can be inferred even from rich demographic data. We're continuously validating new use cases, and for each potential application, we can provide transparency about what the model can and cannot reliably predict.
Can I use this for markets outside of Finland?
Currently, our validated model covers the Finnish population, but we are actively exploring options for expanding to additional markets. If you're interested in a specific population, get in touch. We can discuss timelines and potentially prioritize development based on partner needs.
How do you ensure the model doesn't reflect AI biases?
We build our model from real survey data and demographic statistics, not from LLM training data alone. The underlying population model is statistically grounded, and we use LLMs only for the final response generation step. We continuously validate outputs against real-world data and actively test for known AI biases. Every pilot partner receives detailed documentation about our validation methodology.
What kind of support do pilot partners receive?
Pilot partners work directly with our research team. You'll get dedicated onboarding, help designing your first surveys, interpretation support for results, and regular check-ins. We're looking for partners who want to be involved in shaping the platform, so your feedback directly influences our development priorities.

Ready to make better decisions?

Stop guessing. Start asking. Get insights from our synthetic population model today.