Why do I choose this product?
Households and customer groups, grounded in segmentation data, in different situations.
See the profiles →How it works
Profiles you can ask. Answers you can trace.
Svava gives you profiles of consumers, decision-makers and communities. Four experts lead the conversations, and every answer can be followed back to the source material.
Three perspectives
The same way of working, three kinds of profiles.
The experts, the traceability and the way of asking are the same. What differs is who you ask.
Why do I make this business decision?
Source-backed profiles of those who buy, approve and invest.
See the profiles →What does this mean for us, and does it belong in our culture?
Communities with their own values, places, stories and language. The answer shows where the community holds together and where it comes apart.
See the profiles →Experts & profiles
Experts who lead, profiles who participate.
Svava has two kinds of agents. Four expert agents lead the work and guide a process. Profiles are the data-grounded perspectives those processes engage. Choose an expert for your task, bring in the profiles you want to understand and explore the question.
| Verb | Expert | What it does |
|---|---|---|
| Select | ProfileSelector | Helps you choose profiles based on your business scenario, audience or research question. Starts with your need rather than a list. |
| Understand | In-depthInterviewer | Explores one profile in depth to examine attitudes, behaviours and needs. |
| Test | FocusGroup Leader | Leads a focus group with several profiles to explore different reactions and discuss a topic. You can upload material to test. |
| Co-create | WorkshopFacilitator | Facilitates creative workshops where new ideas are generated, evaluated and developed together with the profiles. |
Mosaic group H · Security-oriented families
Profile compass
Purchasing power0.62
Price focus0.78
Social responsiveness0.54
Digital curiosity0.34
Online shopping0.46
Sustainability values0.58
Channels · receptive to
Direct mailLocal mediaEmailSMSSocial media
Messages · resonate
SecurityFamilyLong-term thinkingFrugalityAspirationStatus
1.7×
over-indexing · house ownership
1.4×
over-indexing · gardening
0.6×
under-indexing · major cities
Compare a three-line persona with a perspective supported by 600+ data points behind it.
The method principle
General answers give general strategies.
Web panels and focus groups often produce findings like “everyone thinks it is important”, true but useless. Strategic value does not come from consensus. It comes from contrast.
The wrong exploration
The fluffy answer.
- “What do you think of our offer?”
- “How do you feel about our new campaign?”
- “How do you see our brand?”
Result: vague answers, low contrast, hard to prioritise.
The right exploration
Strategic data.
- “Rank price, quality, design, security, innovation.”
- “What do you sacrifice first, design or sustainability?”
- “What makes a product feel premium rather than trendy?”
- “What makes our offer more relevant than the competitor’s?”
Result: differences, priorities, conflicts, strategic tensions.
Svava is built for the second type of question. It is not a matter of style. It is the difference between useful and useless customer understanding.
Tension map
Where customer groups diverge.
Audience analysis helps you explore where perspectives diverge. The tension map places profiles on axes you define and highlights conflict zones.
AspirationEveryday prioritiesPredictabilityControl
Anders
Lars
Maja
Sara
Erik
Ingrid
Strategic implication
No shared message territory.
The two quadrants containing your core groups do not overlap. A campaign that resonates with one may alienate the other.
Different messages may be needed to address these contrasting priorities.
Constructed example. The profiles are illustrative.
Traceability
Every conclusion, linked to its source.
Follow a conclusion back to the profile answers and the data points linked to them. Review the source material and assess which assumptions need further investigation.
A readable explanation, linked to source material.
For each profile answer, you can open a report showing which data points in the source material have been linked to the answer. This gives you a concrete starting point for reviewing the interpretation.
The report shows the connection to the source material. It does not reveal the model’s internal reasoning or prove that an interpretation or behavioural prediction is correct.
“Visible source material makes a conclusion reviewable. You assess what it supports and what needs to be tested with real customers.”
H
Mosaic group H
“For us, culture comes alive when we travel together, to a city in southern Europe, a museum or a market. But also at home around the kitchen table, discussing history or cooking the food we brought back.”
Traceability reportIllustrative example · 6 data points
Over-indexesHolidays abroad, a prominent travel pattern1.4×
Over-indexesCultural trips, self-directed activities beyond sunshine1.7×
InterestHistory, prominent in leisure surveysprominent
InterestFood culture, prominent in cooking habitsprominent
ValueFamily, central to the group’s self-descriptioncentral
Language choice“We” phrasing: the profile represents a Mosaic group rather than an individualstyle choice
A closed system
The agents do not search the web. By design.
The agents do not search for new sources on the web during profile conversations. They work with the source material you supply. The AI model’s prior knowledge also contributes to the reasoning.
Defined sources, clear source material.
The supplied source material is limited to the profiles’ data and the material in the conversation. When you review a web page using the Chrome extension, the page is supplied as material; the agents do not search further on the web.
The traceability report shows which data points have been linked to an answer. You can review their relevance without mistaking this for full visibility into how the model arrived at its interpretation.
Defined source material makes it easier to review answers and formulate the next customer question.
The web
Svava’s system
Closed by design
AI modelSegmentation dataYour dataTest material
No web searchWhere you already work
Ask the profiles where you already work.
No integration
You log in and start asking. Up and running the same day.
On the page you are looking at
Let the profiles read a web page, a draft in Figma or a page that has not been published yet.
Claude, ChatGPT or Copilot
Via MCP you ask the profiles directly in the assistant you already use. Your own agents can ask the profile before they answer.
The difference
What you do today compared with Svava.
Most tools give you more data or faster answers. Svava gives you better questions and traceable perspectives to make decisions against.
| Web panel | Focus group | Your own bot | Svava | |
|---|---|---|---|---|
| How often | Once a quarter | 1 to 2 times a year | Depends | ✦Any time |
| In-depth | Multiple-choice | Verbatim, deep | Depends on the data | ✦Verbatim with 600+ data points |
| Iterations per question | 1 | 1 to 2 | Depends | ✦10+ the same day |
| Segment resolution | Demographic | Small sample base | Depends | ✦Customer group and your own data |
| Traceability | Aggregate | Verbatim but unsystematic | Depends | ✦Every conclusion traceable |
| Time to answer | 2 to 4 weeks | 4 to 8 weeks | Variable | ✦Minutes |
Book a demo
Book a demo. Bring a real question.
We build a first example together around a concrete question you are actually working on. 30 minutes. No sales deck.