Jason Duke, Founder, Kronaxis
Tag: Research
I built Panel Studio to replace a specific category of research: the expensive, slow, small sample screening phase that sits at the beginning of most research programmes. I did not build it to replace all research. The distinction matters, and being honest about where synthetic panels fall short is more useful than pretending they solve everything.
Focus groups cost between £30,000 and £80,000 per round in the UK when you factor in recruitment, venue hire, incentives, moderation, transcription, and analysis. They take four to eight weeks from brief to report. They produce twelve to ninety six data points. These are real limitations, and they are the reason Panel Studio exists.
But focus groups also do things that synthetic panels cannot do. Here are the five categories where you should still book the room and hire the moderator.
1. Physical products
If your research question involves how a product feels in the hand, you need hands. Taste, texture, smell, weight, balance, the sound a lid makes when it clicks shut, whether a bottle pours cleanly or dribbles: none of these can be simulated. A synthetic persona can tell you whether they find a product concept appealing based on a description. They cannot tell you that the cap is too tight for arthritic fingers or that the material feels cheap.
This applies to packaging, food and drink, cosmetics, consumer electronics, furniture, clothing, and anything where the physical interaction is part of the product experience. If you are testing a new shampoo formulation, no personality model predicts how it feels on someone's scalp. You need people in a room, with the product, reporting what they experience.
Panel Studio's role here is upstream. Test ten packaging concepts with a synthetic panel. Identify the top three. Then produce physical prototypes of those three and put them in front of real consumers. You have just cut your physical testing costs by 70% without losing the physical testing where it matters.
2. Regulatory submissions
Certain regulators require evidence from research conducted with real human participants. The MHRA requires clinical studies with real patients for medicinal product claims. The FDA has similar requirements. The ASA will scrutinise the methodology behind advertising claims and expects consumer research to involve actual consumers.
Synthetic panel data can support a regulatory case. It can demonstrate that you tested a broad range of personality types and demographic segments. It can show that your claims hold across different population subgroups. But it cannot replace the primary requirement for human participant data when a regulator explicitly demands it.
The same applies to academic publication. If your study design requires human ethics approval and participant consent, synthetic data is supplementary, not primary. Reviewers will (rightly) distinguish between "we tested this with 500 synthetic personas" and "we tested this with 500 recruited participants."
Panel Studio's role here is in the design phase. Test your stimuli, refine your questions, identify which variables to measure, and screen hypotheses before committing to the cost and timeline of formal human research. One synthetic screening pass saves one round of expensive real participant testing that would have produced ambiguous results.
3. Extreme niche markets
Synthetic panels are built from census weighted demographics. They reflect the population distributions of the countries they represent. This means they work well for markets that align with broad demographic segments: UK adults aged 25 to 45, US households with income above $75,000, German consumers in urban areas.
They work less well for extreme niches. If your target market is "left handed violinists over 60 who live in rural Wales," the intersection of those attributes is too narrow for census weighted generation to produce meaningful signal. The panel builder can create personas that match those demographic filters, but the life experiences, preferences, and decision patterns specific to that exact intersection come from the broader population model, not from deep domain data about elderly Welsh violinists.
The threshold is roughly this: if your target market represents at least 1% of the national population, a synthetic panel of 200 or more personas will produce useful signal. Below that, the personality model is doing the work rather than domain specific knowledge, and you should weight the results accordingly.
For niche markets, the better approach is qualitative: find ten actual people in your niche, interview them deeply, and use the insights to calibrate your understanding. Synthetic panels can then test whether those insights generalise to adjacent segments.
4. Group dynamics
Synthetic personas respond independently. Each persona receives a stimulus, processes it through their personality profile and life history, and produces a response. They do not interrupt each other, build on each other's ideas, form coalitions, defer to dominant speakers, or change their minds because someone made a compelling argument in the moment.
Real focus groups are messy precisely because people interact. A well moderated group generates ideas that no individual would have produced alone. Someone makes a half formed observation, someone else builds on it, a third person challenges it, and by the end of the exchange the group has articulated something none of them came in thinking. This emergent property of group interaction is one of the genuine strengths of traditional qualitative research.
If your research question is about how a group negotiates ("how would a family decide which holiday to book?"), how a team reaches consensus ("would this procurement committee approve this vendor?"), or how social influence shapes decisions in realtime, you need a group of real people in a room.
Panel Studio's focus group feature does simulate multiple personas discussing a topic in rounds, and it produces useful structured output. But the simulation is turn based and orderly. Real groups are neither. The chaos is the point.
5. Emotional depth and serendipity
A skilled qualitative researcher does things that no structured system can replicate. They notice that a participant's body language contradicts their words. They follow an unexpected tangent that reveals a latent need the discussion guide never anticipated. They adjust their probing in realtime based on a hunch that there is something deeper behind a surface level answer.
Synthetic personas give personality driven responses that explain the decision mechanics behind their preferences. This is more analytically rigorous than most focus group output. But it is not serendipitous. The persona answers the question asked. It does not volunteer the adjacent insight that a human participant might offer when a good interviewer creates space for it.
If your research is exploratory, where you do not yet know what the right questions are, spending time with real people in unstructured or semi structured conversation is still the best way to discover what you should be asking. Once you know the questions, synthetic panels answer them faster, cheaper, and at scale.
When synthetic panels are the better tool
For everything not covered above, synthetic panels outperform traditional focus groups on every operational metric.
Speed. A Panel Studio stimulus returns results in under twenty minutes. A focus group returns results in four to eight weeks. If speed matters to your decision, and it almost always does, that difference is decisive.
Cost. A 500 persona Panel Studio panel costs under £100 on the Starter tier. A 12 person focus group in London costs £30,000 to £80,000 all in. The cost difference is not incremental. It is structural.
Scale. A focus group gives you 12 data points. A synthetic panel gives you 500. Subgroup analysis that is statistically meaningless with 12 participants becomes reliable with 500. You can segment by personality dimension, age band, income level, region, and gender simultaneously. Try that with twelve people.
Reproducibility. Run the same panel again with the same stimulus and you get the same results. Change one variable and you isolate its effect. Focus groups are unrepeatable: different participants, different moderator dynamics, different day of the week. You cannot A/B test with a focus group.
Causal explanation. Panel Studio reasoning traces tell you which personality dimensions drove each response and why. Focus group transcripts tell you what people said, filtered through social desirability and group dynamics. The gap between stated reason and actual driver is where most qualitative research misleads.
Multi country coverage. Panel Studio supports 20 countries. Run the same stimulus against UK, US, German, French, and Japanese panels simultaneously and compare personality driven responses across cultures. Doing this with real focus groups means five recruitment agencies, five venues, five moderators, five translation passes, and a six figure budget.
The hybrid approach
The most effective research programmes use both. Synthetic panels for hypothesis generation, rapid screening, and quantitative personality segmented analysis. Real focus groups for final validation, emotional depth, and physical product interaction.
This is not a diplomatic both sides framing. It is a cost optimisation. A typical research programme runs three to five rounds of screening before reaching a final concept. Each round costs £30,000 to £80,000 with traditional methods. Replace the first four rounds with synthetic panels and you cut total research spend by 80% while keeping the depth exactly where it matters: at the end, when you are validating one strong hypothesis rather than exploring twenty weak ones.
The £40,000 question is not whether to use synthetic panels or focus groups. It is which rounds of your research programme actually require twelve people in a room, and which can be done better, faster, and cheaper with five hundred synthetic personas who explain their reasoning.
For most programmes, the answer is that one round requires the room. The rest do not.