TL;DR: Toluna Start is an end-to-end consumer intelligence platform: a global panel of 79M+ consumers, a drag-and-drop survey builder, automated targeting, dashboards, an AI probing feature (QProbe), and — increasingly — over a million synthetic personas for simulated research. Koji is an AI-native research platform that does one thing with unusual depth: it runs AI-moderated voice and text interviews with your real customers, then analyzes them automatically into a one-click report. The core difference is philosophical. Toluna is built to push large quantitative studies through a panel quickly, with synthetic personas to scale further. Koji is built to extract qualitative truth from real human voices at interview depth — fast, affordable, and without a synthetic shortcut. If you need national B2C survey volume, Toluna's panel delivers. If you need to understand why your customers do what they do, Koji is the modern choice.
Toluna vs Koji at a glance
| Toluna Start | Koji | |
|---|---|---|
| Category | Consumer insights panel + survey suite | AI-native interview platform |
| Audience | 79M+ panel + your own audience | Your own audience (or routed sample) |
| Primary method | Automated quantitative surveys | AI-moderated voice/text interviews |
| AI depth | QProbe follow-ups + synthetic personas | Conversational AI moderator on every answer |
| Real vs synthetic | Real panel + 1M+ synthetic personas | Always real human respondents |
| Analysis | Dashboards + recommendations | Automatic thematic analysis + report |
| Pricing | ~$30k/yr average, contact sales | Free tier; €29–€79/mo, self-serve |
| Best for | National B2C quant at panel scale | The qualitative "why," fast |
What is Toluna?
Toluna positions Toluna Start as "the industry's first and only end-to-end consumer intelligence platform," supporting both quantitative and qualitative research through self-service or custom programs. Its panel spans 79+ million consumers, with 350+ profiling options to target audiences (or you can bring your own via weblink, email, or embedded widget). The survey builder is drag-and-drop with advanced methodologies — timed exposure, heatmaps, text highlighters — and includes QProbe, an AI-driven feature that detects insufficient answers and auto-generates personalized follow-ups. More recently, Toluna has leaned into over 1 million validated synthetic personas to "accelerate and scale research with AI."
Toluna's strengths are reach and breadth: a large panel, a mature survey toolkit, and dashboards with recommendations. Its trade-offs are cost and orientation. The average Toluna contract runs around $30,000 per year, and reviewers note costs can be steep for small businesses or detailed projects. And while QProbe adds intelligent follow-ups, the platform's center of gravity remains the automated quantitative survey — with synthetic personas increasingly offered as a way to simulate respondents rather than talk to more real ones.
What is Koji?
Koji takes the opposite bet: that the highest-value insight comes from real conversations with real customers, and that AI's job is to conduct and analyze those conversations — not to replace the human on the other end.
- AI-moderated voice and text interviews. Koji's AI interviewer runs a natural two-way conversation, probing follow-ups in real time across the whole study — the depth of a researcher-led interview, at the scale of a survey. See the complete guide to AI qualitative research.
- Six structured question types.
open_ended,scale,single_choice,multiple_choice,ranking, andyes_no— so you get scores you can chart (like NPS or satisfaction via scale questions) and the open-ended reasoning behind them in one conversation. - Automatic thematic analysis. Every transcript is coded and clustered into a per-question codebook automatically — see how to analyze qualitative data.
- One-click reports, grounded in real quotes from real people.
This matters because the market is moving toward qualitative depth, not away from it. 57% of researchers report growing demand for qualitative research — yet online qualitative still represents only about 6% of market research revenue, a gap that exists precisely because qual has historically been slow and expensive to run and analyze. Koji closes that gap with AI moderation and automatic analysis. (See also conversational surveys.)
Real voices vs synthetic personas
The sharpest dividing line in 2026 is authenticity. Toluna's million-plus synthetic personas promise speed by simulating respondents. That can be useful for early directional gut-checks — but synthetic personas can only recombine what models already assume about people; they can't surprise you with the thing you didn't know to ask. Koji's entire premise is the opposite: always real humans, just moderated and analyzed by AI. When the cost of a real conversation drops to minutes and euros, the case for synthetic shortcuts gets much weaker. (We dig into this trade-off in our analysis of AI personas vs real interviews.)
Where Toluna wins
- National B2C panel reach across 79M+ members and 350+ profiling attributes.
- Mature quant tooling — timed exposure, heatmaps, advanced survey methodologies.
- One-stop enterprise suite for teams that want panel, fielding, and dashboards under one contract.
Survey fatigue is a real constraint on that quant machine, though: completion rates fall from roughly 89% at 10 questions to 79% at 40, so even a great panel can't rescue an over-long questionnaire.
Where Koji wins
- Qualitative depth at scale — interview-grade conversations, not just survey grids.
- Speed — from question to insight in hours, not the weeks a managed panel study can take.
- Real respondents only — no synthetic-persona simulation.
- Analysis included — automatic thematic analysis, no separate analyst step.
- Transparent, self-serve pricing — start free, no $30k contract.
Pricing: enterprise contract vs self-serve
Toluna runs on annual contracts averaging around $30,000/yr, with pricing gated behind sales. Koji is self-serve and transparent: a free tier with 10 credits, then Insights at €29/mo and Interviews at €79/mo, with per-conversation credits (text = 1, voice = 3) and a quality gate so only substantive conversations are billed. A team can validate Koji on real customers before Toluna would even return a quote.
Which should you choose?
If your work is national consumer quant at panel scale, Toluna's panel and survey suite are built for it. If your work is understanding real customers in depth, quickly, with the analysis done for you — and you'd rather talk to real people than synthetic personas — Koji is the AI-native platform purpose-built for that. Many teams keep a panel relationship for big quant waves and run their continuous, qualitative discovery in Koji.
Three scenarios where Koji fits better
Continuous discovery. A product team that wants weekly signal cannot wait on a panel field-and-tabulate cycle. Koji runs always-on interviews with real users and refreshes the thematic report as responses land — discovery that keeps pace with your sprints.
Win/loss and churn. These require talking to specific named accounts, not a representative consumer panel. Koji interviews your actual buyers and churned users, probes the real decision drivers, and codes the reasons automatically — the kind of depth a synthetic persona simply cannot fabricate.
Concept and message testing on your own audience. Show a concept, capture a scale rating, then let the AI probe the reasoning behind it — open-ended "why" attached to every score, in a single pass instead of a survey plus a separate follow-up wave.
In all three, the job is not national quant volume; it is depth from the right people, fast. That is where an AI-moderated interview beats both a long panel survey and a synthetic-persona simulation: real humans, real reasoning, analyzed for you in hours rather than weeks.
Start with Koji
Talk to real customers, not simulations — and let AI do the moderating and the analysis. Create a free Koji account, point the AI interviewer at your own audience, and get a thematic report from genuine conversations in hours. From question to insight, 10x faster — no synthetic shortcuts, no analysis backlog, no enterprise contract.
