curly
A voice-first consumer research platform whose AI conducts hundreds of adaptive spoken interviews in hours, then codes and analyses them both qualitatively and statistically, with every finding traced back to a real participant quote.
What is curly?
curly is a consumer research platform in which a conversational voice AI runs the interviews itself. The pitch is a direct answer to a familiar trade-off: quantitative surveys scale but do not explain, qualitative interviews explain but do not scale. curly claims to deliver both at once, advertising 100+ deep customer interviews over a weekend, samples of n=500 and above, and five times more net-new insight than a survey.
The workflow has three stages. Setting up a study starts from an abstract question and some context; the platform helps turn that impulse into a concrete interview guideline, and the site insists that no research PhD is required. Studies support open questions, single and multiple choice, scale questions and A/B tests, and accept rich stimuli — images, video, Figma prototypes or a live product — so participants react to real creative work rather than to a description. Interviews then run in parallel in the participant's browser, in audio and video, for up to sixty minutes each, with the AI deciding where to probe. The company's example is telling: a respondent who scores something four out of five is asked what would have made it a five.
The third stage is analysis. Conversations are converted automatically into a structured codebook, responses are tagged, variables are identified and statistical significance is tested across the whole dataset. Results can be read question by question or participant by participant, interrogated through a chat interface, and exported as reports in which every number links back to the quote, screen interaction and audio file behind it. Quality controls are part of the sell: participants come from established premium panels or the client's own base, and Integrity Agents screen each interview for inconsistency, rushing and fabricated answers.
The platform is browser-only, sold in monthly interview-hour packages, and operated by curly Intelligence GmbH in Munich. Optional human support ranges from a second opinion on study design to a fully run project, and participant recruitment can be bought as a service.
What it does
- Run hundreds of adaptive voice interviews in parallel and get results within hours rather than weeks
- Turn an abstract research question into a structured interview guide with AI assistance
- Ask intelligent follow-up questions that chase the reasoning behind an answer instead of recording a score
- Present images, videos, Figma prototypes or a live product as stimuli and capture reactions to them
- Convert every conversation into a structured codebook automatically, with variables and significance testing
- Slice results by question, participant, segment or hypothesis, and query the dataset in a chat interface
- Generate downloadable reports in which each finding links back to the original quote, screen interaction and audio
When to use curly / When not to
A quick filter to help you decide if curly is the right fit.
When to use curly
- Insights and market research managers at consumer brands who have the budget and the panel but still cannot explain the why behind survey answers
- Qualitative and UX researchers whose interview practice is bottlenecked by recruitment and manual analysis rather than by method
- Product managers and innovation teams validating concepts, features and Figma prototypes before engineering commits to them
- Marketing and brand teams that need to test campaigns, positioning or the drivers behind an NPS score with real consumers rather than an eight-person focus group
- Customer experience, retention and strategy teams investigating churn, win/loss patterns or a new market at a fraction of an institute's cost and timeline
When not to use curly
- Small teams or one-off projects: the entry package is 600 EUR per month billed annually in advance, so the first invoice is 7,200 EUR
- Anyone expecting a free plan or a free trial, since neither exists and self-service sign-up is not open — accounts must be approved by the vendor
- Engineering teams wanting to pipe results into their own stack: there is no public API, no documentation portal and no connector
- Mobile-first users, as the product is a browser application with no iOS or Android app
- Buyers whose procurement requires published terms of service, a downloadable DPA or a security certification such as ISO 27001 or SOC 2, none of which the site publishes
How to use curly
A typical end-to-end flow, from setup to results.
- Try the public self-service demo at demo.curly.de to experience an AI-led interview as a participant, without an account
- Book a call through the site: the Get curly buttons open a Google Calendar scheduler, and dedicated pages offer 15, 30, 45 or 60 minute slots with either founder
- Have your account approved by the vendor — the platform is not open to self-service sign-up
- Bring a decision you need to make: a concept, a feature, a prototype, a live flow or a pricing question is enough to start
- Let the platform turn that question and its context into a structured interview guideline, then set the parameters of the study
- Choose the question types and attach your stimuli — images, video, a Figma prototype or your live site for task testing
- Recruit participants by sharing the interview link with your own audience, or buy curly's panel service
- Let the interviews run in parallel and watch insights appear as they complete
- Slice the results by question, participant, segment or hypothesis, and ask curly to test a specific hypothesis in the chat
- Download the report and trace any finding back to the quote, interaction or audio recording behind it
Pros & Cons
Pros
- Combines qualitative depth and quantitative scale in a single study, with significance testing on the same dataset
- Turnaround measured in hours: the site promises results the next morning and a hundred interviews over a weekend
- Full traceability — every number links back to the quote, excerpt and audio recording that produced it
- Public, readable and degressive pricing, unusual in market research: no base fee, no platform fee, and hourly rates falling from 60 EUR to 32 EUR
- Unlimited studies and unlimited seats on every package, with full product access and no modules
- Data stored in Germany and AI processing kept inside the EU, with the entire sub-processor chain named in detail
- A public self-service demo lets a prospect experience the interview before talking to sales
Cons
- No terms of service and no privacy policy on the site itself — only a legal notice and a cookie policy — which makes contractual review difficult
- Annual billing paid upfront turns the advertised 600 EUR per month into a 7,200 EUR first commitment
- No free plan, no free trial, and no self-service sign-up: accounts must be approved by the vendor
- No API, no documentation portal and no mobile app, so results cannot be pulled into another system automatically
- No published security certification such as ISO 27001 or SOC 2
- No documented way to exclude customer data from model training, and no overall vendor commitment on the subject
- A very young company — the legal entity was registered at the end of 2025 and LinkedIn lists two to ten employees — with three published case studies
Pricing & Plans
There is no free plan and no free trial. Pricing is public and built entirely on interview-hour packages, with no base fee, no platform fee and no entry ticket. The cheapest paid entry point is 600 EUR per month for ten interview hours, which works out at 60 EUR per hour. Larger commitments lower the hourly rate: 960 EUR for twenty hours at 48 EUR per hour, 2,000 EUR for fifty hours at 40 EUR per hour, and 3,200 EUR for a hundred hours at 32 EUR per hour. Beyond a hundred hours, or for complex target groups, more than two hundred interview hours or custom reporting, the price is quoted. One point deserves attention: billing is annual and paid upfront, so the real first commitment at the entry package is 7,200 EUR rather than 600 EUR. Every package includes full product access without modules, unlimited studies, unlimited seats, all major languages and EU hosting. Project support, end-to-end study support and participant recruitment are sold separately on quotation.
- 10 interview hours per month — 600 EUR per month
- 60 EUR per interview hour
- 20 interview hours per month — 960 EUR per month
- 48 EUR per interview hour
- 50 interview hours per month — 2
- 000 EUR per month
- 40 EUR per interview hour
- 100 interview hours per month — 3
- 200 EUR per month
- 32 EUR per interview hour
- More than 100 interview hours per month — priced on request
- Project (support for a standalone study)
- curly Service (end-to-end study support) and Panel (participant recruitment)
- All packages are billed annually
- in a single invoice paid upfront
Data, GDPR & hosting
A consolidated view of how curly handles your data.
GDPR overview
curly states GDPR compliance explicitly and repeatedly: the pricing page carries an EU-hosted, GDPR-compliant line and the strategy page promises GDPR-compliant EU hosting. The participant privacy notice enumerates the full set of data subject rights — access, rectification, erasure, restriction, portability, objection, withdrawal of consent and complaint to a supervisory authority — and names Articles 6(1)(a), 9(2)(a), 28, 45 and 46. Third-country transfers are addressed directly: processing and storage remain in the EU, but the vendor acknowledges a theoretical possibility of US access through its providers' parent companies and invokes standard contractual clauses or the EU-US Data Privacy Framework. Two gaps matter. The site itself publishes no privacy policy and no terms, only a legal notice and a cookie policy. And no data protection officer is named for curly: the DPO fields in the participant notice are placeholders the client fills in.
Who owns the data?
The site publishes no terms of service, so contractual ownership of customer data is never stated. The only document that describes the arrangement is the privacy notice served to interview participants, and it is written from the client's point of view: the client is the data controller and curly Intelligence GmbH acts as its processor under an Article 28 GDPR data processing agreement. On that reading the commissioning brand owns the interview data and curly may only process it on instruction. Tenant separation is claimed to be technical and logical, so one client's recordings are not exposed to another. Nothing published addresses intellectual property in the resulting analyses, reports or codebooks.
Reuse rights
No terms of service are published, so no reuse rights are granted or withheld in writing. What the participant privacy notice does describe is the processing chain. Audio recordings, automatic transcripts, AI-generated analyses, participant master data such as panel identifiers, campaign parameters and a consent log are processed on the basis of participant consent under Article 6(1)(a) GDPR, with Article 9(2)(a) covering any sensitive information a participant volunteers. The named sub-processors are Supabase for the database on Amazon Web Services infrastructure, Google Cloud Platform for the application server and for Vertex AI and Gemini, ElevenLabs for speech recognition and transcription — which in turn relies on Anthropic — Microsoft Azure OpenAI for text analysis, chat, embeddings and Whisper transcription, and AWS Bedrock running Claude Opus for report generation. curly states that data sent through Bedrock is not shared with the model provider and not used to train models, and that ElevenLabs runs under enforced zero retention. It makes no equivalent statement about Azure OpenAI or about its own use of customer data, and it documents no opt-out mechanism.
Data retention & training
Hosting summary
Data is hosted in the European Union, and specifically in Germany. The pricing page states EU hosting and GDPR compliance, and the participant privacy notice gives the detail: the application server runs on Google Cloud Platform, the database on Supabase over Amazon Web Services infrastructure, and the server location is the Frankfurt am Main region. Audio recordings and transcripts are stored exclusively in Germany, in AWS and Supabase eu-central-1 and Google Cloud europe-west3. AI processing stays inside the EU as well: text analysis through Microsoft Azure OpenAI in West Europe, screen recording analysis through Google Vertex AI in the EU multi-region, speech recognition through ElevenLabs under EU data residency with enforced zero retention, and report generation on AWS Bedrock in eu-central-1. The vendor states that no transfer to a third country takes place, while acknowledging a theoretical possibility of US access through its providers' parent companies, covered by standard contractual clauses or the EU-US Data Privacy Framework. Security measures published are TLS 1.3 in transit, AES-256 at rest and strict multi-tenant separation.
Things to keep in mind
Risks and trade-offs to weigh before adopting curly.
- Delegating the interview to an AI removes the human interviewer's judgement: the follow-up questions that are never asked are invisible, and a researcher who stops moderating may gradually lose the craft that lets them spot when a conversation has gone wrong
- Statistical significance applied to a few hundred AI-led conversations is persuasive by construction; a coefficient attached to a quote is easy to present and hard to challenge, which invites more confidence than the method may warrant
- The vendor publishes no terms of service and no privacy policy of its own, so ownership of the data, of the transcripts and of the resulting analyses rests on a data processing agreement that has to be requested rather than read in advance
- There is no documented way to exclude customer data from model training and no overall commitment from the vendor on the subject; the only explicit no-training statement covers one sub-processor leg
- Participants may disclose health, political or religious information during an open-ended conversation, and the vendor says so itself — a real risk in the sensitive consumer categories the product is sold into
- Speed is the selling point, and it can become the pressure: results promised by the next morning encourage decisions taken before anyone has questioned how the sample was recruited or how the question was framed
- The brand name is extremely generic — a live and unrelated product called Curly exists on another domain, and the .de domain itself was somebody else's personal homepage until recently — so verify you are dealing with curly Intelligence GmbH of Munich
Setup & Integrations
Technical difficulty
Very low on the technical side, moderate on the access side. Nothing to install: the platform and the interviews both run in the browser, studies are distributed through a shareable link, and there is no API, connector or deployment to configure. Setting up a study is advertised as taking minutes, starting from an abstract question, with the platform generating the interview guideline and the site stating that no research PhD is required. The friction is administrative rather than technical: sign-up is not self-service and an account must be approved by the vendor. Paid human support is available for complex studies.
Deployment
Integrations
Supported languages
Behind curly
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does curly actually do?
How fast are the results?
What does it cost?
Is there a free plan or a free trial?
Who recruits the interview participants?
Which languages are supported?
Where is the data hosted?
How long is interview data kept?
Is there an API or a mobile app?
How does curly prevent fake or careless answers?
Should you pick curly?
curly attacks a genuine and well-framed problem: qualitative research does not scale, quantitative research does not explain. Its answer is coherent from end to end — an adaptive voice AI that conducts the interview, automatic coding into a structured codebook, significance testing, and traceability down to the audio file behind every number. The quality controls are unusually explicit for this category, with panel-sourced participants and automated screening for fabricated answers.
The pricing is public, readable and degressive, which is rare in market research and welcome. It is also less gentle than it looks: billing is annual and paid upfront, so the advertised 600 EUR per month is in practice a 7,200 EUR commitment, and there is neither a free plan nor a trial to soften the decision. The self-service demo partly compensates.
The reservations are mostly about paperwork rather than product. The vendor publishes a detailed sub-processor chain, German data residency and a three-year retention rule in the privacy notice served to interview participants — but publishes neither terms of service nor a privacy policy of its own, carries no security certification, and documents no way to exclude customer data from model training. The legal entity is very young and the team small, which is normal at this stage but worth knowing before signing an annual commitment.
For an insights, product or brand team that already runs research and is limited by throughput rather than by method, curly is worth a serious look — starting at the smallest package, and with a legal review of documents that will have to be requested rather than downloaded.
- Choosing a selection results in a full page refresh.
- Opens in a new window.