# iMario > Synthetic audiences built from real people. Ask anyone anything before you decide: iMario turns your own customer data and 5.5 billion people in 61 markets into synthetic audiences calibrated on real data, then runs surveys, interviews, focus groups and tests on them in minutes. iMario turns real human data into Synthetic Individuals: simulated people, organised into audiences, that a team or an AI agent can question before a decision ships. Which price, which message, which market first. Every audience is built on three layers, each calibrated on real data: Foundation is who is in the audience and in what proportions, sampled from real population distributions; Mind is what they believe, fitted to real survey data from the same population; Behavior is what they have actually done. Each individual keeps a memory of what it has said. Every answer says whether it came from that person's own data or from a modelled profile. The brand line is The Human Layer of AI: the same reusable individuals serve an insights team running a survey, a product team testing a flow, a sales team rehearsing a call, and an AI agent that consults them through MCP or the API before it acts. Every docs page is also served as Markdown at its URL plus ".md", and the docs links below point to those copies. llms-full.txt holds all of them in one file. **Audiences** - Your customers: connect Zoom, HubSpot, Salesforce, Intercom, Qualtrics, Google Drive and other systems, or upload interview transcripts, survey exports, CRM records and support tickets. When a record, answer or transcript says who it is about, through a customer ID, an email or a named attendee, iMario files it under that customer, so one customer's call and survey answer belong to one individual. Connectors are read-only, each permission is approved before anything is read, and iMario never writes back. Your data stays private to your workspace and is never used to train models for other customers. - Any market: 5.5 billion simulated people across 61 markets, including the US, Canada, the UK and EU, China, Japan, Korea, India, Southeast Asia, Brazil, Australia and the Gulf, built from official statistics and calibrated on real survey data. Pick a market, describe the group you need, and the audience exists. A market audience can be cut into narrower sub-groups, and so can most official audiences. - Official audiences: groups iMario has already built, each from real data about that group, such as survey answers and the way its members talk. Current ones: physicians in the US, China and the UK; new parents and high earners in the US; US professionals who use AI at work; urban graduates in China; consumers and parenting families in Japan. Some of these groups are slow and expensive to recruit: 300 real physicians can take weeks and cost tens of thousands. New ones are added regularly. - Languages: ask in the language you write in and respondents answer in it. **Research** - One box instead of a canvas: type what you want to learn, attach what you want tested, and choose who answers. A run keeps going after you close the tab. - Methods: a single question, a survey, interviews with follow-up probes, a focus group where participants react to each other, a preference (A/B) test of text or images, a creative test on an ad or design, and feedback on a live web page. Testing an AI agent against synthetic customers is coming soon. - Explore: see who gave an answer, cross it with any other question, save that group as its own audience and ask it the next question, or ask one respondent why in a conversation. - Report: a written report over the runs you choose, with every claim traced back to respondent quotes. Focus groups are read as a group. Reports can be shared as a link or exported. - Validate: send the same questionnaire to real people from inside the result and compare the two side by side. **Accuracy** - On iMario's public benchmark, synthetic audiences match real survey answer distributions with 89% accuracy (88.9, scored as 1 minus the total variation distance). The benchmark covers 11 populations, 922 questions and about 47,000 real respondents: Pew American Trends Panel 2024, Pew Global Attitudes 2025 in eight countries, CGSS 2023 in China and the Stack Overflow 2025 developer survey. - Two real samples of the same population agree about 93%, so 89% is about 95% of what a human rerun achieves. Off-the-shelf model personas score 61 to 64% on the same sets. - With a recent real survey on the same topic to reference, the score passes 90%. When the audience is built from your own interview transcripts, the score reaches 96%. - Asked the same question in 12 different wordings, an individual gives the same answer 95% of the time. - The method, datasets and raw files are published and reproducible. **Pricing** Credits only. There are no plans or tiers, and free and paid credits run the same product. - Free: 1,000 credits when you sign up, no card required. Credits never expire. - One-time top-ups: $0.05 per credit, from 500 to 4,999 credits. - Monthly packs: 5,000 credits for $200, 10,000 for $400 or 20,000 for $800 a month, which is $0.04 per credit. Cancel any time; unused credits roll over. - Research, per respondent per question: single-choice 1 credit; multi-select and ranking 1 per option; open questions, image and creative tests 2; interview questions with follow-up probing 5. A web page test is 5 per respondent. A focus group is 5 per participant per topic. - Reports: 50% of the first 1,000 run credits a report covers, 25% up to 5,000 and 10% above, with a minimum of 20 credits. A report generation that fails is refunded. - Explorer and follow-up: an AI read of a selected group is 10 credits; following up with a respondent is 1 credit per message. - Free of charge: building audiences and the individuals in them, and writing surveys and discussion guides. - Examples: one open question to 50 people is 100 credits ($5). A 300-person survey of 12 single-choice questions is 3,600 credits, and its report is 1,150 more. - Audience access is bought separately, in dollars. Each workspace includes two markets, and each further market is $200 once. Official audiences are a one-time purchase: new parents, urban graduates and the two Japanese audiences are $500 each, high earners and US professionals who use AI at work are $1,250 each, and physician audiences are arranged with the team. Studies on them use credits as usual. - Enterprise: the same product plus a forward deployed engineer who builds your audiences and studies, white-label reports, scheduled data refresh, data residency, SAML SSO, SCIM, RBAC, audit logs and a dedicated API. Priced on committed volume. **Use iMario from an AI agent** - MCP server: `https://mcp.imario.ai/mcp` (Streamable HTTP). ChatGPT, the Claude apps, Claude Code and Codex connect by approving the connection in iMario. Scripts, CI jobs and your own agents use the workspace API key from Settings > MCP & API. - One-line setup for Claude Code, Codex or Cursor: paste `Connect iMario to this agent: imario.ai/SKILL.md`. The setup sheet tells the agent to register the server, install the `imario` skill and ask you to approve the connection in your browser. - The `imario` agent skill teaches an agent to pick an audience, estimate the cost, run the study and report counts and quotes. Install it with `npx skills add laugherxiao/imario-skills`. - Every study is estimated before it runs, one daily limit covers every connected app, and a run that gets no answers costs nothing. A study started from an agent also appears in the iMario web app, where you can explore it further. **Contact** - Sales: sales@imario.ai - Support: support@imario.ai - LinkedIn: https://www.linkedin.com/company/imario-ai - X: https://x.com/imario_ai ## Product - [Home](https://imario.ai): Synthetic audiences built from real people, from your own customer data and 61 markets. - [What are Synthetic Individuals?](https://imario.ai/synthetic-individuals): What a synthetic individual is, how iMario builds one, how accurate it is and what it is used for. - [Your customers: synthetic audience from your customer data](https://imario.ai/synthetic-audience): Turn CRM records, surveys and interviews into Synthetic Individuals your teams and agents can ask any time. - [Any audience: simulate the world population](https://imario.ai/simulate-world-population): 5.5 billion people across 61 markets, built from official census statistics and kept calibrated against real survey data, plus the official audiences. - [Discover: synthetic market research](https://imario.ai/synthetic-research): Map a market before there is a plan. Read segments, needs and buying motivations across thousands of people in an afternoon. - [Validate: concept and pricing testing](https://imario.ai/concept-testing): Put a price, a package or a concept in front of Synthetic Individuals built from your own customers. See who takes it, who trades down and where revenue peaks. - [Launch: creative pretesting](https://imario.ai/creative-testing): Show drafts of an ad, an email, a post or a public statement to the audience it is for, and read how each one lands and with whom before it goes public. - [Optimize: customer journey testing](https://imario.ai/optimize-customer-journey): Walk an audience through an onboarding, a program or a service flow step by step, and read where they stop and why. - [Service: support script and AI agent testing](https://imario.ai/ai-agent-testing): Run a support script against synthetic customers in every temperament, and find where it breaks and what it over-promises. - [Accuracy benchmark](https://imario.ai/synthetic-audience-benchmark): The method, datasets, raw files and both accuracy metrics, 1 minus total variation and 1 minus mean absolute error. - [Benchmark repository](https://github.com/laugherxiao/imario-benchmark): The public benchmark data and scoring code. - [Pricing](https://imario.ai/pricing): Credits, monthly packs, worked examples and Enterprise. - [Blog](https://imario.ai/blog): Product news, case studies and comparisons. ## Agent setup - [Setup sheet for agents](https://imario.ai/SKILL.md): Paste its address into Claude Code, Codex or Cursor to connect iMario and install the `imario` skill. - [iMario agent skill](https://github.com/laugherxiao/imario-skills): The `imario` skill, its reference files and install instructions. - [Full documentation text](https://imario.ai/llms-full.txt): Every current docs page in one markdown file. ## Docs: Getting Started - [What is iMario?](https://imario.ai/docs/getting-started/overview.md): Choose an audience, run research, and inspect the evidence behind a decision. - [Your First Study](https://imario.ai/docs/getting-started/quickstart.md): Ask one question of an existing audience, inspect its result, and choose a useful next step. - [Core Concepts](https://imario.ai/docs/getting-started/core-concepts.md): Understand audiences, individuals, tasks, runs, materials and reports. - [Is iMario Right for You?](https://imario.ai/docs/getting-started/is-imario-right-for-you.md): Decide when modelled audience research helps and when you need direct evidence from people. ## Docs: Research - [Choose a Research Method](https://imario.ai/docs/research/choose-a-method.md): Select the format that matches your question, stimulus and desired result. - [Ask a Single Question](https://imario.ai/docs/research/single-question.md): Run one focused question and decide what to investigate next. - [Run a Survey](https://imario.ai/docs/research/survey.md): Build or import a questionnaire, review its questions, and run it with a chosen audience. - [Run Interviews](https://imario.ai/docs/research/interview.md): Explore reasoning and follow-up questions with a discussion guide. - [Run a Focus Group](https://imario.ai/docs/research/focus-group.md): Hear simulated participants discuss topics together and inspect how the conversation develops. - [Compare Alternatives](https://imario.ai/docs/research/preference-test.md): Run a preference test on text or creative options with a consistent audience. - [Test a Creative](https://imario.ai/docs/research/creative-test.md): Show an image to an audience and explore what it communicates. - [Get Feedback on a Page](https://imario.ai/docs/research/url-feedback.md): Review a rendered page with an audience and distinguish page feedback from usability testing. - [Compare Audiences](https://imario.ai/docs/research/compare-audiences.md): Keep questions and materials consistent when comparing segments or markets. - [Run, Review and Follow Up](https://imario.ai/docs/research/run-and-follow-up.md): Check a run before submitting, follow its progress, and distinguish explanations from new research. ## Docs: Audiences - [Choose an Audience](https://imario.ai/docs/audiences/choose.md): Select an existing population, an official audience, or a segment built from your own data. - [Markets and Official Audiences](https://imario.ai/docs/audiences/catalog.md): Understand audience access, source coverage and creating a segment from an existing population. - [Build an Audience from Your Data](https://imario.ai/docs/audiences/build-from-data.md): Upload customer records and supporting evidence, review field matching, and create a reusable audience. - [Review and Update an Audience](https://imario.ai/docs/audiences/review-and-update.md): Check sources, population composition and the effect of syncing or rebuilding. - [Choose a Panel Size](https://imario.ai/docs/audiences/sample-size.md): Set respondent counts by research method, audience coverage and budget. ## Docs: Connect your data - [Connect a Data System](https://imario.ai/docs/synthetic-individuals/create/connect-data-system.md): Build an audience from your own data. Connect Zoom, Google Drive, Qualtrics or HubSpot, approve exactly what iMario reads, and import the recordings, surveys and files you pick. - [Connect Zoom](https://imario.ai/docs/synthetic-individuals/create/connect-zoom.md): Add the iMario app to Zoom, import the transcripts of the cloud recordings you pick, and remove the app and its data again. - [Connect Salesforce](https://imario.ai/docs/synthetic-individuals/create/connect-salesforce.md): Import contact records so an audience is built from your real customers. An admin creates a read-only External Client App once, inside your own org. - [Connect Qualtrics](https://imario.ai/docs/synthetic-individuals/create/connect-qualtrics.md): Import your Qualtrics survey responses so an audience answers from your own data. - [Connect Intercom](https://imario.ai/docs/synthetic-individuals/create/connect-intercom.md): Import the support conversations you pick from Intercom, so an audience answers in the words your customers used unprompted. - [Connect Zendesk](https://imario.ai/docs/synthetic-individuals/create/connect-zendesk.md): Import support conversations so an audience speaks in your customers' own unprompted words. - [Connect Feishu or Lark](https://imario.ai/docs/synthetic-individuals/create/connect-feishu.md): Feishu Minutes transcripts ground your audience in real customer calls. - [Connect Tencent Meeting](https://imario.ai/docs/synthetic-individuals/create/connect-tencent-meeting.md): Import the transcripts of your Tencent Meeting cloud recordings. ## Docs: Results & Reports - [From Results to Decisions](https://imario.ai/docs/run-and-analyze/overview.md): Read research results, explore their evidence, and decide whether to generate a report. - [Explore Results](https://imario.ai/docs/run-and-analyze/explorer.md): Select answers, inspect overlapping groups and break down results in Explorer. - [Generate and Read a Report](https://imario.ai/docs/run-and-analyze/reports.md): Choose completed runs, generate a synthesis, and trace findings back to responses. - [Share and Export](https://imario.ai/docs/run-and-analyze/share-and-export.md): Send a report or exported results with enough context for another person to interpret them. ## Docs: Materials & Individuals - [Materials & Connections](https://imario.ai/docs/materials/overview.md): Put each source in the right place: a research task, an audience, or an individual knowledge base. - [Use the Library](https://imario.ai/docs/materials/library.md): Create and reuse surveys, discussion guides, images, documents and reports. - [What is a Knowledge Base?](https://imario.ai/docs/knowledge-base/about.md): Provide reusable background to individuals without confusing it with audience evidence. - [Build a Knowledge Base](https://imario.ai/docs/knowledge-base/build.md): Add files, links or text and wait for processing before using the material. - [Manage a Knowledge Base](https://imario.ai/docs/knowledge-base/manage.md): Review sources, organize files and understand how updates affect connected individuals. - [About Synthetic Individuals](https://imario.ai/docs/synthetic-individuals/about.md): Inspect and interact with a modelled person while keeping their source and limitations clear. - [Work with an Individual](https://imario.ai/docs/synthetic-individuals/manage.md): Inspect a profile, supply context and follow up on a response. - [Inspect an Individual](https://imario.ai/docs/synthetic-individuals/manage/structure.md): Read identity, reflection, capabilities and memory. - [Connect Knowledge to an Individual](https://imario.ai/docs/synthetic-individuals/manage/connect-knowledge-bases.md): Supply relevant background and check how it supports later responses. - [Connect Agent Skills](https://imario.ai/docs/synthetic-individuals/manage/connect-agent-skills.md): Attach skills, authorize accounts and review proposed tool actions. - [Understand Memory](https://imario.ai/docs/synthetic-individuals/manage/memory.md): Inspect remembered context and its effect on later interactions. - [Chat with an Individual](https://imario.ai/docs/synthetic-individuals/manage/chat.md): Follow up on a response and inspect knowledge or tool activity. - [Individual Sources](https://imario.ai/docs/synthetic-individuals/create.md): Understand current audience creation and source-specific individual workflows. ## Docs: Methods & Evidence - [Interpret Evidence](https://imario.ai/docs/methods/evidence.md): Distinguish source data, modelled answers, respondent agreement and real-world validation. - [Read the Public Benchmark](https://imario.ai/docs/methods/benchmark.md): Understand what distribution matching, consistency and qualitative coding measure. - [Validate with Real Data](https://imario.ai/docs/methods/validate.md): Compare synthetic findings with observed responses while keeping the question and population aligned. ## Docs: Workspace & Credits - [Organize Your Workspace](https://imario.ai/docs/workspace/organization.md): Keep audiences, materials and research in the right workspace and project. - [Pricing & Credits](https://imario.ai/docs/workspace/pricing.md): Understand research rates, report generation, credit purchases and audience access. - [Track Usage and Purchases](https://imario.ai/docs/workspace/usage.md): Check the available balance and trace research, report and refund events. ## Docs: MCP & Agent Setup - [MCP & Agent Setup](https://imario.ai/docs/agent-setup/overview.md): Connect an external agent to iMario, install its companion skill, or add capabilities to an individual. - [Connect the iMario MCP Server](https://imario.ai/docs/agent-setup/mcp.md): Approve a connection or create a workspace key, configure a client and verify access before running research. - [Install the iMario Agent Skill](https://imario.ai/docs/agent-setup/install-skill.md): Add the companion research instructions to an external agent after connecting MCP. ## Docs: Playbooks - [Test an Ad or Copy](https://imario.ai/docs/playbooks/test-ad-or-copy.md): Compare messages and inspect the reasons behind a preference. - [Explore a Pricing Decision](https://imario.ai/docs/playbooks/pricing-study.md): Investigate price objections and packaging before testing real purchasing behavior. - [Review an Onboarding Flow](https://imario.ai/docs/playbooks/validate-onboarding.md): Find comprehension problems in onboarding materials before testing with real users. - [Compare a Concept Across Markets](https://imario.ai/docs/playbooks/cross-market-concept-testing.md): Keep a concept consistent while examining differences across market audiences. - [Explore Hard-to-Access Audiences](https://imario.ai/docs/playbooks/reach-hard-to-access-audiences.md): Use simulated answers, and the data behind them, to prepare better questions for people who are hard to recruit. - [Rehearse a Sales Conversation](https://imario.ai/docs/playbooks/rehearse-sales-conversation.md): Practice discovery and objections against a modelled buyer perspective. - [Work with a Synthetic Colleague](https://imario.ai/docs/playbooks/synthetic-employee.md): Combine an individual perspective, background material and approved tool use. ## Docs: Help - [Troubleshooting](https://imario.ai/docs/resources/troubleshooting.md): Resolve missing inputs, inaccessible pages, stalled runs and unexpected results. - [Frequently Asked Questions](https://imario.ai/docs/resources/faq.md): Find current answers about research, audiences, reports and costs. - [Support](https://imario.ai/docs/resources/support.md): Contact the team with the context needed to investigate a problem. - [Send Feedback](https://imario.ai/docs/resources/feedback.md): Report an issue or describe the workflow a product change should improve. - [Documentation Updates](https://imario.ai/docs/resources/changelog.md): Track this manual separately from product release announcements. ## Blog - [iMario 2.0: The Human Voice Behind Every Decision](https://imario.ai/blog/introducing-imario-2-0): Published 2026-09-15. The 2.0 release: audiences from the customer data you already own, 5.4 billion simulated people across 59 markets, iMario inside ChatGPT, Claude and Codex, and usage-only pricing. - [How Vitasoy Tests New Ideas with Synthetic Consumers](https://imario.ai/blog/case-vitasoy-test-new-products-before-launch): Published 2026-08-27. Case study: earlier feedback on flavor concepts, packaging and new products from historical customer knowledge and interviews. - [How Rand and Rand Tests Agendas with Synthetic Audiences](https://imario.ai/blog/case-rand-and-rand-test-event-agendas-before-launch): Published 2026-08-27. Case study: pressure-testing event agendas with simulated attendees and sponsors. - [How IFOP Combines Synthetic Consumers with Traditional Research](https://imario.ai/blog/case-ifop-combine-synthetic-consumers-with-traditional-research): Published 2026-08-27. Case study: extending traditional research methods with synthetic consumers. - [A Men's Grooming Brand's 3 Launch Decisions in 7 Days](https://imario.ai/blog/case-mens-beauty-three-decisions-one-week): Published 2026-05-15. Case study: packaging, claim and pricing studies on one synthetic audience. - [MatrAIx's 8.3 Billion Personas: Substance or Sizzle?](https://imario.ai/blog/matraix-8-3-billion-personas-substance-or-sizzle): Published 2026-08-10. A line-by-line reading of a claimed simulation of the world's population. - [The 10 Best AI Focus Group Tools for 2026](https://imario.ai/blog/best-ai-focus-group-tools-2026): Published 2026-06-15. Synthetic-panel, AI-moderated and real-human focus group tools compared. - [The 8 Best SyntheticUsers Alternatives for 2026](https://imario.ai/blog/synthetic-users-alternatives-2026): Published 2026-06-09. Synthetic-respondent, AI-moderated and real-human research tools compared. - [iMario vs Delve.ai](https://imario.ai/blog/imario-vs-delve): Published 2026-05-08. A market research capability comparison. - [iMario vs SyntheticUsers.com](https://imario.ai/blog/imario-vs-syntheticusers): Published 2026-05-06. A capability comparison. - [iMario vs Aaru](https://imario.ai/blog/imario-vs-aaru-synthetic-research): Published 2026-03-25. Two paths to human behavior simulation. ## Optional - [No One Can Predict Anyone, Not Even You](https://imario.ai/blog/no-one-can-predict-human-behavior): Published 2026-06-01. The say-do gap, and why you can forecast a crowd but not a person. - [Five AI Shifts in Product Development](https://imario.ai/blog/five-ai-shifts-in-product-development): Published 2026-05-18. A reading of McKinsey's piece on AI in the product development lifecycle. - [The Five Layers Behind an iMario Research Report](https://imario.ai/blog/five-layers-behind-imario-research-reports): Published 2026-04-28. How thematic analysis and consulting-style findings combine in a report. - [Introducing iMario 1.0](https://imario.ai/blog/introducing-imario-1-0): Published 2026-04-13. The 1.0 release, since replaced by 2.0. - [iMario vs Base LLMs](https://imario.ai/blog/imario-vs-llms): Published 2026-03-23. Mode collapse and identity drift in base LLMs compared with iMario. - [Harness Engineering Is the Missing Layer Between Models and Products](https://imario.ai/blog/harness-engineering-is-the-missing-layer-between-models-and-products): Published 2026-03-16. Pipeline design, validation loops and stateful orchestration. - [Synthetic Users vs. Traditional Research](https://imario.ai/blog/synthetic-users-vs-traditional-research-roi-2026): Published 2026-02-23. A cost, time and learning-speed framework. - [Beyond "Average": Building a Million Synthetic Individuals](https://imario.ai/blog/beyond-average-how-imario-is-building-a-million-synthetic-individuals): Published 2026-02-09. Why average personas fail in simulation. - [What Are Synthetic Individuals and Why They Can Represent Real Humans](https://imario.ai/blog/what-are-synthetic-individuals-and-why-they-can-represent-real-humans): Published 2026-01-26. An explainer, and how they differ from role-play. - [The Human API: Why We Are Building iMario.ai](https://imario.ai/blog/the-human-api-why-we-are-building-imario-ai): Published 2026-01-19. Why human understanding is the bottleneck, and why iMario exists.