--- title: "Panels and Methodology FAQ | Minds" canonical_url: "https://getminds.ai/faq/panels" last_updated: "2026-06-24T15:39:47.902Z" meta: description: "Customer, client, user, and expert panels. How to build them, how big they should be, how to ask questions, how to read aggregated results." "og:description": "Customer, client, user, and expert panels. How to build them, how big they should be, how to ask questions, how to read aggregated results." "og:title": "Panels and Methodology FAQ | Minds" "twitter:description": "Customer, client, user, and expert panels. How to build them, how big they should be, how to ask questions, how to read aggregated results." "twitter:title": "Panels and Methodology FAQ | Minds" --- Minds May 9, 2026·Faq·Minds Team # **Panels and Methodology FAQ** Customer, client, user, and expert panels. How to build them, how big they should be, how to ask questions, how to read aggregated results. Everything about how panels work in Minds. For a deeper walk-through, see the [Guide on Panels](https://getminds.ai/guide/panels) and the blog post on [AI focus groups](https://getminds.ai/blog/ai-focus-group). ## What a panel is ### What is an AI panel? An AI panel is a group of AI personas queried together. You ask one question, all personas respond in parallel, and the platform aggregates the answers into: - _Scale ratings_ (1 to 10): distribution charts and group averages - _Categorical_ (yes/no, multiple choice): percentage breakdowns - _Qualitative_ (open-ended): clustered themes Panel sizes typically run 8 to 100 personas depending on the question and the confidence you need. ### What are the four panel types? | Panel type | Who's in it | Used by | | --- | --- | --- | | Customer Panel | Your target customers | Marketing, product, founders | | Client Insight Panel | Your client's customers | Agencies, consultants | | User Panel | Your product users | Product teams, UX | | Expert Panel | Domain experts (CMO, VC, engineer, lawyer) | Founders, strategists, anyone needing senior advice | Mechanically all four work the same way; the difference is who you put in the panel. ### What is a customer panel? A panel of personas representing your target customer segments. Used to test campaigns, messaging, pricing, positioning, product concepts, ad creative, landing pages. See [AI customer panels](https://getminds.ai/blog/ai-customer-panels). ### What is a client insight panel? A panel of personas representing your client's customers, used by agencies and consultants. Build once per client, reuse across briefs. See the blog [adding market research to agency retainer using AI](https://getminds.ai/blog/adding-market-research-to-agency-retainer-using-ai). ### What is a user panel? A panel of personas representing your product's actual users. Used for UX research, feature pre-testing, onboarding flow validation, churn diagnosis. See [AI user research](https://getminds.ai/blog/ai-user-research). ### What is an expert panel? A panel of domain experts (CMOs, VCs, engineers, lawyers, designers). Used when you need senior advice, perspective, or a sanity check from someone who's seen the pattern before. See [AI expert panel](https://getminds.ai/blog/ai-expert-panel) and [the AI advisor](https://getminds.ai/blog/the-ai-advisor). ## Sizing ### How big should a panel be? Use case driven: - _8 to 15_ personas: fast directional read, "does this hook land?" - _30 to 50_ personas: confident segmentation work, "do urban vs suburban diverge here?" - _50 to 100_ personas: quantitative-style distributions, "what's the price elasticity?" For most decisions, 15 is plenty. Larger panels cost more and take longer; the marginal value of the 50th persona is small. ### Why not just one persona? One persona is fine for a quick sanity check or a one-on-one expert deep-dive. For research, you want a distribution. The interesting answer is rarely "everyone agreed"; it's "here's where they split, and here's why." ### How long does a panel run take? A 15-Mind panel responds in roughly 1 to 3 minutes. A 100-Mind panel takes longer (5 to 10 minutes typical). The platform shows responses as they arrive; you don't wait for the slowest persona to start reading. ## Building ### How do I build a panel? Two paths: 1. _Pick existing Minds_ from your library 2. _Describe the audience_ in plain text ("enterprise CTOs in San Francisco, SaaS, 500+ employees, currently evaluating Snowflake alternatives") and let Minds generate a representative panel for you You can edit, swap, or remove individual Minds before running queries. Panels are reusable; you build a panel once and query it for years. ### Can I import my existing personas? Yes. Paste in a persona document, upload a PDF, drop in interview transcripts, or feed in a CRM export. Minds builds a usable Mind from any reasonable source. See [import customer panels into Minds](https://getminds.ai/blog/import-customer-panels-into-minds). ### Can I import LinkedIn profiles? Yes. Paste a LinkedIn URL when creating a Mind. The Mind absorbs the public profile data and structures it through the personality model. See [LinkedIn customer profile to AI persona](https://getminds.ai/blog/linkedin-customer-profile-to-ai-persona). ### Do I need to update panels over time? For long-running research programs, yes. Refresh personas when your target market shifts (new competitor enters, regulatory change, generational handoff). For one-shot pre-testing, the panel you build today is fine to use today. ## Asking questions ### How do I ask a good panel question? Show specific stimulus, not abstract framing. _Bad:_ "What do you think about CTAs in B2B SaaS?" _Good:_ "Here's our pricing page. What stops you from clicking 'Talk to sales'?" _Bad:_ "Is our messaging clear?" _Good:_ "Read this homepage hero. In your own words, what does this product do?" Specific stimulus produces specific, useful responses. Abstract framing produces abstract, useless responses. ### Can I direct individual personas? Yes. Use `@name` in the chat to address a specific Mind. `@Sarah what do you think about this positioning?` returns Sarah only. Without a mention, everyone in the chat responds. ### What kind of stimulus can I show a panel? Anything visual or textual: - Landing page screenshots - Pitch decks and PDFs - Product images, packshots, mocks - Competitor ads, campaign creative - Interview notes, raw transcripts - Pricing pages, ad copy, email drafts - Short videos (typically up to a few minutes) ### Can I run two panels side-by-side? Yes. Add two Groups to the same chat to compare segments (Gen Z vs Millennials, US vs Germany, free users vs paying users). The chat shows responses from both panels, side-by-side, with percentage and distribution breakdowns per segment. ## Reading results ### How are panel responses aggregated? Three ways depending on the question type: 1. _Scale (1 to 10):_ distribution histogram + group average 2. _Categorical (yes/no, multi-choice):_ percentage breakdowns 3. _Qualitative (open):_ clustered themes with representative quotes You can drill into any individual Mind's response from the aggregate view. ### What is the Alignment score on a panel answer? Every panel answer has an _Alignment_ dropdown in the header, with a 0–100% score per group: - _High_ (67%+) — segment answered consistently with its persona definitions - _Medium_ (34–66%) — mixed; worth reading the individual responses - _Low_ (under 34%) — read every response before acting on the aggregate The score is the average response reliability of the Minds in that group for that specific question. Each Mind's answer is re-scored against its own persona definition (how on-character it was); we average per group. It is a temporary metric. The proper group-alignment model — closeness to empirical research findings for that segment — is in development. ### Why does the Alignment score load after the chart? Alignment is computed after each Mind's answer is generated. The chart renders first so you see the panel result immediately; the Alignment dropdown shows a loading state on each row for a few seconds while the per-Mind scores arrive, then fills in the per-group average. For messages older than this feature (or for the v1 API responses where Alignment is computed inline before return), the dropdown shows the score immediately on open. ### Can I export panel results? Yes. Export to CSV, PDF, or share via a public link. Useful for client decks, internal summaries, and sharing with stakeholders. ### Can I share a panel result publicly? Yes. Toggle link sharing on a panel and share the URL with prospects, clients, or partners. Use the share for sales calls, agency pitches, or as a teaser for a deeper engagement. ## Related - [Synthetic research basics FAQ](https://getminds.ai/faq/synthetic-research) - [Research methods FAQ](https://getminds.ai/faq/research-methods) - [Comparisons FAQ](https://getminds.ai/faq/comparisons) - Guide: [Panels](https://getminds.ai/guide/panels) - Blog: [AI focus group](https://getminds.ai/blog/ai-focus-group) - Blog: [AI expert panel](https://getminds.ai/blog/ai-expert-panel) ## **Frequently asked questions** ### **What is an AI panel?** An AI panel is a group of AI personas queried together. You ask one question, all personas respond in parallel, and the platform aggregates the answers into distributions (scale ratings), percentage breakdowns (multiple choice), and clustered themes (open-ended). Panel sizes typically run 8 to 100 personas depending on the question. ### **What are the four panel types?** Customer Panels (your target customers), Client Insight Panels (your client's customers, used by agencies), User Panels (your product users for UX and feature work), and Expert Panels (domain experts like CMOs, VCs, engineers). Each is built the same way; the difference is who's in the room. ### **How big should a panel be?** 8 to 15 for fast directional reads. 30 to 50 for confident segmentation work. 50 to 100 for quantitative-style distributions where you want statistical-feeling breakdowns. Larger panels cost more and take longer; for most decisions, 15 is plenty. ### **How do I build a panel?** Two paths. Pick existing Minds from your library, or describe the audience in plain text ("enterprise CTOs in San Francisco, SaaS, 500+ employees") and let Minds generate a representative panel for you. You can edit, swap, or remove individual Minds before running queries. ### **How do I ask a good panel question?** Show specific stimulus, not abstract framing. Drop in the actual ad creative, landing page screenshot, pricing page, or pitch deck. Ask "what stops you from clicking the CTA?" instead of "what do you think about CTAs?" Specific stimulus produces specific, useful responses. ### **Can I run two panels side-by-side?** Yes. Add two Groups to the same chat to compare segments (Gen Z vs Millennials, US vs Germany, free users vs paying users). The chat shows responses from both panels, side-by-side, with percentage and distribution breakdowns per segment. ### **What is the Alignment score on a panel answer?** Every panel answer has an Alignment dropdown showing a 0-100% score per group, labelled High (67+), Medium (34-66), or Low (under 34). It is the average response reliability of the Minds in that group for that question — a measure of how on-character each Mind's answer was against its own persona definition. High alignment means the segment answered consistently with its persona; lower alignment is a signal to read the individual responses before acting on the aggregate. This is a temporary stand-in for a future group-alignment metric that scores closeness to empirical research findings. ### **Why does the Alignment score load after the chart?** Alignment is computed after each Mind's answer is generated, by re-scoring every response against its persona. The chart renders first so you see the panel result immediately; the Alignment dropdown shows a loading state on each row for a few seconds while the per-Mind scores arrive, then fills in the per-group average.